An Assessment of Energy Resources For Global Decarbonisation PDF

4CMR Working Paper Series
An assessment of energy resources for global decarbonisation

Jean-Francois Mercure, Pablo Salas
Cambridge Centre for Climate Change Mitigation Research (4CMR), Department of Land Economy, University of Cambridge, 19 Silver Street,
Cambridge, CB3 1EP, United Kingdom
Abstract
This paper presents an assessment of global economic energy potentials for all major natural energy resources.
This work is based on both an extensive literature review and calculations based onto natural resource assessment
data. In the first part, economic potentials are presented in the form of cost-supply curves, in terms of energy flows
for renewable energy sources, or fixed amounts for fossil and nuclear resources, using consistent energy units that
allow direct comparisons to be made. These calculations take into account, and provide a theoretical framework for
considering uncertainty in resource assessments, providing a novel contribution aimed at enabling the introduction
of uncertainty into resource limitations used in energy modelling. The theoretical details and parameters provided in
tables enable this extensive natural resource database to be adapted to any modelling framework for energy systems.
The second part of this paper uses these cost-supply curves in order to build a tool for analysing global scenarios
of energy use, in the context of exploring the feasibility global decarbonisation using renewable energy sources. For
such a purpose, a theoretical framework is given for evaluating either flows of stock energy resources for given price
path assumptions for the related energy carriers, or the prices of energy carriers given energy demand assumptions.
Results of both approaches are used in order to produce a complete comparison of global energy resources. The
particular case of the feasibility of global decarbonisation by the end of the century is explored. Since the scale of the
required amount of energy flows from renewables is comparable to the sum of the technical potentials, the associated
scale of global land use for energy production is found to be large. For complete decarbonisation, without energy
demand reductions, 7 to 12% of the global land area could be required for energy production activities, emphasising
the importance of improving energy consumption patterns and intensity of the global economy.
The third part of this work is an appendix that provides all missing details, equations and databases necessary to
understand and reproduce the work of Part I. This part is therefore aimed at enabling energy modellers to reproduce
exactly and use in their own work the database that was constructed in this work.
Keywords: Global energy resources, Climate change mitigation, Energy resources, Global Decarbonisation
Corresponding
Pablo
author: Jean-Franc ois Mercure, email address: jm801@cam.ac.uk

Salas, email address: pas80@cam.ac.uk
4CMR Working Paper no 2
September 25, 2013
Cambridge Centre for Climate Change Mitigation Research (4CMR)
Contents
I Energy Resources
1 Introduction
2 Definitions
3 Methodology
3.1 Economic potentials .
3.2 Cost-supply curves .
3.3 Distributions . . . .
3.4 Uncertainty . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6
6
7
8
10
4 Renewable energy resources

4.1 Wind energy . . . . . .
4.2 Solar energy . . . . . .
4.3 Hydroelectricity . . . . .
4.4 Geothermal energy . . .
4.5 Bioenergy . . . . . . . .
4.6 Ocean energy sources . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
11
13
14
15
15
17
5 Stock resources
5.1 Fossil fuels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Fissile materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
17
20
6 Summary of energy resources
21
7 Conclusion
21
II Global Decarbonisation
23
8 Introduction
23
.
.
.
.
9 Methodology and definitions

9.1 The economic potential of natural resources . .
9.2 Scenarios for the exploitation of stock resources
9.3 Calculation of energy flows from stocks . . . .
9.4 Constant global production to reserve ratios . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
24
24
25
26
28
10 Global overview of stock energy resources

10.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.2 Flows from stock energy sources for exogenous prices . . . . . .
10.3 Price paths for exogenous flows . . . . . . . . . . . . . . . . . .
10.4 Real systems operate somewhere between these limiting situations
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
29
29
30
31
32
11 Supply and demand for energy resources: the feasibility of global decarbonisation
11.1 Global flows of secondary energy carriers from all sources . . . . . . . . . . . . . . . . . . . . . . .
11.2 Ranges of total primary energy demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11.3 The complete set of cost-supply and cost-flow curves as a tool for energy planning . . . . . . . . . .
11.4 The decarbonisation of the global economy before 2100 . . . . . . . . . . . . . . . . . . . . . . . . .
11.5 The effect decarbonising the current global energy system and the associated strain on the environment
35
35
36
36
38
39
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12 Conclusion
40
III Appendix
42
13 Introduction and use of the Appendix
42
14 Derivation of distribution functions and cost-supply curves

14.1 Distribution function for the hierarchical type of resources . . . . . .
14.2 Distribution function for nearly identical resources . . . . . . . . . .
14.3 Cost-supply curve expressions . . . . . . . . . . . . . . . . . . . . .
14.4 Parameterisation formulas . . . . . . . . . . . . . . . . . . . . . . .
14.5 Demonstrating the validity of the functional forms using IMAGE data
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
42
42
43
44
45
45
15 Cost-supply curve calculation methodology per resource type

15.1 Definition of world regions . . . . . . . . . . . . . . . . .
15.2 Wind and solar energy . . . . . . . . . . . . . . . . . . .
15.3 Hydropower . . . . . . . . . . . . . . . . . . . . . . . . .
15.4 Geothermal energy . . . . . . . . . . . . . . . . . . . . .
15.5 Bioenergy . . . . . . . . . . . . . . . . . . . . . . . . . .
15.6 Ocean energy . . . . . . . . . . . . . . . . . . . . . . . .
15.7 Oil . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15.8 Natural gas . . . . . . . . . . . . . . . . . . . . . . . . .
15.9 Coal . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15.10Uranium . . . . . . . . . . . . . . . . . . . . . . . . . . .
15.11Thorium . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
45
45
46
46
47
48
48
49
50
50
50
51
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
16 Data tables and figures
52
Acknowledgements
59
Bibliography
59
Part I
Energy Resources
1. Introduction
Energy policy decisions for the planning of new energy generation capacity designed to respond to future demand
require information regarding the engineering feasibility and the requirements of such systems in terms of capital
investments, but also, crucially, the availability of natural resources. Meanwhile, future energy systems are expected,
in many contemporary policy frameworks, to evolve towards their gradual decarbonisation, in order to decrease anthropogenic interference with the climate system (see for instance Edenhofer et al. [1]). From an energy perspective,
the decarbonisation of the sector involves a transfer of supply from traditional fossil fuel based technologies towards
low GHG emission energy generation capacity such as renewable energy systems or nuclear reactors. Such a transfer
requires changes in the technologies used through substitution processes, a subject extensively studied in the past (for
instance [24]). However, these transformations also require changes in the use of primary energy sources. Realistic
energy scenarios of the future can only be designed in a way that does not exceed natural sources and flows of energy
which are available in all regions of the world. Therefore, assessments of the potential of natural energy resources are
essential to energy planning and policy.
Meanwhile, many models exist that generate scenarios for the evolution of global energy systems in the future
(For a recent brief review of IMAGE/TIMER, MERGE, E3MG, POLES and REMIND see Edenhofer et al. [1]).
Such models, in order to generate scenarios that are realisable, must take into account the limits to each type of
natural resource. However, while some energy models do not currently take explicit account of limits to resource
flows, most of them do not consider their associated uncertainty. Of particular interest in this work is the EnergyEnvironment-Economy Model at the Global level (E3MG), a macroeconometric model of the global economy, which
calculates economic activity in 42 industrial sectors within 20 regions of the world [59]. In a previous paper, a new
sub-component designed for use in E3MG was introduced, modelling the trend of investor behaviour in the power
sector facing the choice of various technologies, using their Levelised Cost Of Electricity production (LCOE), which
includes the cost of natural resources, and the resulting technological substitutions [4, 10]. In this model, the limitation
of natural resources occurs through the use of cost-supply curves, requiring a complete set for each type of natural
resource considered in order to properly constrain the model.
Many studies and reviews have been published that summarise what was known of global energy potentials at
time of their publication [1117]. These reports however do not provide any economic structure to energy potentials.
Without underlying individual cost structures associated with energy resource extraction or clarifications the use of the
concept of economic potentials, natural resource limit values are ambiguous. As the key for strategic energy planning
lies precisely with the cost structure of energy production, such assessments are of limited use for energy systems
modelling and policy-making without additional assumptions over their cost distributions.
Every energy source is limited, either in its total amount that can be consumed, for stock resources, or in the
total energy flow it can produce at any one time, for renewable resources. Resources tend to be exploited in order of
their cost of extraction. As consumption gradually progresses to higher and higher levels of exploitation of particular
resources, additional units consumed tend to incur increases in production cost. Therefore, the economic potential is
better defined as a f unction of cost, rather than as a constant value. As the costs of production increase with the levels
of use, developers increasingly seek alternatives, where they exist, through an evaluation of the opportunity cost.
Thus, the economic potentials for all energy resources available within a particular market depend onto one another
and cannot be determined individually without knowing the alternatives. Additionally, since every particular market
for energy is composed in a particular way, economic potentials vary geographically. As resource use evolves and
depletion progresses, all economic potential values in a market change with time, an effect that stems from changing
individual costs of energy production, and and that has repercussions onto the rate of consumption of every natural
resource.
As discussed in previous work, the economics of energy resources when used, for instance, for electricity production can be modelled in a simplified manner using a complete ensemble of cost-supply curves, which express the
cost of resources at various levels of exploitation [4]. In such a framework, the marginal cost of electricity production
for every individual natural energy resource using specific power technologies may be compared using a framework
such as the LCOE at every level of natural resource use, in order to compare their profitability. In such a model, the
consumption of particular resources is limited by increasing marginal costs, and the potential depletion of resources
becomes naturally represented since the cost gradually diverges at a levels of exploitation closer and closer to the
technical potential. Cost-supply curves for global resources of wind, solar and biomass energy have been calculated
previously using the land use model IMAGE 2.2 [1822], and are used in this way in the TIMER energy sub-model.
Additionally, global cost curves for fossil fuels have been produced by Rogner [23], an influential work which is
unfortunately becoming increasingly outdated. However, no comprehensive global assessment of all major energy
resources which provides an underlying cost structure currently exists in the literature.
Additionally, assessments of natural energy resources inherently possess high uncertainties, which must be taken
into account in order to generate confidence levels in model outputs. For instance, the global bioenergy potential has
been estimated to lie between around 310 to 660 EJ/y, by Hoogwijk et al. [20], between 0 and 650 EJ/y by Wolf
et al. [24] and between 370 and 1550 EJ/y by Smeets et al. [25]. These particular uncertainties stem from those
on future projections of food demand and levels of technology advancement in the agricultural sector used in these
calculations. These types of uncertainties are present in all assessments of renewable energy resources. Similarly,
uncertainty arises with knowledge on amounts of stock resources, such as uranium and fossil fuels, where lower and
lower levels of confidence are associated with larger and larger quantities. These uncertainties originate directly from
the cumulative amount of effort that has been deployed to explore geological occurrences, and express the fact that it
is never possible to know with certainty the detailed composition of the crust of the Earth. Using a review of literature,
it is possible, and appropriate, to define ranges of energy potentials rather than specific values, therefore defining areas
in the cost-quantity plane for where actual cost-supply curves are likely to lie.
The first part of this work proposes a theoretical framework and a computational methodology for building natural
resource assessments readily useable in models of energy systems, by using a combination of cost-supply curves and a
treatment of uncertainty. This methodology is then applied to produce a cost-quantity analysis for every major natural
energy resource, those with a potential larger than 10 EJ/y. As part of this work, cost-supply curves were produced
for 13 types of resources for every one of the 20 world regions specified in E3MG, and form a new sub-model
for natural resource use and depletion. Underlying potentials were however defined for 190 individual countries,
and can be aggregated to any other particular set of world regions. For the sake of presentation in this paper, the
cost-supply curves were aggregated into global curves, providing a world view. Since data for all 190 countries
could not realistically be provided here, tables of parameters are provided in the supplementary material that enable
to reconstruct the cost-supply curves for a set of 14 world regions that were assumed the most representative of the
requirements of the international modelling community. Additional resource specific information regarding theoretical
derivations, additional methodology and justifications are also provided.
Part I of this work follows consistently a methodology that can be reused as presented by the modelling community.
To this end, definitions of the concepts used are first given followed by a concise description of the approach, detailing
the cost-quantity analysis of resources with the associated uncertainty. Following this, a theoretical characterisation of
the statistical properties of resource occurrences is given in order to enable the use of functional forms for interpolating
resource data. This methodology is then used to produce cost-supply curves for renewables and stock resources. A
summary of all major energy resources is given in the last section.
2. Definitions
Renewable and stock energy resources: Natural sources of energy that may be found in one of two forms: stocks,
where energy may be extracted from fixed amounts of geologically occurring materials with specific calorific
contents; renewable f lows, where energy may be extracted from continuously producing onshore or offshore
surface areas with wind, solar irradiation, plant growth, river flows, waves, tides or various forms of heat flows.
Theoretical Potential: Total quantity of energy stock or flow estimated to exist or stem from a particular natural
process, disregarding it technical recoverability.
Technical potential: Total quantity of energy stock or flow estimated to exist or stem from a particular natural process, recoverable using a specific technique, disregarding its level of technical difficulty and the associated
costs.
Economic potential: Quantity of energy stock or flow estimated to exist or stem from a particular natural process,
recoverable at exploitation costs that are competitive compared to all other alternative ways of producing the
same energy carrier. Since cost value considered competitive at any one time changes continuously, the economic potential is expressed as a quantity of stock or flow of energy function of cost. However, this concept is
used more conveniently in models when expressed as the inverse of this function, the cost-supply curve.
Cost-supply curve: Function of the cost of energy flow or stock from a particular resource given that a certain
quantity is already in exploitation or has already been consumed (the marginal cost). In this work, the costsupply curve and the economic potential used interchangeably.
Uncertainty range: Area in the cost-quantity or cost-flow planes in which actual real cost-supply curves have a 96%
probability of lying. This would correspond to two standard deviations, 2, if the distributions were normal,
but they are in general skewed. Real values have a 2% probability of occurring below the range, and a 2%
probability of lying above.1
Productivity: Amount of energy stock or flow produced per unit of land or sea surface area (wind, solar, biomass,
wave energy), bore depth or digging effort (oil, gas, coal, nuclear fuels, geothermal energy) or construction
effort (hydroelectricity, tidal energy).
Hierarchical resource distribution: Statistical type of natural resource productivity distribution, with productivity
values that strongly depend on the number of simultaneous positively contributing physical factors. Resource
producing items of this statistical type within one kind of resource (windy areas, rivers, mines, wells, etc)
possess widely different productivities which enable their ranking in order of resource quality.
Distribution for nearly identical resources: Statistical type of natural resource productivity distribution in which
resource producing items possess nearly identical productivity values, which do not depend on the simultaneous
occurrence of several factors. Producing items of this type within one kind of resource (for instance sunny or
fertile plots of land) have nearly identical properties, cannot be ranked and can be exchanged for one another.
3. Methodology
3.1. Economic potentials
Natural occurrences of energy resources are found in different forms, with varying productivity levels or require
various levels of effort for their extraction, which enable their transformation into usable energy carriers with different
levels of profitability. These variations together lead to particular distributions of costs for their utilisation. Naturally,
resources with the best qualities for energy production, and thus lowest extraction costs, are likely to be considered
first by energy firms under financial constraints. Therefore, deriving economic potentials for energy resources involves
the task of classifying and ranking different occurrences of specific resources in order of cost.
Information on energy resources is scarce and irregularly distributed, possibly inconsistent, and thus must be
organised and classified in order to produce a consistent and complete set of economic potentials. Data may be
patchy and incomplete, in which case assumptions are required in order to interpolate through missing parts. Such
assumptions are taken in this work in the form of functional forms for the ranking of resources in terms of their cost
of extraction. These are derived theoretically from basic statistical properties of resources. They have been carefully
verified against several sets of data for specific types of natural resources which do not take any assumptions over
the distribution of resources (wind, solar, two types of biomass resources as well as with uranium). They have been
assumed to hold true for all other types of resources (for fossil, geothermal, hydroelectric and ocean resources).
Methods of assessment differ and produce different results or ranges of results. In the absence of justified criteria
onto which to base a choice of particular studies over all others, resource assessments must be considered equally,
the collection of which can be used to generate uncertainty ranges. This allows to decouple this work from specific
assumptions used in specific studies.
2 probability range correspond to erf( 2) = 95.45%, yielding 2.28% as a probability of values occurring above or below the range. The
values of 96% and 2% are used instead for convenience.
1 The
Cost Distribution
0
C0
b)
Cost
Technical Potential
Cumulative Distribution
Upper Range
Probable Range
Lower Range
a)
Cost
C0
c)
d)
2%
C0
0
Quantity
Cost
Technical Potential
Cost
96%
2%
C0
0
Quantity
Figure 1: a) Sketch of a hypothetical distribution of cost ranked amounts of energy or energy flow units available in various cost ranges. The
uncertainty over the amount available in each cost range is indicated with a colour shading: the top of the dark histogram represents the minimum
amount which has a 98% probability of being exceeded while the top of the white histogram indicates the maximum amount associated with a 2%
probability of being exceeded. The most likely quantity is intermediate, represented by the top of the grey histogram. b) Cumulative distribution
of energy resources, with uncertainty shown as vertical error bars. c) The marginal cost, or cost of extracting an additional unit of energy stock or
flow given that a certain quantity has already been exploited, commonly called the cost-supply curve, with uncertainty represented as horizontal
error bars. d) Cost-supply curve defined as a probability distribution, where the red curves indicate the limits of a 96% confidence level region in
the cost-quantity plane, while the blue curve corresponds to the most probable cost-supply curve. The assumption is therefore taken that there is a
2% chance that the cost-supply curve lies below the upper boundary, and a 2% chance that it lies below the lower boundary.
The methodology presented here builds upon the approach defined in our earlier work [4]. The economic potential
of resources is defined using the cost-supply curve, which expresses the quantity of resource available for any cost
value considered economic, or competitive with all other alternatives. Such curves are derived from cost rankings of
resources and resulting distributions. The cost variable, however, stems from varying levels of technical difficulty for
extracting resources, or alternatively, the productivity of energy producing resources such as plots of land, mines, oil
wells, rivers, etc. Therefore, continuous distribution functions for the amounts of resources available in nature are
defined in terms of their productivity. Two empirical forms for these distributions are defined and used throughout.
Confidence ranges are derived from uncertainty analysis. The combination of both is used to construct probability
densities for the location in the cost-quantity or cost-flow planes where the real cost-supply curves would be situated if
it were possible to determine them with certainty.2 These probability distributions may be used as inputs to uncertainty
analysis (such as Monte Carlo simulations) in energy systems modelling.
3.2. Cost-supply curves
The ranking of resources in terms of their productivity, required for building cost-supply curves, can be done using
a set of histograms of the quantity of energy stocks of flows that can be obtained within various ranges of productivity
2 Note that the use of uncertainty ranges in the cost-quantity/cost-flow plane relaxes the constraints of using specific functional forms, since it
allows variations in the particular forms of the functional dependences within the ranges.
values. The productivity variables may be converted into costs, which result in a new set of histograms representing
the amount of energy that can be produced at costs within various cost ranges. This is shown in Figure 1 a), with a
typical distribution of energy resources, which decreases at low cost values due to the decreasing number of resources
of exceptional quality, and to high cost values due to the decreasing productivity or recoverability of the resources.
The shading is a representation of the confidence associated to their potential availability. The top of the dark grey
distribution shows the lowest quantity of assured resources, assumed to be exceeded with a probability of 98%. The
top of the white histogram represents the upper range of speculative resources, those assumed to be exceeded with a
probability of 2%. The quantity which is the most likely to be available lies between these extremes, shown with the
top of the grey histogram.
The amounts in each cost range thus possess an uncertainty.3 In order to determine the quantity that can be obtained
at or below certain cost values, the cumulative distribution function is calculated, shown in panel b). This sum
converges towards the technical potential. The uncertainty increases approximatively cumulatively with increasing
cost values through the root of the cumulative sum of the squares of the individual uncertainty values, shown with
blue error bars.
The marginal cost of resources, or the cost of extracting an additional unit of resource given that a certain quantity
has already been used, corresponds to the inverse of the cumulative sum, shown in panel c). Thus, the cost of additional
units diverges when the number of units used approaches the technical potential, at the point of resource depletion.
Using the uncertainty ranges, or error bars, to define two additional curves, assumed to define the upper and lower
96% confidence limits, a probability density can be defined in cost-quantity space for the location where the real
cost-supply curve would lie if it were possible to know it with certainty. This is shown in panel d), where the red
curves define the uncertainty area and the blue curve represents the most probable of all possible cost-supply curves,
the mode of the distribution. Such probability densities are normally skewed, since the uncertainty over undiscovered
resources lies at higher quantity values.4 Note that the uncertainty is assumed to vanish at the contemporary position
in the cost-quantity plane, since current costs and levels of exploitation are well known.
3.3. Distributions
Natural resources are scattered around the planet in different forms with different probabilities for the cost of their
exploitation. Complex processes underlie the formation of these distributions, however, they may follow certain statistical trends, the nature of which stems from the nature of the resource. One particular property affects significantly
these statistical distributions, whether the productivity of resource producing units (rivers, plots of land, wells, mines
etc) tend to be very similar, which makes their ranking difficult, as opposed to resource types which have strong ordering. For example, solar energy can be produced using photovoltaic (PV) panels, and these panels may be installed
with equal ease almost anywhere, and scarcely populated regions of similar solar irradiation will have large potentials
for solar energy situated within very narrow productivity ranges where every plot of land has nearly the same productivity value. These are described by a particular type of statistical distribution for nearly identical resources. The
properties of wind energy potentials follow a very different structure. Plots of land within a geographical region are
unlikely to possess the same average wind speed, and can therefore be ranked in order of productivity, and thus of cost
for the production of wind power. As opposed to solar energy sites, wind power production sites can be described by
a hierarchical statistical distribution type.
Resource distributions for nearly identical resources are sharply defined in a narrow range of productivity, but
cut off at a maximum value, which corresponds to the best possible conditions for energy production. Meanwhile,
resources with hierarchical distributions occur in large numbers in low productivity ranges, and in ever lower amounts
as productivity increases. This property stems from the large number of positively contributing factors which are
required simultaneously in order to have a resource producing unit with high productivity. Such a property results in
a distribution that decreases exponentially with increasing productivity, but cut off below a certain low productivity
value, where it is simply assumed that no energy can be obtained with any reasonable amount of effort. This is shown
in Figure 2, top panel, where a typical distribution for nearly identical resources is shown in green, and a hierarchical
distribution is given in blue. Similarly to panel a) of Figure 1, the colour shading indicates uncertainty.
3 No error bars are present for the cost variables, since an uncertainty on costs corresponds to an uncertainty on how to distribute energy quantities
between existing cost ranges, which translates into an uncertainty in the quantity in each range.
4 Thus, the most probable cost-supply curve is neither the mean or the median of the skewed distribution, it is the mode, or maximum.
2
Solar Power:
Identical
Distribution
Wind Power:
Hierarchical
Distribution
Productivity ( MW/km )
Technical Potential A
Cost ( USD/MWhy1 )
Cost Distribution [ EJ/(USD/MWhy ) ] Productivity Distribution ( km2 )
C0
1
Energy Flow ( MWhy
Cost ( USD/MWhy
Figure 2: Depiction of two types of natural resources, based on their statistical properties. T op. A typical sharp distribution for nearly identical
resource types is shown in green, while the broad blue distribution is for hierarchical resources, from equations (1) and (2). Both are expressed in
terms of productivity. Different colour shadings represent uncertainty, as in panel a) of Figure 1. Bottom. Same distributions expressed as functions
of cost, from equations (5) and (4) through equation (3). Associated cost-supply curves are given in the inset. Note that the technical potential was
adjusted to be the same for both curves for visual clarity.
These resource distributions are well described by the following density functions:
( A
e d > 1
,
f ()d =
0
1
g()d =
A e
2
(2 ) 2
22
< 2 ,
2
(1)
(2)
where is the productivity (see definition above), A is the technical potential, is the width of the distribution and 1
is the minimum usable productivity, in the first case, and 2 is the maximum productivity available in the second case.
Costs are related to the productivity by an inverse relationship,
C=
Cvar
+ C0 ,
(3)
with which the distributions can be transformed into cost-quantity space. The scaling factor Cvar corresponds to a cost
per unit of land or sea area (wind, solar, biomass, wave energy), digging effort (oil, gas, coal, nuclear fuels, geothermal
energy) or construction effort (hydroelectricity, tidal energy), and the ratio Cvar / has units of cost per unit of energy
produced. C0 corresponds to the sum of costs which do not depend on the productivity, per unit of energy produced.
The conversion of these productivity distributions into cost distributions is described in detail in section S.2 of the
supplementary material, and yields the following:
B
CC
AB
0 dC
(CC
C > C0
2e
0)
,
(4)
f (C)dC =
0
C C0
g(C)dC =
A e
2B
(CC0 )2
2B2
dC
C > C0 ,
C C0
(5)
where A is the technical potential, B a scaling factor and C0 a cost offset, the set of three parameters required to
define every distribution given in this work. These functions are illustrated in Figure 2, where the inset shows the
associated cost-supply curves. It is observed that for a similar technical potential, the curve for nearly identical
resources possesses less curvature up to very near the technical potential than those for hierarchical resources, a
property that stems from their lack of ordering, and results in similar cost values for most of the resources.
These functional forms have been found to reproduce very closely the cost-supply curves calculated by Hoogwijk
et al. using the land use model IMAGE, whose work does not assume any functional dependence on cost for its
distributions [18, 19, 22]. Distributions were calculated by producing cost ranked histograms of calculated potential
renewable energy flows (wind, solar and biomass) at every point of a 0.5 0.5 grid of the earths onshore land. Thus,
their form originates purely from statistical properties of the aggregation and ranking of the resources modelled. Using
least-squares non-linear fits, the cost-supply curves in their work were found to agree very well with one or the other
of the functions given above, depending on the nature of the resources: solar energy and agricultural land are well
represented by the distribution for nearly identical resources, while wind power and rest land are well represented
by the hierarchical distribution. Additionally, the distribution for hierarchical resources was found to agree well with
observed cost distributions of uranium as reported by the International Atomic Energy Agency (IAEA) [26]. Examples
of non-linear fits of these functions to IMAGE data are given in section S.2.5 of the supplementary material.
No such global cost ranked data exist for the remaining types of natural resources that could enable fits of distributions distribution. Choices of distributions were therefore taken as assumptions, made based on the physical nature
of the resources. Potential basins that could be created for hydroelectricity possess very individual characteristics,
which makes them hierarchical. Geothermal resources, however, were treated as a hybrid mixture of the two, since
good geothermal sources in active volcanic areas such as Iceland can be ranked, but large amounts of very similar
sites can be found in non active areas. Stock resources however were treated slightly differently. Different oil and gas
occurrence subtypes (e.g. conventional gas, shale gas, clathrates, etc) originate from different geological processes
which have no strong relation to one another, and should therefore be treated independently. These resource subtypes
are characterised with different costs of extraction. They were assigned one hierarchical distribution per subtype, and
subsequently aggregated. This resulted in composite cost-supply curves with complex structures. Coal, uranium and
thorium resources were considered to occur as a single subtype, supported by the fact that the data for uranium was
found to follow well the hierarchical distribution.
3.4. Uncertainty
The methodology used in this work for treating uncertainty is fundamental to this analysis of economic potentials
as it allows the incorporation all available information, even when sources are inconsistent or conflicting. Inconsistencies can be found between assessments for most individual natural resources, and are the result of the use of
different approaches and assumptions, which can be determinant for the technical potential values derived. This is
most obvious in resources such as wind power, solar energy and bioenergy, where the total amount of appropriate land
depends highly onto competing activities, making the assumptions in the evaluation of the land suitability factors the
main drivers of uncertainty. Other such assumptions are world population and the associated food demand, levels of
technological development and changes in agricultural productivity. Resource assessments are uncertain by nature,
since it is not possible to know with certainty the complete geological content of the crust of the earth, or to predict
the weather and associated wind, sunshine and rainfall with perfect foresight. Thus the comparison of ever larger
numbers of natural resource assessments is the key to define ranges of confidence, and these are as important as their
associated most probable potential.
This work uses a consistent methodology to define probability distributions for cost-supply curves. Three costsupply curves are derived from resource assessment data, where two are used to define the 96% confidence region in
the cost-quantity plane, and one taken as the most probable of all possible curves. In all plots of this work, the most
probable cost-supply curves are given in blue and the 96% confidence limits are displayed in red. Uncertainty ranges
are almost always asymmetric, since upper ranges are intrinsically characterised by smaller amounts of accumulated
knowledge.
Uncertainty is treated differently for renewable resources compared to stock resources. For renewables, costsupply curves were obtained from the literature or calculated and taken as the most probable curves, while the 96%
confidence limits were obtained by scaling the technical potential to the limits of its uncertainty range, defined by
an ensemble of different studies.5 In the case of stock resources, all resource assessments provide classifications
associated with cost ranges and various levels of confidence. In these cases, three cost-supply curves were calculated
by assigning probabilities to uncertainty classifications, as in panel a) of Figure 1 (i.e. reserves were assumed to
exist with a 98% probability, while reserves plus all speculative resources were assumed to be available with a 2%
probability). Individual methodologies for all types of resources are described in the supplementary material.
4. Renewable energy resources
4.1. Wind energy
Wind speeds depend strongly on altitude as well as on landscape topologies, the climate and the type of land
cover, or roughness. In general, wind speeds increase logarithmically with elevation at low altitude (see for instance
Srensen [27]), and, for a specific elevation and geographic location, occur statistically following a well defined
Weibull probability distribution which decreases both towards zero and large wind speeds (for instance Grubb and
Meyer [28]). Average wind speeds on sites useable for energy production, for instance in the United Kingdom, range
between 5.1 and 9 m/s at 10m elevation, the lower boundary determined by technology and the upper limit by the
decreasing supply of locations with large wind speeds (for instance in [27]). The power density offered by land areas
must be calculated using technical characteristics of particular turbines. For a particular site, one integrates over all
wind speeds the product of the site wind speed probability distribution and the turbine power curve. However, a
correlation exists between the yearly averaged wind speed and the number of full load hours [29]. The minimum
possible distance between a turbine and its neighbours in a wind farm are determined by losses produced by the wake
of neighbouring turbines which results in lower wind speeds and increased turbulence, and scales with the turbine
rotor diameter, limiting the density of energy that can be extracted per unit of land area (for instance in the work of
Mackay [30]).6 Hoogwijk et al. assumed a maximum density of energy production of 4 MW/km2 [19].
Various research groups have calculated the global distribution of wind power, resulting in a range of values
between 72 and 2509 EJ/y for onshore wind power [19, 28, 3134], and about 57.4 EJ/y for offshore wind power
[16]. A global onshore value of 346 EJ/y has been derived Hoogwijk et al. [19], in whose work, used for the present
analysis, estimations of average wind speeds were applied to points on a global onshore grid, as well as the land
suitability for the installation of wind farms using the land use model IMAGE 2.2.7 Energy potentials obtained from
yearly averaged wind speeds determined on every point of the grid were subsequently aggregated into various costsupply curves for specific regions of the world, according to the land suitability factor of each point. Cost values were
determined using a present value calculation including fixed and variable costs, capacity factors and energy densities
associated to particular land areas.
Wind farm sites follow very strongly the distribution for hierarchical resources (see for instance the exceptional fit
of Figure S.4.1 in the supplementary material), since good sites with average wind speeds exceptionally suitable for
energy production are geographically scattered, and the majority of areas in any region of the world possess mediocre
average wind speeds, allowing a strict ordering of resource units (see for instance the European Wind Atlas [36]). The
5 This
is done in order to avoid inconsistencies where curves calculated independently, for instance by fitting data, could cross in some cases.
turbines intercept a larger wind front, they are also spaced further apart in two spatial directions. Thus, while the power production
of large wind farms scales with the square of the length of the blades, it scales inversely with the square of the distance between turbines. These
two effects almost cancel each other out, except for the fact that taller turbines intercept higher wind speeds at higher altitudes.
7 For details on IMAGE see [35].
6 While larger
Solar Energy
Wind Energy
1000
1000
LCOE (USD / MWh)
LCOE ( USD/MWh )
800
600
400
200
0
0
3384 EJ/y
346 EJ/y
134 USD/MWh
500
1000
1500
Energy Flow ( EJ/y )
800
600
438 USD/MWh
400
200
0
0
2000
4000
8000
12000
Energy Flow (EJ/year)
Hydroelectricity
Geothermal Energy
1000
66 EJ/y
800
LCOE (USD / MWh)
LCOE (USD / MWh)
1000
600
400
200
36 EJ/y
800
600
400
200
104 USD/MWh
68 USD/MWh
0
0
0
0
50
100
Biomass Primary Energy
A2 B2 B1
Cost (USD / GJ)
A1
A2: 305 EJ/y

B2: 321 EJ/y
B1: 447 EJ/y
A2: 679 EJ/y
6
4
2
0
1.11 USD/GJ
400
800
1200
40
60
80
100
Ocean Energy
1000
LCOE (USD / MWh)
10
20
23 EJ/y
800
600
400
200
0
0
203 USD/MWh
20
40
60
Figure 3: Cost-supply curves for renewable resources: wind power, solar energy, hydroelectricity, geothermal power, biomass and ocean energy.
profitability of a wind farm venture depends strongly on the quality of the wind resource, determined through capacity
factor and average turbulence values. For a fixed turbine investment cost, low capacity factors increase dramatically
the cost per unit of electricity produced.
Figure 3 presents the global economic potential for wind power. It gives an aggregate cost-supply curve, using
for the most probable curve the data calculated by Hoogwijk et al., which involves the most detailed methodology for
determining the suitability factors, resulting in a technical potential of 346 EJ/y [19]. While Lu et al. estimated an optimistic technical potential of 2509 EJ/y by calculating wind potentials over the global onshore area and excluding low
wind areas by restricting capacity factors to values above 20% [31], Archer et al. calculated a potential of 2257 EJ/y
in an assessment where the land included was restricted to class 3 wind energy sites but did not include alternate uses
of the land, a value taken as the upper boundary of the uncertainty range [33]. Meanwhile, an estimate of 72 EJ/y was
obtained using an evaluation of the number of sites with average wind speeds above 5.1 m/s, but with an arbitrarily
chosen value of 4% of that land available for wind turbine installation, in order to account for alternative land uses,
taken as the lower boundary of the uncertainty range [34]. All other existing studies result in values within this range
[28, 32].8
It is to be noted that using up a large fraction of that potential results in large areas becoming covered by wind
farms. Since typical individual wind turbines currently have capacities of 3 MW but capacity factors of about 25-35%,
compared to 80% for coal power stations, replacing one coal power station of 1 GW for wind energy requires 700 to
1100 turbines, covering an area of 500 to 1500 km2 , compared to about 1 km2 for the original power station9 . Even
though agricultural land used by wind farms may still be cultivated, and therefore a competition for land with agriculture does not directly occur, strong emissions reductions pathways based on substituting fossil fuels for renewables
likely implies that, given the large number of wind turbines required, these would invade permanently traditional rural
landscapes.
In the case of offshore wind power, although distributions of wind speeds at offshore locations tend to have a
higher median, the air flow possesses less turbulence and wind speeds vary less in time, all of which contribute to
produce a higher power density in terms of geographical area, the total area where such turbines can be installed is
small compared onshore areas, unless floating turbines become widely available [37]. The potential of offshore wind
energy was evaluated to 22 EJ/y by Hoogwijk et al. and to 57.4 by Krewitt et al., both based on the work of Fellows et
al. with power density values of 10 and 12 MW/km2 [16, 32, 38]. This potential could be at best only approximately
six times lower than the most probable potential of onshore energy given in Figure 3, and costs per unit energy are
significantly higher [39]. Due to the lack of reliable and consistent cost data, a cost-supply curve for offshore wind is
not presented in this paper.
4.2. Solar energy
Solar radiation over the Earth surface is of about 1.2105 TW, or 3.6106 EJ/y [40]. The fraction of that energy
that can be harvested with existing systems has been estimated by several studies, and ranges between about 1340
to 14800 EJ/y [18, 21, 41]. Even though the total generation of energy from solar technologies has been increasing
steadily during the last two decades, they still represent very low percentages in their respective categories; solar heating systems account for 0.3% of the total energy used for heating in 2008, and solar electricity generation represented
only 0.06% of the total electricity generation during the same year [42].
Solar energy can be harvested using either of two existing technologies, photovoltaic (PV) devices (see [43]) or
concentrated solar power (CSP) (see the International Energy Agency (IEA) Energy Technology Systems Analysis
Programme (ETSAP) [44]). Single crystal silicon photovoltaic diodes currently have light conversion efficiencies of
up to 25%, while III-V semiconductor cells such as GaAs systems have efficiencies of up to 28%, and solar cells
using concentrated sunlight can convert light as efficiently as 43.5% [45]. The resulting electricity generation energy
density ranges between 5 and 100 MW/km2 , depending on the type of devices used and the geographical location. On
the other hand, CSP technology uses a traditional steam turbine where water was heated using sunlight concentrated
with various arrangements of mirrors, and have efficiencies of around 13-24% and energy densities near 25 MW/km2
[14, 44]. These systems are however restricted to high irradiance areas, and therefore have a lower global technical
8 Fellows
9 The
et al. concludes with an estimate for 2020 of 148 EJ/y, while Grubb and Meyer calculated a global potential of 191 EJ/y [28, 32].
variation originates from both assumed ranges in capacity factors of 25-35% and turbine densities of 2.2 to 4 MW/km2 [19, 30].
potential. Solar energy is well represented by a distribution for nearly identical resources, since within areas of similar
irradiance and average cloud coverage, its cost does not depend strongly on the nature of the land, having identical
productivity values that depend solely on the chosen technology, and the opportunity cost of the land is the limiting
factor to its technical potential.
Figure 3 shows the global economic potential for solar energy, using PV as technology. It is an aggregation of
curves determined in various world regions, based on the work of Hoogwijk [18], but rescaled to match the values
of de Vries et al. [21]. The global technical potential is of 3384 EJ/y. Regional cost-supply curves were drawn from
an analysis performed using the land use model IMAGE 2.2 to determine land suitability factors and the opportunity
cost of the land at every point of a global grid, while regional estimates of solar irradiation were used to determine the
energy potential, which have been taken here for the most probable cost-supply curve. The lower boundary curve of
the uncertainty range, with a technical potential of 1340 EJ/y, assumes that land availability does not increase from
2000 levels, while the upper boundary curve of the uncertainty range, with a technical potential of 14778 EJ/y, stems
mostly from increases in land availability following future reductions in the amounts of land required for agriculture
through improvements in productivity [21].
The use of a large fraction of the technical potential for solar energy using PV systems signifies covering up large
amounts of land with solar panels. Mackay calculates an average productivity value of 10 MW/km2 for the United
Kingdom, while Hoogwijk used values between 6 and 25 MW/km2 , depending on the geographical location, with
capacity factors of up to 50% [18, 30]. When compared to coal power plants of capacity of 1 GW and capacity factor
of 80%, this implies that the replacement of one such power plant by solar panels requires an area between 50 and
500 km2 .
4.3. Hydroelectricity
Hydropower stems from water pressure gradients that are produced by the run-off of rainfall through landscape
topographies, using dams to restrict water flow and accumulate water at elevation level higher than that given by the
landscape, producing a potential for electricity production using turbines (see for instance [46]). Hydroelectricity is
the most deployed renewable electricity technology, with a global installed capacity of close to 1 TW, which produced
around 2% of the total primary energy supply in 2008 [42]. As a fraction of its total technical potential, it is also
the most developed of all renewable resources, to the extent that around 23% of the global hydroelectric technical
potential is currently in use. However, its exploitation around the world is not even: 25% of the European technical
potential has already been developed, while Africa uses only 7% of its hydroelectric resource [47].
Costs of hydroelectric systems are highly site-specific and were found to have varied between around 400 to
4500 USD2002/kW in an extensive global analysis done by Lako et al. [48]. These values are influenced by many
different factors, which include material and labour costs, but also critically the opportunity cost of the land. The
latter refers to the consequences of flooding large areas of land, and the resulting displacement of communities and
agricultural activities, and thus varies strongly from region to region. For this reason, the deployment of hydropower
is often decided on political rather than financial grounds. Hydroelectric resources were assumed in this work to
follow a hierarchical distribution, since available natural basins that can be flooded possess vastly different geographic
characteristics that make them unique and produces strong ordering.
Figure 3 shows the global cost-supply curve for hydroelectricity, where the dotted line indicates the cost at the current deployment of 12 EJ/y, 23% of the modest value of the most probable technical potential of 66 EJ/y, calculated in
this work from data gathered in the World Atlas and Industry Guide 2011 [47]. The high deployment to potential ratio
is an indication that the remaining number of suitable sites for building dams is relatively limited. The intersection of
this curve with the current total hydroelectricity generation value yields a cost of production of about 68 USD/MWh.
However, the development of hydropower projects hardly follows an order based onto cost, but follows instead an
order dictated by political considerations, which are out of the scope of this work. Therefore, this value is only indicative, and projects with LCOE values between 23 and 460 USD/MWh have been recently developed [49]. The
use of this cost-supply curve in modelling involves the inevitable assumption of development following a cost order.
In long term scenarios, this is reasonable, since the development of the limited number of remaining available sites
involves either increasing opportunity costs in inhabited areas due to increasing local populations, or increasingly
large transmission costs associated with increasing distances to uninhabited areas. The cost-supply curve was derived
using the theoretical, technical and economic local potential values from [47]. Since the definition of the economic
potential in IJHD is not given, the (asymmetric) range of the distribution of recent cost values in the work of Lako et
al. was interpreted (in 2008 dollars) as what is currently assumed economic. The remaining technical potential (above
the economic potential) was assumed to involve higher costs. The upper boundary curve of the uncertainty range was
derived by considering the aggregated global theoretical potential of 148 EJ/y from the data in the World Atlas and
Industry Guide, while the lower boundary curve of the uncertainty range was derived by assuming that no additional
construction of hydroelectric dams occurs in the future, limiting future hydroelectric generation to 12 EJ/y.
4.4. Geothermal energy
Geothermal resources, stored beneath the Earths surface in the form of heat, are heat sources constantly replenished by the radioactive decay of isotopes of uranium, potassium and thorium (see for instance [50, 51]). Although
geothermal heat has been used since prehistory, and its utilization for electricity generation commenced at the beginning of the last century, its current deployment is small in comparison with other sources of energy. It currently
accounts only for 0.3% and 4% of the total electricity generation and heating production respectively [42, 52, 53].
Geothermal resources are classified in four categories: hydrothermal (liquid and vapour dominated), hot dry
rock (where fluids are not produced spontaneously), magma (molten rock in regions of recent volcanic activity)
and geopressured (hot high-pressure brines containing dissolved methane) [54]. The most commonly used type is
hydrothermal, although high expectations exist regarding the development of Enhanced Geothermal Systems (EGS),
oriented towards the hot dry rock type through hydraulic stimulation [55]. According to estimations made by Aldrich
et al. based on a report of the Electric Power Research Institute (EPRI), the estimated geothermal heat accumulated
under the continental masses to a depth of 5 km depth is of approximately 1.46108 EJ, most of it assumed to be
stored in rocks and water, with a proportion of 6:1 in favour of the former [56, 57]. Even though this is a vast amount
of heat, only a small part is recoverable for productive purposes.10
While geothermal resources are available all over the world, their accessibility differs from site to site according
to various technical characteristics including the geological structure of the ground and the depth and type of heat
reservoirs. In the vicinity of tectonic plate boundaries, narrow zones characterised by significant volcanic activity
(so-called volcanic belts), geothermal gradients are particularly high, between 40 and 80 C per km of depth, enabling
the extraction of high temperature geothermal resources. On the other hand, areas with low volcanic activity are
characterized by low and uniform geothermal gradients: around 25 C per km of depth [56, 57]. The extraction
of geothermal resources in active areas are highly site-specific, and thus were assumed to follow a distribution for
hierarchical resources [59]. Meanwhile, geothermal gradients in the rest of land masses have very similar properties
and costs, and were therefore assumed to follow a distribution for nearly identical resources. Cost-supply curves were
produced for both types of land and both hydrothermal and EGS technologies, generating four curves which were
subsequently aggregated in each world region.
Stefansson found a high correlation between the number of active volcanoes in a particular region and the estimate
of the size of hydrothermal resources for electricity generation in the same region [60]. Using the total number of
volcanoes active in the world, discarding those located on the sea floor or in arctic regions, he estimated a global
installable hydrothermal electricity producing capacity of approximately 200 GW (producing 6.0 EJ/y of electricity
with 95% capacity factor). Using this information, along with the statistical analysis between wet and dry systems
developed by Goldstein et al. [61], Bertani estimated the total geothermal installable electricity producing capacity of
1200 GW (36 EJ), including hydrothermal and EGS technologies [62, 63].
The global aggregation of curves yields the cost-supply curve presented in Figure 3. The associated global technical potential of 36 EJ/yr, involves a 95% capacity factor. Cost values were obtained from the International Energy
Agency (IEA) [64]. The lower and upper boundaries of the uncertainty range, of 4 and 111 EJ/y, are explained in
section S.3.4 of the the supplementary material.
4.5. Bioenergy
Bioenergy, energy derived from plants, is currently the most widely exploited renewable energy resource, with
51 EJ/y, 10% of global annual primary energy use [42]. The combustion of biomass derived fuels is nearly carbon
10 Note that the average replenishment of the geothermal heat underground is several orders of magnitude inferior to the stock of heat currently
available: around 65 mW/m2 at the continental level, producing an average thermal energy recharge rate of about 315 EJ/yr [58]. This value can be
considered as the theoretical potential of geothermal energy if viewed in terms of sustainable extraction of geothermal resources over an extended
period. However, the amount of time over which geothermal resources could be used at higher rates than this is likely to be more than one thousand
years.
neutral if CO2 uptake during plant growth is taken into account, minus losses occurring in transformation processes.
Thus, biomass based technologies provide an important emissions mitigation potential. While biomass combustion
using integrated gasification combined cycle technology (BIGCC) is expected to become the most efficient biomass
based electricity production method [65], the combination of biomass and carbon capture and storage technology has
been shown to produce negative CO2 emissions [66], thus providing the potential for reductions of atmospheric CO2
concentrations, or for compensating other emissions. Moreover, the emissions factors of some power plants using
conventional coal technologies are being be reduced by co-firing coal and biomass fuels. Meanwhile, liquid biofuels
derived from biomass, such as ethanol and biodiesel, have the potential to replace oil-derived transport fuels with
minimal changes in vehicle internal combustion engine technology and jet engines [67].
Biomass currently used for electricity and biofuel production largely originates from forestry and agricultural
residues, and other forms of commercial or household mixed solid waste (MSW). Volumes of waste available could
amount up to 100 EJ/y but are highly uncertain and not studied here11 [25, 68]. The larger share of bioenergy potential
lies with the production of dedicated biomass crops. Global technical potentials for primary bioenergy range between
0 and 1550 EJ/y [20, 21, 24, 25, 69]. Bioenergy crops include perennial woody short rotation coppiced trees, such as
willow, poplar or eucalyptus, perennial grasses such as miscanthus, elephant grass and switchgrass, starch rich crops
such as wheat, corn, sugar beet and cane, and oil rich crops such as rapeseed and palm. Depending on their nature,
they can be transformed into energy carriers by using, among many processes, combustion, gasification or anaerobic
decay for electricity production, fermentation or the Fischer-Tropsch process for transport fuel production [70].
Biomass production for energy purposes makes use of agricultural land and thus may have a high opportunity
cost. The technical potential that lies in agricultural land is large, but energy production from biomass is in direct
competition for land with food production, a situation which has the potential to drive significant increases in world
food prices [71]. Following the approach which has been used recently by many authors (for instance in [20, 22,
24, 25]), the explicit assumption is taken in the present work that future bioenergy production uses no more than
leftover land after the global food demand has been met, a premise that is difficult to justify in the absence of specific
legislation and further investigations, but it avoids the complex problem of simulating food and biomass prices.12
Thus the bioenergy potential is obtained by subtracting from the total biomass potential the amount required by the
food demand, based on population growth curves and dietary assumptions.
Hoogwijk et al. evaluated the use of the land at each point of a global grid yearly up to year 2100 using IMAGE 2.2,
in which leftover agricultural land was termed abandoned land [20, 22]. The reported cost-supply curves were
observed in the present work to follow a distribution for nearly identical resource units using non-linear fits of eq. (5)
to their data. In addition to agricultural land, however, other types of geographical areas with lower productivities exist
which can be used for particular bioenergy crops. These were labelled rest land by Hoogwijk et al. and contribute
a significant global technical potential. They were found to follow the distribution for hierarchical resources by using
fits of eq. (4) to their data. Examples of such fits are described in section S.2.5 of the supplementary material. Land use
depends strongly on assumptions regarding world population, diet habits, global urbanisation and trade of agricultural
products. The four main SRES scenarios, A1, A2, B1 and B2 (see [72]), were taken as assumptions for all exogenous
variables in these calculations, and results are presented for each. Large differences arise between scenarios, with
technical potentials ranging between around 302 EJ/y for the A2 scenario to 676 EJ/y for the A1 scenario, which
result in large uncertainties for values of the global biomass technical potential. This work has however been revisited
more recently taking into account additional assumptions of future water scarcity and land degradation by van Vuuren
et al., yielding lower estimates of between 120 and 300 EJ/y [69]. Other ranges have been estimated using different
methodologies, with more pessimistic projections of 0 to 648 EJ/y by Wolf et al., and optimistic values of 367 to
1548 EJ/y by Smeets et al. [24, 25]. The low end of the range given by Wolf et al. stems from high projected food
demand and low agricultural productivity, while the high end is due to mostly vegetarian diets and high productivity.
Meanwhile, the high end of the range of Smeets et al. originates from super high agricultural productivity, high
availability of the land and landless animal production systems.
11 Waste amounts depend on efficiency of biomass use (such as in food or timber production) and therefore subject to significant changes
depending on future policy. These are therefore difficult to model, as are their associated costs of production.
12 Bioenergy potentials could in principle be larger if global food demand is not met. However, it will not be lower if global food demand is
indeed met. The problem of simulating food prices is complex as it involves modelling both local food markets, underreported in developing
countries, and efficiency changes associated with changes in food prices.
Figure 3 presents the global economic potential of bioenergy in terms of primary energy before conversion to
electricity or liquid biofuels (derived from the data in [20, 22, 24, 25]), using both abandoned and rest land, but
includes also a small component from bagasse of 3 EJ/y. Four cost-supply curves are given, calculated by Hoogwijk
et al. for the A1, A2, B1 and B2 SRES scenarios in 2050, shown as solid curves13 [22]. A value near zero was taken
for the lower boundary of the uncertainty range, consistent with the low end of the range calculated by Wolf et al.
while the high end of the range calculated by Smeets et al. was taken for the upper boundary of the uncertainty range
[24, 25]. For a decarbonisation scenario, the cost-supply curve derived for the B1 SRES scenario was considered the
most probable cost-supply curve, but for other types of scenarios, choices of curves consistent with particular working
assumptions should be made.
4.6. Ocean energy sources
The term ocean energy denotes renewable energy produced using seawater as a resource, where unlike for wind
energy or hydroelectricity, not only the kinetic energy of seawater can be used to produce electricity, but also temperature gradients in the ocean and salinity differences near river mouths. Using ocean energy as a general classification
type, it can be divided into four main sources of energy [73, 74]:
Wave energy, driven by transfers of energy from the wind to the surface of the ocean,
Tidal energy, driven by the rise and fall of sea levels due to gravitational forces (tidal range) and the resulting
water currents,
Ocean Thermal Energy, driven by temperature gradients between upper and lower ocean layers,
Salinity Gradient energy, derived from salinity gradients between ocean and fresh water at the mouths of
rivers.
Section S.3.6 of the supplementary material provides a review of theoretical potentials for these sources, resulting
in a total that could be as high as 600 EJ/y. Technical potentials however are much lower and uncertain, since the
current development status for ocean energy technologies excluding tidal is preliminary, and cost data is in some
cases unavailable. Specific geographical and configurational requirements for tidal and salinity gradient technologies
involves, as it is the case for hydroelectricity, calculating the technical potential by summing the potential values
from a large number of individual studies. Such studies have not been performed exhaustively on a wide scale yet.
Meanwhile, wave and ocean thermal are based onto global extrapolations carried out using physical measurements.
Global energy potentials calculated in various studies are given in table 1, and additional details are given section S.3.6
of the supplementary material.
Energy potentials for ocean thermal and salinity gradient energy are theoretical and highly uncertain, and no
reliable cost estimates were found. These types of resources were therefore not included in the present calculations
for the most probable cost-supply curve, due to the risk of generating misleadingly optimistic potentials given the
lack of reliable information.14 Wave and tidal systems are better established. Therefore, using cost values obtained
from the IEA and the ETSAP, a cost-supply curve for ocean energy based on an aggregation of separate cost-supply
curves for wave and tidal energy was produced, and is given in Figure 3, with a small technical potential of 22.5 EJ/y
[49, 73]. The lower and upper boundaries of the uncertainty range were obtained from the extremal values of 8 and
72 EJ/yr respectively given in table 1.
5. Stock resources
5.1. Fossil fuels
As it occurs with all types of exhaustible natural resources, fossil fuel resources and reserves are known to continuously expand, even though they are gradually consumed. This is due to periodic resource discoveries and improvements in the methods of extraction. Therefore, what is considered economical to extract changes every year. Reserves
13 Values projected for the year 2050 were used since amounts of land potentially available in the future are expected to increase due to increased
agricultural efficiency.
14 The costs for ocean thermal and salinity gradient energy systems are likely to be much higher than those of tidal and wave energy. This would
result in a piecewise cost-supply curve featuring an additional step at high costs.
Natural Gas
Oil
10
67 000 EJ
Conventional Oil
2
0
0
40
60
Resource ( 1000 EJ )
Tight Gas
0
0
80
Shale Gas
Coalbed Methane
4
2
20
Methane
Hydrates
Oil Sands
6
4
46 000 EJ
Oil Shales
Cost (USD/GJ)
Cost (USD/GJ)
10
Conventional Gas
20
Hard Coal
10
220 000 EJ
Cost (USD/GJ)
Cost (USD/GJ)
37 000 EJ
6
4
2
6
4
2
100
200
300
0
0
400
20
40
Uranium
2
60
Thorium
0.1
1360 EJ
4680 EJ
0.08
1.5
Cost (USD/GJ)
Cost (USD/GJ)
100
Soft Coal
10
0
0
40
60
80
0.5
0.06
0.04
0.02
0
0
0.5
1
1.5
2
2.5
0
0
4
6
8
10
Figure 4: Cost-supply curves for fossil and nuclear resources, including oil, gas, hard coal, soft coal, uranium and thorium. Hard coal includes
anthracite and bituminous coal, defined as coal with a calorific content above 16 500 kJ/kg. Soft coal corresponds to sub-bituminous coal and
lignite, and includes all coal with a calorific content lower than 16 500 kJ/kg.

Technology
Wave
Energy
Min.
EJ/y
6.3
Mode
EJ/y
Max.
EJ/y
18.9
65
Tidal
Energy
1.8
Total
Ocean
Thermal
Energy
Salinity
Gradient
Energy
Total
8.1
3.2
3.6
22.5
7.2
72.2
32
85
5.8
7.2
17.1
62.8
83
240.2
Study
[75]
[34]
[11]
[76]
[76]
[34]
[77]
[77]
[78]
[79]
[16]
[80]
Table 1: Technical potentials for different types of ocean energy used to define the cost-supply curve. The uncertainty ranges are defined using the
Min and Max values, while Mode represents the most probable value.
are distinct from resources, the former referring to the resources that are known to exist with almost complete certainty and to be economical to extract, while the latter refers to those which are thought to exist with various degrees
of confidence, and those currently thought too expensive to extract. As technological improvements and additional
knowledge affect the economics of different methods of extraction, there is a flow from resources towards reserves,
and thus reserves expand [23, 81]. Meanwhile, discoveries continuously add to resources. As prospecting activities
for hydrocarbon sources remains very active, this makes the production of cost-supply curves more difficult than for
renewables, since is at best a snapshot in time of what is known to exist and recoverable with current technology.
In order to assess global energy potentials, it is nevertheless necessary to explore cost-supply curves for fossil fuels,
even if they are derived from current knowledge, and therefore expected to change in the future. It is unlikely that
fossil fuel resources turn out smaller than what is currently expected to exist. On the contrary, it is probable that they
turn out significantly larger as methods of extraction are devised for types of occurrences which were until recently
not thought possible to use, such as gas hydrates or oil shales. Therefore, the cost-supply curve uncertainty ranges
are highly asymmetric. The associated extraction costs, which increase as low cost conventional sources are depleted,
nevertheless decrease due to technological improvements, and it is therefore not immediately obvious whether costs
are likely to go up or down in the future.
Global cost-supply curves have been calculated previously by Rogner [23]. These results have been used extensively by the energy modelling community; however they are becoming increasingly outdated. This section provides
an update to the work of Rogner, but using an approach emphasising uncertainty, and thus, following the spirit of the
current treatment of renewable resources, in opposition to the approach of Rogner, the results of this section should
be interpreted as ranges rather than specific values.
The economic potentials of fossil fuels are given in Figure 4, showing in order liquid hydrocarbons, gaseous hydrocarbons, hard and soft coal, the last two being classified using their calorific content.15 For oil and gas, different
types of occurrences considered in this assessment are indicated with text. These are associated with independent
distributions of the hierarchical type, aggregated to produce composite curves. Due to the wide use and global diffusion of fossil fuel extraction technology, extraction cost ranges were assumed to be the same for all regions of the
world. In the case of coal, less information was found over differing types of mines and associated costs, and single
distributions were used, where costs were assumed to vary little with the amount extracted. This is unlikely to matter
in the long run given the very large scale of the resource, and limited expectations of its depletion.
Uncertainty ranges were determined using resource classifications, and in some cases where this is unavailable,
their nature. Oil occurrences obtained from the World Energy Council and the German Federal Institute for Geosciences and Natural Resources are classified as either reserves or resources, with the exception of oil shales, which
15 Hard coal includes anthracite and bituminous coal, while soft coal includes sub-bituminous coal and lignite, the last two having lower calorific
contents than the first two. The limiting calorific value used to separate the two categories is of 16 500 kJ/kg [13].
are as a whole considered resources only [13, 15]. Four types of oil resources were considered, conventional (crude)
oil, oil shales, oil sands and extra-heavy oil. Cost ranges were obtained from the IEA [82]. Gas occurrences follow a
similar trend, but a larger number of types of resources were considered: conventional gas [13], shale gas [83], tight
gas [11, 13], coalbed methane [84] and methane hydrates [85]. The respective cost ranges were obtained from the
ETSAP [86]. Large amounts of methane are known to exist dissolved in aquifers [11], but were not included due
to the lack of reliable data. Information for coal was derived from a mixture of data [13, 15]. Complete details on
the methodology underlying these curves, as well as region specific data tables can be found in the supplementary
material.
5.2. Fissile materials
Five sources of fissile materials for nuclear reactors are known to exist. These are enumerated in order of cost. The
first comes from stocks of highly concentrated 235 U (uranium) or 239 Pu (plutonium) originating from decommissioned
nuclear arsenals diluted with 238 U. The second source is lightly enriched 235 U/238 U produced from mined natural
deposits. The third originates from U and Pu recovered from spent fuel (using the PUREX process [87]). The fourth
source is thorium (Th) using the 232 Th/233 U fuel cycle. The fifth source is U which occurs in very low concentrations
in seawater. Producing a cost-supply curve involves creating a scenario for the nuclear sector, and requires careful
consideration of uncertainty. Additionally, if ingenious use of fast reactors is invoked for the future, fuel efficiencies
of up to 50 times larger could be obtained, altering dramatically these expectations.
In order to construct a cost-supply curve for U and Pu, the nuclear industry was assumed to continue to use current
methods and thermal reactors, and therefore, only fuels originating from naturally mined U and from nuclear arsenals
were considered. Many authors stress that deposits of Th worldwide are three times larger than those of U [8790].
However, less efforts at prospecting for Th ore have been carried out and as a consequence, the current reasonably
assured reserves of Th, in tonnes of natural Th, are lower than those of U [26], a situation which is likely to change
if interest in Th grows. The nuclear fuel cycle for Th being more efficient than that of thermal reactors based on U,
it leads to larger amounts of energy per tonne of natural Th and thus leads to lower fuel costs per unit of energy.
Costs only include the extraction costs given by the IAEA, without the inclusion of enrichment or transformation
components.
Detailed resource data from the IAEA for naturally occurring U and Th were used to construct two cost-supply
curves and associated uncertainty ranges [26]. The data are classified into four levels of certainty and four cost ranges.
Such resources generally increase naturally in size with increasing costs of extraction, as well as with uncertainty, an
effect produced by the hierarchical ordering of natural resource consumption and by the decreasing amount of effort
which has been spent on prospecting activities for resources more and more difficult to exploit. For the conversion
of resources from tonnages to energy values, an average conversion efficiency for thermal U reactors of 159 TJ/t was
used, determined from the 2008 electricity production of 2611.1 TWh from a global fleet capacity of 273.7 GW, with
a capacity factor of 80%, which used 59 065 t of natural uranium [26]. Meanwhile, the burnup rate for Th reactors
was derived from the value of 24 000 MWd/t reached by the experimental Indian model [88], equal to about 2100 TJ/t.
Panels e) and f ) of Figure 4 present the resulting global economic potentials for U and Th. Uncertainty ranges for U
were obtained by considering only reasonably assured reserves (RAR) for the lower boundary of the uncertainty range,
RAR and inferred reserves for the most probable cost-supply curve, and all of the RAR, inferred, prognosticated and
speculative resources for the upper boundary of the uncertainty range.
The uncertainty ranges are highly asymmetric due to the tendency of the size of speculative resources to increase
with the level of uncertainty. It is observed that in terms of energy, reserves of Th are larger than those of U, and that
these are also less expensive per unit of energy, due to the higher burnup rate of the Th system. It must be emphasised
that U resources could, in principle, be used with much higher burnup rates, were fast reactors to be deployed globally.
The resources of U do not include seawater U, as data over these are scarce and highly speculative.16 Finally, it is to be
noted that the fuel component of the levelised cost for nuclear reactors is very small compared to the investment costs,
which results in a very small influence of the fuel costs onto the decision-making, unless nuclear resources become
depleted.
16 U is present dissolved at very low concentrations in seawater (3-4 ppb, from [26]), giving rise nevertheless to large amounts of U given the size
of the terrestrial body of seawater. Water turnover due to currents is very slow however, making it highly speculative whether a significant portion
of seawater U can be recovered, and the costs involved are very high [87].
6. Summary of energy resources

Resource
Name
Wind
Solar
Hydro
Geotherm.
Biomass
Ocean
Oil
Gas
Hard Coal
Soft Coal
Uranium
Thorium
Type
Flow
Flow
Flow
Flow
Flow
Flow
Stock
Stock
Stock
Stock
Stock
Stock
Dist.
Hierarch.
Identical
Hierarch.
Hybrid
Hybrid
Hierarch.
Hierarch.
Hierarch.
Hierarch.
Hierarch.
Hierarch.
Hierarch.
Use
EJ/y
.72
.04
12
0.23
51
.002
170
109
139
30
-
Technical Potential
L
M
72
346
1340
3384
12
66
4
36
0
447
8
23
9
67
7
46
24
220
5
37
0.83
1.36
1.74
4.68
U
2257
14778
148
111
1548
72
98
106
419
75
3.43
12.27
Units
EJ/y
EJ/y
EJ/y
EJ/y
EJ/y
EJ/y
103 EJ
103 EJ
103 EJ
103 EJ
103 EJ
103 EJ
Table 2: Summary table for all energy resources.

S tock/Flow indicate whether resources are renewable flows or stocks.
Hierarch./Identical/Hybrid identifies the type of statistical distribution assigned. U se refers to current yearly consumption of these resources. L
indicates the lower boundary of the uncertainty range. M indicates the most probable technical potential. U indicates the upper boundary of the
uncertainty range.
Table 2 provides a summary of all types of global energy resources, classified by type (renewable flows or stocks),
to which a type of statistical distribution it is assigned (for hierarchical or nearly identical resources, or a hybrid
mixture of both), along with technical potential values. The potential values are given with their lower and upper
boundaries of the uncertainty ranges. For comparison, current consumption of these resources is given based on data
from the IEA [42]. Note that biomass resources are expressed in terms of primary energy, which become smaller
when converted into electricity according to the efficiency of transformation.
7. Conclusion
This paper presents an assessment of global economic energy potentials for all major energy resources, those with
a potential larger than 10 EJ/y. These were given in the form of cost-supply curves, adding an economic structure
to energy potentials, and therefore providing them an unambiguous definition. Additionally, these were provided
using a probabilistic construction that allows a simple representation of uncertainty. The curves were calculated
using assumptions over the cost distribution of resources using functional forms based on statistical properties of
resource types. The set of energy potentials include six types of renewable energy sources, wind, solar, hydroelectric,
geothermal, biomass and ocean energy, as well as four types of fossil fuels, oil, gas, hard and soft coal, and two nuclear
materials, uranium and thorium. While the potentials for renewable resources were determined predominantly based
onto an extensive review of the literature, potentials for stock resources were determined directly using resource and
reserve assessment data.
The cost-supply curves calculated in this work were produced in order to be used by the global energy modelling
community, for the purpose of constraining models of the energy sector in order to produce realistic scenarios of
future energy use. It is hoped by the authors that this work will supersede outdated studies currently used and provide
a consistently calculated update for all types of energy resources. In particular, the large set of regional cost-supply
curves underlying the aggregate curves presented in this paper form the core of a new model for natural resource use
and depletion for the global E3 model E3MG, to be used through the family of technology models FTT. Other regional
aggregates can be provided by the authors.
Resource assessments, however, change continuously as new information becomes available, and as new and more
sophisticated studies are carried out. Therefore, while the potentials presented here may become outdated in the future, however the methodology presented will not. The simple and robust theoretical framework and computational
methodology presented here can be very useful for reporting natural resource assessments, and enable direct use in
models of energy systems. It would therefore be appropriate to reuse this approach with new data as they become
available, either by rescaling the curves given (the values for the A parameters) or by recalculating new sets of parameters. In either cases, this approach provides a consistent and general framework for limiting all types of natural
resources in energy models.
Part II
Global Decarbonisation
8. Introduction
Energy flows from natural resources are a necessary component of all sectors of the world economy. Although
the economic output of the energy sector accounts only for small fraction of the world gross domestic product (GDP),
changes in the prices of energy carriers have pronounced consequences on the output of most economic sectors (see
for instance [91]). Since the price of energy carriers are reflected in the prices of goods and services originating from
energy intensive sectors, such changes can lead to increased inflation, decreases in economic output and reduced paces
of economic development. Many attempts have been made to capture such interactions between energy, the economy
and the environment (E3) in computer models (see for instance [1] and the various models reviewed). While many
models of E3 interactions do not incorporate explicit representations of natural resource use and depletion, or the
physical limits to available energy flows, very few represent endogenously exploitation costs and none of them feature
a particular emphasis onto uncertainty in the economic or technical potentials of natural resources. This is also the
case with E3MG, and for this reason, Part I of this work introduces a methodology and a database in order to address
this issue.
Using aggregates obtained from this newly developed energy resource database, as seen in the graphs of Part I,
comments can be made about the state of the energy resources of the world and their future prospects. The least
cost production method for electricity is currently based on the combustion of coal [49], while the least cost natural
resource for the production of transport fuels is oil (see for instance [42] p. 355). Fossil fuels are also dominant
for activities involving heat production, whether for human dwellings or used in industrial processes such as cement
and glass production or iron smelting [92]. Activities involving fossil fuel combustion currently generate significant
greenhouse gas (GHG) emissions that interfere with the climate system [93]. New climate policy strategies in several
nations increasingly target the reduction of the scale of fossil fuel combustion activities, through their replacement by
energy production methods using renewable energy resources. Additionally, strategies in energy policy targeting the
creation of sustainable energy systems require the replacement of flows of stock energy resources by flows of energy
from renewable energy resources. While various authors and lobby organisations have recently claimed that the world
could be entirely powered by renewable energy systems before the end of this century (notably by the WWF, see [17]),
their results rely on unpublished or estimated energy potentials that lack in transparency and rigour. It is not yet firmly
established whether any of these strategies are feasible in terms of available energy and material flows from nature,
land area and time.
In order to assess the feasibility of such strategies, and/or to model them with computer simulations, it is necessary
to use estimates, and their associated uncertainty, of available energy flows in order to evaluate potential substitutions
between sources. Such comparisons must however be done in terms of their economic feasibility, and costs associated
with each energy source vary widely. Energy potentials and associated cost structures were given in Part I in the form
of cost-supply curves for fossil fuels (coal, gas and oil), nuclear fuels (uranium and thorium) and renewable energy
resources (wind, solar, biomass, geothermal, hydroelectric and ocean energy). These cost-supply curves have the
form of probability distribution densities in cost-supply space, presented with uncertainty range boundary curves as
well as the most probable cost-supply curve. While renewable resources were given in terms of energy flows, stock
resources (fossil and nuclear fuels) were presented in terms of fixed amounts. However, in order to evaluate possible
substitution between energy sources, scenarios of stock resource use must be drawn in order to transform stocks into
flows and depletion timescales.
Part II of this work presents comparable potential flows of energy from all resources and an analysis of the feasibility of the substitution of energy flows between all possible sources, in view of the process of global decarbonisation.
Since the comparison of stock and renewable resources requires a dynamic treatment of stock resource use, a theory
is presented with which this can be done, and is applied in the model of technology substitution in the power sector
FTT:Power [4]. This framework allows to project either a flow of stock resource given a price path in time, or a price
path in time given a projected demand curve. It is applied however dynamically in FTT:Power, and provides a first
step towards defining an exogenous model of stock resource commodity prices (which excludes geopolitical pressures
and artificial demand related to speculation and hoarding).
While this paper is intended to be a tool for the comparison of global energy resources, emphasis is given onto
the feasibility of the decarbonisation of the global economy. It involves in a first step the definition of a model for
stock resource extraction given a price for the associated carriers. This model is used in order to derive scenarios
of flows of stock resources based on their respective cost-supply curves. These flows are subsequently compared to
those available from renewable energy resources. Using a profile for future total primary energy demand, various
scenarios for the allocation of resources can be derived given energy policy assumptions. As such, this work is
intended to be used by scientists and policy-makers alike in order to estimate possible futures for the energy sector
while respecting global technical potentials of natural resources. Since all values are given in terms of ranges and
probability distributions, emphasis is given on the uncertain nature of such projections for the energy sector.
The feasibility of the decarbonisation of the global economy is explored, in terms of the scale of renewable energy
resources that are necessary to replace the total amount of fossil fuels that are projected to be demanded by the end
of this century. The results paint a possible but difficult scenario where resources are used up to levels close to their
technical potentials, and the resulting significant global use of land is estimated. From these considerations, reflections
are drawn regarding the necessity to reduce the scale of energy consumption and improve the efficiency of the global
energy sector in order to achieve global decarbonisation.
9. Methodology and definitions
Cost
Technical Potential
9.1. The economic potential of natural resources
Supply
Figure 5: Sketch of the economic potential of energy resources as given by their cost-supply curves. Resources are allocated in such a way that the
sum of energy flows equals the demand (solid black line). However, for all opportunity costs to be zero, some resources will be utilised to higher
levels than others (solid colour line). When projecting resource use for a future where energy demand increases, new levels of resource use (dashed
colour lines) define their economic potential (colour arrows), as opposed to their technical potential, the level of use where a cost-supply curve
diverges. The new levels of resource use define a new overall marginal cost (dashed black line), deemed economic.
The economic potential of an energy resource depends on its opportunity cost and that of all other energy resources.
If the opportunity cost is negative, meaning that additional development is less expensive per unit of energy produced
than all other alternatives, then it is optimal to consume an additional unit. If that cost is positive, there is at least one
alternative which is more profitable. As argued in previous work [4], although it is not appropriate to assume rational
behaviour for every individual investor, investors in energy generation capacity have the tendency to follow this
rule, on average, when taking decisions,17 leading to a gradual minimisation of the overall cost of producing energy,
17 This relaxed assumption allows for unknown/unknowable sets of motivations influencing individual investor decisions, leading to distributions
in behaviour, of which the median is taken.
through substitutions, which occur following investor decisions. In such a framework, without changes occurring in
the global pool of knowledge on natural resource potentials (leading to changes in the cost-supply curves), energy
policy (e.g. taxes/subsidies) and the costs of technology (influenced by, for instance, learning-by-doing), the level of
use of all resources would at all times converge towards an equilibrium situation where their opportunity costs are nil.
However, the energy system is characterised by time scales that are much longer than the typical life span of energy
policy decisions and the evolution of knowledge. Therefore, this equilibrium never has time to take place, and the cost
of energy production is never really completely minimised.
It is nevertheless the case that some natural energy resources have higher or lower opportunity costs than others,
and that this varies on the global and regional levels. In order to determine which allocations of natural resources
generate economically efficient energy systems in particular regions of the world, starting from a particular state of
the sector, additional demand for energy flows should be allocated between resources according to their opportunity
cost. The economic potential of energy resources is therefore defined as the following. Assuming that energy demand
increases with time, using a particular projection, or a range of projections, of future energy demand, and allocating
the additional demand compared to its current value to natural resources according to their opportunity costs, new
levels, or ranges of levels, of natural resource use can be derived, as well as a new overall marginal cost, since
additional energy units are likely to be more expensive at higher levels of use. The difference between their future
individual levels of use and their current values defines their economic potential given the future date for which the
projected demand was taken (see figure 5). Opportunity costs are defined by comparisons of the cost-supply curves
associated to each resource at the current level of resource use, and the economic potentials depend on one another.18
Therefore, economic potentials can be increased or decreased using various forms of government support such as
taxes, subsidies or an emissions trading scheme, which generate preferences for particular resources in comparison to
others, by shifting up or down their cost-supply curves with respect to the others. By doing so their economic potential
can vary from zero to some useable value, in the case of support (e.g. solar energy), or from a useable value to zero,
in the case of taxation (e.g. coal based electricity).
This is represented formally in a model for energy technology substitution based on cost considerations, which
calculates rates of technological change, and therefore the change in the use of the associated natural resources.
Many such models based on overall cost optimisation (and an effective global social planner) have been defined
and used, for instance the TIMER model using a logit substitution function [94], MESSAGE and MARKAL using
an optimisation of the total cost of energy production [95], while FTT:Power represents investor behaviour seeking
lowest cost opportunities using a dynamic family of logistic differential equations [4, 10]. All these models require
cost-supply curves in order to constrain their output scenarios.
9.2. Scenarios for the exploitation of stock resources
A fixed energy demand, when met by renewable energy resources, results in fixed levels of use of these resources.
However, when met by stock resources, it results in particular rates for their depletion. The gradual depletion of stock
resources generates gradual increases in their cost of exploitation, an effect which is due to the natural average tendency towards the exploitation of resource occurrences with lower extraction costs first, and costs increase following
their depletion. Part I of this work presented cost-supply curves for fossil and nuclear resources. In order to allocate
energy demand between all potential energy sources, rates of exploitation of stock resources must be assumed, which
result in particular lengths of time after which these resources reach complete depletion, and depend on the price of
the associated energy commodities. This section presents a mathematical model that generates a relationship between
rates of exploitation of a given finite resource base and the price of the associated commodity. This theory is designed
to be used to treat the use and depletion of stock energy resources in a dynamic model of energy systems, and is
applied in FTT:Power. While the use of this framework is crucial for future work using the combination of E3MG and
the FTT family of technology models, it will also be used here to evaluate limiting scenarios of stock energy resource
use. This is done in order to explore the current state of global energy resources.
In a world with a perfectly efficient and competitive energy market, fast rates of resource extraction and no speculation over the future price of these commodities (or seeking rent maximisation through speculation), only the resources
18 Rigorously speaking, every Joule of energy should be allocated to energy resources following their opportunity costs, all of which change
every time a Joule of energy is allocated to one or another of the available resources.
with lowest extraction costs would be traded at any time. In such a situation, resource depletion would proceed in
perfect order following the increasing cost of extraction. In reality, the demand for energy commodities is met by
resource types with extraction costs within a certain range, delimited by the price of energy carriers, with a corresponding range of margins of profit. For example, the price of oil determines which of known oil fields are economic
to exploit, and the remaining fields are reserved for a future in which a higher price of oil is expected. The upper limits
for the extraction cost values that are still considered economical given the prices of energy carriers are defined by the
differences between these prices and the sum of all fixed costs such as those associated to transformation processes
and transport, and the minimum profit margins that industries will consider, plus the extraction cost. Increases in
the prices of energy carriers enable wider ranges of natural resource types to be exploited, for instance low grade or
unconventional fossil fuels, which are not profitable under low price conditions.
This behaviour can be summarised by stating that increases in the price of energy carriers unlock additional energy
resources. While low cost stock resources become increasingly depleted, increases in price of energy carriers enable
the exploitation of additional high cost resources in order to supply the demand. It may thus be inferred that price
paths in time produce supply paths in time for energy resources. It is however the demand for energy carriers that
determine their prices: they adjust in such a way that the supply resulting from the sum of resources unlocked meets
the demand.19
Following the terminology of McKelvey [81] and Rogner [23], reserves are continuously consumed but also
continuously expanded at the expense of resources. Reserves are defined as those currently economical to exploit, and
the boundary delimiting reserves from resources is defined by prices, which evolves in order to maintain a certain level
of reserves with existing resources. As we show below, it is remarkable that on the global level, the ratio between the
rate of consumption of oil and gas to the size of their associated reserves has been nearly constant in recent history.
This aspect strongly strengthens the assertion that it is the evolution of prices that enables the size of reserves to
follow the magnitude of their respective consumption levels, which we take as a starting postulate in order to define
the theory given below.
9.3. Calculation of energy flows from stocks
Flows of stock resources and associated depletion can be calculated using time paths for their associated carrier
prices which unlock just the right amount of energy at every time step that is not already produced by other stock
resources or renewable energy systems, in order to supply the total energy demand. This increase in price is associated
with the marginal cost of production for this energy resource.
Following the first panel of figure 1, we take n0 (C) as the initial cost distribution of a particular type of resource
(e.g. oil, gas, coal or uranium), and a time dependent resource distribution n(C, t) which represents the amount of
resource left. We take the assumption that the rate of extraction of stock resources at any time (and price of the
associated energy commodity P(t)) is proportional to the amount left which is considered economical to extract,
therefore at costs below the extraction cost value that delimit reserves from resources. We take a step-like function
f (P(t) C), which equals one below this boundary and zero above, as the probability of extraction of resources at cost
C. This is shown in panel a) of figure 6.
While remaining resources are n(C, t), reserves correspond to n(C, t) f (P(t) C). If the constant fraction 0 of
reserves are extracted at any time, the flow of resources is therefore as follows:
dn(C, t)
= 0 n(C, t) f (P(t) C),
dt
(6)
The time dependent energy carrier price P(t) drives the evolution in cost C of the boundary between reserves and
resources. For a constant or increasing commodity demand, as the size of reserves decreases following consumption,
this flow decreases and generates an upwards movement of P(t), shown in panel b), where the distribution of resources
19
Strictly speaking, this is true over long term averages. Short term fluctuations originate in large parts from stockpile hoarding or bottlenecks
in the production chain. In the long term, fluctuations in demand and supply originating from hoarding activities even out to zero, since the size of
stockpiles can only be finite (i.e. artificial demand created by hoarding activities in times of energy security uncertainty can only last for as long as
there is storage space), and supply mismatches tend to be corrected by building new processing capacity.
Cost
Resource Density
a)
c)
Exogenous Flow
d)
Flow
Marginal Cost
Exogenous Marginal Cost
b)
Resource Density
Cost
Time
Time
Figure 6: Process of stock resource extraction. a) Original (current) resource distribution as a function of extraction cost (black curve) and
probability function for resource extraction given the price of the resource commodity (red curve), its maximum being equal to 1. b) Dynamic
process of resource depletion as the price of the resource commodity increases. The black curves corresponds to the distribution of resources left
after increasing amounts of time have passed, associated with increasing prices for the energy commodity. c) Flow curve and uncertainty range
for the amount of resource unlocked by an increasing exogenous price of the commodity which allows the resource exploitation to proceed up to
an upper limit marginal cost (function of time). The area between the red curves indicates the 95% confidence level region, while the blue curve
represents the most probable curve. d Marginal cost of production and its uncertainty range of the commodity for an exogenous demand (function
of time). The values diverge at depletion, which, due to uncertainty, can occur at different values of cumulative production.
decreases in the low cost range and the boundary f (P(t), C) moves to the right. This produces a time dependent supply
(or flow) F(t) which is the sum of the production in all extraction cost ranges during the unit time dt:
Z
dN(t)
dn(C, t)
F(t) =
=
dC,
(7)
dt
dt
0
providing a connection between commodity prices P(t) and commodity flows F(t).
This theory can be used in several ways, depending which variable is considered exogenous and which is endogenous. Thus for an assumed time dependent price P(t), the flow F(t) is straightforward to calculate numerically
using a discrete time step using equations 6 and 7. This is shown in panel c), where a range of time dependent flows
is represented associated with the uncertainty in the amount of resources that actually turn out to be in place, using
a linearly increasing commodity price. This flow is low at low price values, where no resources are economical to
exploit, and at high prices, where all resources have been consumed.
Conversely, for an assumed commodity demand, the price value that unlocks just the right amount of resources
can be obtained by trial and error using an optimisation procedure. Panel d) shows the result of such a calculation with
a range associated with the uncertain amount of resources in place, where lower prices result from higher amounts
resources and the reverse, using as an example a large constant commodity demand. This results in a price value that
gradually increases as resources are consumed, accelerating when remaining resources are small compared to the level
of consumption, forcing the exploitation of expensive resource occurrences. It eventually diverges as depletion sets in.
In both examples, the red curves delimit the 96% confidence region associated with resource assessment uncertainty,
while the blue curve denotes the most probable values.
700
100
Reserves to Prod. ratio ( y )
Reserves to Prod. ratio ( y )
120
NAM
SAM
EUR
ME
AFR
ASP
World
80
60
40
20
0
1980
1985
1990
1995
2000
2005
2010
NAM
SAM
EUR
ME
AFR
ASP
World
600
500
400
300
200
100
0
1980
1985
1990
1995
2000
2005
2010
Figure 7: Reserve to production ratios for stock resources for Oil (Le f t) and Gas (Right). This ratio corresponds to the inverse of 0 , in years. In
the case of oil, this converges globally towards 42 y after around 1987, and is constant at 62 y for gas throughout the period.
9.4. Constant global production to reserve ratios

Insight for justifying the theory presented above can be obtained by considering the size of reserves for energy
resources and their rate of exploitation. Reserves continuously expand, but are continuously consumed as well [for
instance, see 81]. The size of global oil and gas reserves has never been constant in history, neither has global
production. However, the ratio of the two, when taken at the global level (only), is approximately constant, given in
figure 7, for which reserve and production data were obtained from the BP Statistical Review of World Energy [96].
Reserves concern a particular subset of specific resources (oil, gas, etc) that is both very probable to be in place
and feature costs of extraction that are within a range deemed competitive given the price of internationally traded
commodities. The rate at which reserves are consumed is equal to 0 , which controls the behaviour of equation 6.
On the regional level, various non-constant values for 0 are observed (see the curves of fig. 7 for various regions of
the world). Different regions have different energy policies related to their own political and geophysical situations.
However, their aggregated output supplies the international demand for resources, which sets the price. This price
moves up and down in order to adjust the upper value of cost that enables resources to be extracted; it defines the
size of reserves out of the resource base. A large amount of trade occurs between regions of the world, but overall
the value of 0 , on the global level, remains remarkably constant. This supports the fact that the perceived price level,
excluding effects from speculation and hoarding, evolves such that the ratio between production and reserves remains
approximately constant, by unlocking just the right amount of resources to include into reserves in order to supply
demand, given that reserves are exploited at a rate of 0 . It also reflects the maximum rate at which reserves can be
(physically and politically) extracted.
In the case of oil, (left panel of fig. 7), the inverse value of 0 converges towards 42 years and becomes constant
after 1987. Its evolution before this date appears related to the oil shocks of the late 1970s, where OPEC was formed
and Middle Eastern production decreased faster than reserves. In the case of gas, (right panel of fig. 7), this ratio is
constant at 62 years throughout the period. These values correspond to the time it would take to deplete reserves if the
price did not increase in order to unlock additional resources. In the case of coal and uranium, no historical reserve
data is given by BP, but a similar process is likely to take place, and for which the 2008 values are of 1
0 of 122 and
16 years respectively. These values are assumed to remain true for the future in the calculations that follow.
The significance of this phenomenon is not as surprising as it may initially appear. BP reserve data corresponds
to a perception by the industry of the current outlook of energy resources, and their expectations regarding prices and
global demand. It is natural to expect energy firms to increasingly expand their own conception of the economical
limit of reserves by considering to develop new projects that were considered prohibitively expensive in the past, in
order to maintain reserves to a certain level, and that this level should evolve following global demand. Examples of
this abound and include arctic oil exploration, canadian tar sands, ultra-deep offshore rigs, etc.
20
15
Gas
Oil
Coal
Uranium
10
0
2000
2020
2040
2060
2080
2100
Energy Demand per fuel type ( EJ )
Marginal Cost of Resources ( USD/GJ )
1000
800
Renewables
Nuclear
Gas
Oil
Coal
29%
600
400
34%
200
21%
2%
0
2000
14%
2020
2040
2060
2080
2100
Figure 8: Assumptions for calculations of stock resource flows and prices. Le f t Exogenous linear marginal cost component of price assumptions
used to calculate flows. right Exogenous energy demand components used to calculate marginal cost components of prices. Note that here, the
nuclear contribution to the total primary energy demand (TPED) appears as 2% instead as the 6% quoted in IEA TPED data [42] (IEA data for
nuclear electricity production is 10 EJ). This is due to IEAs use of an arbitrary efficiency factor for converting 10 EJ of nuclear electricity into
30 EJ of primary nuclear fuel, out of 514 EJ of TPED. The value used here is 10 EJ out of 494 EJ of TPED, since the conversion factor of nuclear
reactors is already included in the resource data of Part I.
10. Global overview of stock energy resources

10.1. Assumptions
The exploration of possible of stock resource flows can be performed using either prices as exogenous in order
to calculate resulting flows, or using flows (energy demand) as exogenous in order to calculate commodity prices,
excluding in both cases the effects of hoarding and short term demand fluctuations. Values produced here correspond
to marginal costs of energy production, and whatever margins of profit and other cost component considerations may
be added to them in order to derive real carrier prices.
We present first a calculation of stock resource flows given assumptions for energy carrier prices (which are, in
reality, assumptions on the marginal cost of energy production allowed by the carrier prices, profit margins and other
fixed cost components, all of which are excluded). The assumptions are given in the left panel of figure 8. Values for
0 are given in the preceding section. Given these, starting marginal costs required for supplying current demand were
evaluated by finding which value of P generates a dN/dt that equals the current demand. Following this, rates for the
increase of the carrier prices were used that generate an increase in supply consistent with current total primary energy
demand increase, but were maintained linear throughout the century. The resulting flows are given in the following
section.
Second, we present a calculation of the required marginal costs of energy production for each stock resource for
a scenario where total primary energy demand increased to nearly 1000 EJ/y, but where the current shares of energy
demand for these resources (coal, oil, gas, uranium) were maintained until 2100 (i.e. the structure of the current
energy system is maintained). This value of energy demand for 2100 is consistent with many recent projections, a
review of which is given in section 11.2. The results are given in section 10.3. Excluding unknown/unknowable price
fluctuations that may occur in the future given hoarding activities and related speculation, supply restrictions due to
international conflicts and other political events and associated speculation and other geopolitical phenomena that
cannot be modelled, these evaluations of marginal costs are appropriate values that can be used to determine what
future energy carrier prices could be if maintaining the current energy system, from the best of current knowledge
and uncertainty.20 These rigid assumptions explicitly exclude technology substitution in both sets of calculations, an
aspect which is explored in section 10.4.
20 Note that since storage space for commodities is finite, artificial demand generated by speculation and hoarding cannot increase indefinitely
and stored commodities eventually must return to the market. This therefore generates cyclic price fluctuations that even out to zero in the long run.
Oil resources
Natural gas resources
2100
2100
2080
2080
2060
2060
2040
2040
2020
2020
200
400
600
800
Resources available ( EJ/y )
100
Coal resources
2100
2080
2080
2060
2060
2040
2040
2020
2020
200
300
400
700
Uranium resources
2100
100
200 300 400 500 600

10
15
20
Figure 9: Flows of oil, natural gas, coal and uranium, as calculated using equation 6 and the cost-supply curves given in figure 4 of Part I. The 95%
confidence level region is situated between the red curves, while the blue curves indicate the most probable flow values. The curves start in 2008 at
current energy consumption values.
10.2. Flows from stock energy sources for exogenous prices

Scenarios for flows of stock resources were produced for oil, gas, coal and uranium using the resource distributions
underlying the cost-supply curves given in figure 4 of Part I, current rates of resource exploitation (values for 0 ) and
linear extrapolations of energy carrier marginal costs of production. In the case of coal, the sum of the distributions for
hard and soft coal was used. The results are given in figure 9. In all cases, the curves start in 2008 at the current global
production values reported by the IEA [42],21 with zero uncertainty. As time progresses, the increasing uncertainty in
resources assessments at higher levels of use produces ever larger ranges of possible resource production values, or
ranges of possible consumption paths, delimited by the red curves. These however must converge back to low values
where peaking and depletion occurs.
Different results are obtained depending on the size of the various stocks. In the case of oil, which includes all
types of unconventional oil, a peak in production occurs at around 2060, after which depletion begins. A similar
21 Note that the IEAs value for nuclear electricity production of 10 EJ was used, not its reported value of primary nuclear fuel of 30 EJ. Efficiency
factors for thermal reactors have already been taken into consideration in the cost-supply curve for uranium.
situation occurs with natural gas, which includes all types of unconventional gas and methane hydrates, peaking later
near 2080. However, coal resources are very large and depletion does not occur. It can only do so very far outside
the time horizon of 2100. Resources of natural uranium, as reported by the IAEA [26], are found to become depleted
rapidly within the current century after peaking before 2025 22 .
Potential flow values vary highly between resource types. Projected flows from oil resources are the largest, giving
however a faster rate of depletion compared to coal and natural gas. This is due to the current large rate of extraction to
resource ratio 0 . Their massive expansion occurring after 2020 is due to the large scale exploitation of unconventional
oil such as the tar sands and oil shales. While natural gas resources are smaller than those of oil, their depletion is
projected further into the future due to a lower extraction rate to resource ratio 0 . Their massive expansion after 2040
is related to large scale exploitation of unconventional sources such as shale gas.
In the case of coal, the rate of extraction is similar to that of gas, but their reserves are much larger, projecting the
depletion far beyond 2100. In the case of uranium, given the small resource base, the low burn-up rates of current
thermal reactors and the high value for 0 , the expected flow is very small compared to those of fossil fuels, but
the depletion is projected to occur rapidly within this century. This indicates that dramatically higher conversion
efficiencies are necessary to extend the resource base beyond the end of this century, which can be achieved with fast
breeder reactors [87, 97].
Note that these flow paths are not forecasts in any way. They are possible scenarios of resource use, given known
resources and realisable extraction rates and price paths, even though the price of carriers is hardly likely to follow
a linear trend. Nevertheless, for any resource extraction path, the cumulative flow up to complete depletion must be
equal to the technical potential of the resource, a requirement that has been carefully verified in these calculations.
Therefore, for higher rates of extraction and resulting higher resource flows, depletion must occur slightly sooner than
depicted here, and conversely, for slower resource use, depletion may occur slightly later.
10.3. Price paths for exogenous flows
Conversely to the previous section, the reverse problem may be posed where one looks for the appropriate price
of resources that unlocks just the right amount of resources to meet an exogenous demand. In this case, the demand,
or resource flow, is given as exogenous and the price, or marginal cost, is evaluated. This is done by performing and
inverse calculation using a price optimisation of eq. 6, such that the value of eq. 7 is equal to the demand, separately
for each type of energy carrier (oil, gas, coal, U). In such scenarios, it is possible that at a certain point in time,
given the values of 0 , the remaining resource base cannot meet anymore the demand. In such a case, the price
values gradually runaway to infinity, signifying that all resources situated at all possible prices of extraction are under
intense exploitation. Such a situation is very unlikely to occur, since the opportunity cost of using these expensive
resources would become very large and other technologies and energy resources would be more cost effective, leading
to technology substitution, switching away from that resource. Alternatively, the global economy may also readjust
its energy demand in order to avoid diverging prices of energy commodities. The divergence of prices therefore stems
from rigid commodity demand values that do not respond to price signals.
Figure 10 presents the results of such an exercise, using the assumptions described in the right panel of fig. 8, where
the current composition of the energy sector is maintained with a total energy demand scaling up to near 1000 EJ/y
in 2100. The blue curves correspond to the most probable resource bases given by the blue curves of fig. 4 of Part I.
Meanwhile, the red curves delimit the 96% confidence region, where the upper red curves correspond to the lower
bounds for resources, while the lower red curves correspond to the higher end of the resource ranges. Therefore, in
all cases, the marginal cost values calculated in the low end of the resource ranges diverge, while most of the curves
for the upper ranges do not.
In the case of oil and gas, gradual increases are observed in the marginal cost values, with a change in slope occurring between 2020 and 2030. This is related to the price enabling the accession of large unconventional resources,
which include predominantly oil sands and shale gas respectively. The availability of these large resources tend to
damp out possible future increases in price. Meanwhile, in the case of coal, the marginal cost value is hardly affected
by demand at all, unless depletion occurs (as in the low resource limit), which is very unlikely. This is due to the very
large resource being situated in a narrow range of extraction costs.
22 this excludes the reuse of fissile material available in nuclear waste, which would probably become economical be f ore the complete depletion
of natural uranium resources
Oil and Bitumen

8
6
4
2
0
2000
10
2020
2040
2060
2080
2100
Coal and Lignite

8
6
4
2
0
2000
2020
2040
2060
2080
2100
Marginal Cost of Uranium ( USD/GJ )
Marginal Cost of Coal ( USD/GJ )
Marginal Cost of Oil ( USD/GJ )
10
Marginal Cost of Gas ( USD/GJ )
10
Natural Gas
8
6
4
2
0
2000
2
2020
2040
2060
2080
2100
2060
2080
2100
Natural Uranium
1.5
0.5
0
2000
2020
2040
Figure 10: Marginal cost calculations performed for four stock energy resources using the assumptions described in the right panel of fig. 8, which
maintains the current composition of the energy sector, while expanding the total energy demand up to nearly 1000 EJ
Finally, in the case of uranium, the marginal cost is expected to diverge, and the resource to run out, over the
whole resource uncertainty range, if the current share of energy demand supplied by nuclear is maintained up to 2100.
At the current uranium burn-up rates, the resource base is insufficient. This indicates that, either the nuclear industry
will be phased out before 2100, or that much higher efficiency rates in resource used per unit of electricity produced
will be achieved, involving a much higher recycling of nuclear waste than occurs at present. Future energy systems
will therefore adopt either of these solutions in order to avoid this projected fuel price divergence.
10.4. Real systems operate somewhere between these limiting situations
Although the calculations given in the last two sections provide insight on the scale of available resources and on
the process of their gradual consumption, they both depict limiting situations that are very unlikely to occur. This is
due to the facts that:
1- There exists a feedback between prices and demand in the global economy
2- Technology/resource substitution processes occur that enable reductions of the demand for specific commodities.
Prices are not likely to remain strictly linearly increasing as in section 10.2, and the composition of the demand is
not likely to remain fixed in the future, as in section 10.3. Effectively, as depletion progresses, marginal cost of
exploitation increase and prices increase, and these induce gradual technology switching and/or reductions in overall
energy demand. Complete technology switching away from a particular fuel occurs when the price of this fuel makes
50
40
40
30
30
20
20
10
10
1980
2020
2060
2100 1980
2020
2060
0
2100
30
30
20
20
10
10
1980
2020
2060
2100 1980
2020
2060
Electricity Generation ( PWh )
50
Emissions ( Gt )
Emissions ( Gt )
Nuclear
Oil
Coal
Coal+CCS
Gas
Gas+CCS
Biomass
Biomass+CCS
Hydro
Wind
Solar
Geothermal
Ocean
Electricity Generation ( PWh )
0
2100
Figure 11: Example of changes in the power sector which could generate strong increases in the marginal cost of natural gas and uranium, in
comparison to a baseline scenario, where in the top panels electricity generation is given, while in the bottom panels the associated emissions are
shown. On the le f t panels is shown the baseline scenario, which consists primarily in assuming that current policies extend into the future. On the
right is shown a mitigation scenario where carbon pricing exists (starting at 22$/tCO2 and increasing by 2% per year ) and support for the wind
industry and CCS technology.
its use uneconomical. Therefore, the prices can never diverge as long as technology switching options exist since
switching away occurs before the price diverges at complete depletion, and thus stock resources are never depleted
entirely. Technology switching is however constrained by capital lifetimes and can therefore take some time to take
place, in particular in the power sector.
The inverse calculation problem described in section 9.3 to derive a marginal cost of production given an exogenous demand was introduced into the model of the global power sector FTT:Power, which specifically simulates
technology switching given plant lifetimes and dynamic rates of technology diffusion, described in length in [4]. It
thus provides an appropriate testing ground for this theory, an exercise that also generates direct insight on the effect
to prices of the future composition of the power system constrained by natural resources.
Figures 11 and 12 present the results of simulations performed with a version of FTT:Power that features the
above theory to derive the cost changes associated with fuel use given the resource base provided in Part I for all four
non-renewable resources, oil, coal, gas and uranium, and the values of 0 evaluated above. In order to evaluate these,
the total demand for these commodities was required, and therefore components of the demand unrelated to electricity
production had to be assumed, described below. In all other respects, the model assumptions are very similar to those
Oil
Coal
Gas
Biomass
Energy Demand ( EJ )
Nuclear
250
250
200
200
150
150
100
100
50
50
Energy Demand ( EJ )
Marginal Costs ( USD/GJ )
Marginal Costs ( USD/GJ )
0
0
2020 2040 2060 2080 2100 2020 2040 2060 2080 2100
8
8
0
2020 2040 2060 2080 2100 2020 2040 2060 2080 2100
Figure 12: T op panels exogenous assumptions for the demand for coal, gas, oil, uranium and biomass, that does not originate from the power
sector, for the baseline (le f t) and the mitigation (right) scenarios given in figure 11. Bottom panels Marginal cost of production for coal, gas,
oil and uranium that result from the global demand and the theory presented above, along with the marginal cost of biomass production which is
obtained directly from the cost-supply curve.
of the baseline presented in [4].23 These simulations were performed for two sets of policy assumptions, a baseline
scenario and a mitigation scenario. The top panels of figure 11 present the electricity generation by type of technology
for both scenarios, the baseline on the left and the mitigation scenario on the right. In these plots, the dashed vertical
line provide a visual reference to the present and the data to the left of these lines corresponds to historical data, while
on the right are given the calculated projections. Meanwhile, in the bottom panels, the associated emissions from fuel
combustion in the power sector are given, where the horizontal dashed line indicates the 1990 level.
The demand for energy commodities (coal, oil, gas and uranium) was calculated using endogenous values for fuel
demand by the power sector, and exogenous values for fuel demand from the rest of the economy. Demand values for
the rest of the economy are given in the top panels of figure 12 for each scenario. In the baseline, the demand for oil,
originating primarily in the transport sector, was assumed to peak late in the century, motivated by a gradual transition
to alternative transport technologies. The demand for gas, originating primarily in the industrial and buildings sectors,
was assumed to increase gradually up to 2100, although slowing down due to gradually increasing overall efficiency in
parallel with an increasing demand for heating services. The non-power demand for coal, originating primarily in the
industrial sector, was assumed to rapidly peak and gradually decrease due to technology switching and increased use
of natural gas. Finally, the demand for biomass was assumed to gradually increase, at a rate accelerating in the second
23 The
baseline here features no carbon pricing and no subsidies or taxes on any technology.
half of the century due to a higher diffusion of biofuels for transport. In the mitigation scenario, oil demand peaks
rapidly due to massive technology switching in the transport sector towards biofuel and electric cars. The demand for
biomass increases sharply to supply this additional demand. The non-power demand for natural gas however remains
relatively constant and the demand for coal sharply declines due to massive technology switching and electrification
of industrial processes.
The bottom panels present the resulting marginal cost of production of coal, gas, oil and uranium in these model
runs for both scenarios. In the baseline scenario, a strong increase in the cost of natural gas is observed, associated to
an increasing global demand, forcing the price to enable the extraction of shale gas and more expensive resources. This
increase is however damped past 2050 due to the large amount of resource available at these cost ranges. Meanwhile,
the marginal cost of coal hardly changes, irrespective of the sharply increasing demand, reflecting the sheer scale of
low cost coal resources. Nuclear reactors are mostly decommissioned and see a sharp decline in the baseline scenario,
leading to a decrease in the price of uranium. Finally, the price of oil increases due to increasing depletion, however
the main component of the demand does not originate in the power sector but the exogenous transport demand, the
analysis of which is outside the scope of this work.
In the mitigation scenario, strong support is given to wind energy through a subsidy (35% of the LCOE throughout
the simulation period), as well as through the pricing of CO2 emissions (starting at 22$/t and increasing by 2% per
year up to a value of 140 USD/t in 2100), while moderate support is given to electricity production using capture
and storage (CCS) technology (10% of their respective LCOEs).24 The introduction of large amounts of variable
renewable electricity into the grid requires increases in the amount of flexible type of generation, which can be
provided for by, for instance, gas turbines, oil plants or hydroelectricity. This motivates a massive expansion of the
gas turbine technology into the electricity market, which eventually dominates. The carbon pricing however motivates
the installation of CCS on all gas turbines by 2060, reducing drastically emissions. Meanwhile, the pricing of carbon
makes coal technologies gradually come out of favour, while the nuclear industry maintains a constant market share.
However, with the assumption that nuclear reactors only use natural uranium and do not recycle waste, the resource
base is seen unable to maintain the share of nuclear capacity and a strong price increase for natural uranium is
observed, which generates a decline of the nuclear industry starting at around 2070. The massive expansion of gas
turbines generates a stronger increase in the cost of gas compared to the baseline, which is however damped slightly
due to the large amounts of shale gas available. The cost of coal resources gradually decreases following the decline of
coal electricity generation. With a smaller demand for oil resources by the transport sector as assumed exogenously,
the price of oil initially increases but stabilises and decreases slightly in the middle of the century.
This section demonstrates that the combination of the theory presented above for treating dynamically the consumption of stock resources, with a model of technology substitution, enables to effectively project future marginal
costs of energy commodities given exogenous demand values from the rest of the economy. However, since technology substitution can occur in all sectors of the economy, additional flexibility in demand values exists in the global
economy. Therefore, while the power sector amounts to a very significant share of global fuel use, more accurate
calculations for energy commodity prices can be performed using a combination of this theory with a complete family
of models of technology substitution for all major fuel users of the global economy, for instance the FTT family, the
development of which will enable to remove one by one the exogenous demand assumptions given above. This is a
substantial project which is currently under way and will be the subject of forthcoming publications.
11. Supply and demand for energy resources: the feasibility of global decarbonisation
11.1. Global flows of secondary energy carriers from all sources
Flows from renewable resources were calculated in Part I, using reviews of the literature for the definition of
confidence ranges, and functional forms for the resource distributions. These are given in figure 3, and consist of
secondary energy in the form of electricity, with the exception of biomass. Biomass resources may be used either for
the production of electricity using thermal or gasification power plants, or for the production of biofuels for transport,
requiring further efficiency factors. Two cost-supply curves are therefore given in figure 13, which were calculated
24 Additional subsidies are given to biomass based electricity and solar technologies, of 35% and 50% respectively, without success. These are
effectively pushed out of the market by wind and gas turbines.
Bioelectricity
A2
800
B2 B1
A1
Biofuels
A2: 101 EJ/y

B2: 106 EJ/y
B1: 148 EJ/y
A1: 224 EJ/y
600
400
200
0
155 USD/MWh
100
200
300
400
A2
250
Cost (USD / GJ)
Cost (USD / MWh)
1000
B2 B1
200
A1
A2: 116 EJ/y

B2: 122 EJ/y
B1: 169 EJy
A1: 258 EJ/y
150
100
49 USD/GJ
50
0
100
200
300
400
500
Figure 13: Cost-supply curves for secondary biomass carriers, (le f t) electricity and (right) biofuels, both of which involve respective conversion
efficiency factors.
using the same primary biomass resource assessment of Hoogwijk et al.[22], using the four SRES scenarios A1, A2,
B1 and B2 (denoted in colour), but presented as alternative secondary energy carriers, electricity or biofuels, which
compete for the same primary resource. While the cost-supply curves of figure 3, excluding that for bioenergy but
including the left panel of figure 13 can be compared to the flows of stock energy resources of figure 9 for gas, coal
and uranium, the right panel of figure 13 may be compared to the flows of oil resources from panel a) of figure 9.
The conversion efficiency factors for bio-electricity and biofuels used were obtained using, in the first case, 2008 IEA
Energy Balances data for primary biomass use and electricity outputs [98], and the assumption of Hoogwijk et al. in
the second [22], and are of 33% and 38% respectively.
11.2. Ranges of total primary energy demand
Projections of the total primary energy demand (TPED) up to 2100 are intrinsically highly uncertain. Such calculations require many assumptions regarding global economic development, population growth and energy policy,
among many other factors. Additionally, such results depend on the type of model used for their calculation. In order
not to adopt the assumptions underlying any particular model, it is possible, and desirable, to use an aggregation of
various recent studies. Projections of energy use up to 2100 made by five European models (MERGE, E3MG, POLES,
TIMER and REMIND) have been collected in the ADAM model comparison exercise, under assumptions for three
types of scenarios according to their associated final atmospheric CO2 concentrations, of 400 and 550 ppm as well as
for a baseline scenario [1]. The resulting TPED projections are given in figure 14. Additionally, projections of the
TPED up to 2035 by the IEA [42] are given using circles for three different scenarios (which are not the same as those
adopted in the ADAM project), the 2008 value being of 514 EJ/y (or 494 EJ/y if nuclear energy is accounted as done
above in section 10.1). The fraction of TPED used for producing power projected by the IEA up to 2035 is given
using squares, the 2008 value being of 193 EJ/y which results with of 73 EJ/y of electricity. Note that in mitigation
scenarios, the TPED is expected to decrease partly due to accounting effects: since the contribution of renewables to
the TPED is counted in terms of electricity, the transfer of supply from fossil fuels to renewable or nuclear sources
reduces the accounted TPED by the amount lost in the transformation of fossil fuels into electricity.
11.3. The complete set of cost-supply and cost-flow curves as a tool for energy planning
The set of cost-supply curves for renewables, along with the set of cost-flow curves for stock resources, were
derived to be used in energy planning. By looking at the scale of resource use, and the increases in cost that additional
exploitation entails, one can devise various ways with which transfers between energy sources could occur in the
future. In particular, one can derive the effect of a potential supply of governmental support or taxation for various
Energy Demand ( EJ/y )
Projected energy demand, ADAM Project

1200
1000
Baseline
550 ppm
400 ppm
IEA Current policies
IEA New policies
IEA 450 ppm
800
600
400
IEA Power Generation
200
0
2000
2020
2040
2060
Year
2080
2100
Figure 14: Range of total primary energy demand up to 2100 derived from projections by five models as part of the ADAM project, MERGE,
E3MG, POLES, TIMER and REMIND (adapted from [1]), and projections of total primary energy demand and the fraction used for electricity
production from the World Energy Model of the IEA [42].
energy sources, shifting up or down any of these curves. As an example, a large potential of solar energy could
be unleashed, but it would require a very important amount of support. Additionally, the cost-flow curves for stock
resources indicate that increases in carrier prices are to be expected in scenarios where their expansion is projected,
in order to unlock additional more expensive resource types. These increases may produce incentives to switch to
renewables.
As one particular type of energy policy, aimed towards the reduction of GHG emissions, one may be interested
to explore the replacement of certain amounts, possibly the major part, of fossil fuels by renewable energy sources.
This transfer corresponds to the electrification of most fossil fuel consuming activities except for cases where the
combustion of bioenergy could be used. The feasibility of such plans must be evaluated using efficiency conversion
factors, summarised in table 3. Such a substitution occurs in the form of electricity as a secondary energy carrier,
and thus involves larger absolute amounts in units of energy when measured in hydrocarbon form. When replacing
electricity produced by coal, gas and oil by renewables, the conversion factors are on average of about 42%, 57% and
45% respectively [49]. However, for uses of fossil fuels where the heat is directly used (for instance the metallurgical
industries, boilers, furnaces, but with the exception of building heating), their electrification, where possible, may not
necessarily produce large changes in efficiency, and thus would likely require amounts of electricity corresponding to
their calorific content. In the case of building heating, efficiency improvements of 200-500% can be obtained by using
heat pumps, 300% being the most realistic expected value [30].
Meanwhile, if the fraction of the supply of oil used for road transport (about 54% [99]), was to be replaced by
electricity in a scenario where the global road transport fleet becomes electrified, the amount of electricity required
would likely be lower than the amount of oil used, in units of energy, due to the low efficiency of the internal combustion engine. It is however unlikely that the aviation and shipping fleets can be electrified [100], since distances
travelled without refuelling are large and the associated amounts of energy required to be kept onboard are better
provided for using biofuels rather than electricity. The improvement in energy consumption of plug-in hybrid cars,
when powered by the electricity grid, compared to reference diesel/petrol cars was found to be of a factor around 4
[101], with 0.37-0.43 MJ/km for hybrids compared to 1.8-1.9 MJ/km for internal combustion engines (note however
that the world average fuel consumption of the passenger road transport fleet is of about 35 MJ/km [100]). Their CO2
emissions were found to be lower than the references by a factor of about 2 when the electricity is produced by coal
power plants, or higher for lower emissions power systems. In the case however where oil is replaced by biofuels,
similar engines remain in use (internal combustion or jet engines), and if efficiencies do not improve, the replacement
of oil by biofuels occurs by equal amounts.

Coal
Electricity
Industry
Buildings
Land Transport
Aviation
Shipping
Other
Weighed conversion
Current use
Electricity eq.
Remaining
Total 2008
Total 2100 (EJ)
Fract.
65%
19%
4%
0%
0%
0%
12%
Gas
Conv.
42%
100%
33%
100%
60%
139 EJ
83.4 EJ
Fract.
39%
18%
23%
0%
0%
0%
20%
Oil
Conv.
57%
100%
33%
100%
68%
107 EJ
72.8 EJ
219 EJ
241 504 EJ
Fract.
7%
8%
8%
54%
7%
13%
10%
Conv.
45%
100%
33%
25%
100%
37%
170 EJ
62.9 EJ
34 EJ
Table 3: Conversion factors for electricity quantities equivalent to fossil fuels for various types of energy use. Fract. indicates the faction of fuel
use in various sectors [42], while Conv. indicates the associated conversion factor for the equivalent amount of electricity. In the process of global
decarbonisation, where most of these sectors become electric, the required amount of electricity can be calculated using these values to convert
amounts of fossil fuels into equivalent amounts of electricity. This is given for a range of projected energy demand.
11.4. The decarbonisation of the global economy before 2100

The current TPED reported by the IEA for 2008 is of 514 EJ [98]. It can be seen from figure 9, and from IEA
data [98], that about 416 EJ is supplied in the form of fossil fuels, with 139 EJ of coal, 170 EJ of oil and 107 EJ of
gas (see table 3). Out of the remaining 98 EJ, 51 EJ are supplied by biomass sources, 30 EJ by nuclear reactors25 and
11 EJ by hydroelectricity. The potential process of decarbonisation of the global economy requires a transfer of at
least these 416 EJ of fossil fuels towards renewables or nuclear energy, within their resource limits. Additionally, one
may observe from figure 9, panels a), b) and c), in comparison to figures 3 and 13, all panels, that very large amounts
of energy flows from fossil fuels (>2000 EJ/y) are available at relatively low costs, in comparison to renewable fuels.
Thus, decarbonisation is not a straightforward, since energy demand is expected to grow significantly in all scenarios
up to 2100.
In a hypothetical process of global decarbonisation, the use of coal and gas is likely to be phased out in most energy
intensive activities through their electrification, assuming that the power sector is also undergoing decarbonisation.26
The use of liquid fuels, however, can only be partially electrified since a significant amount is used in types of
transportation activities that are not straightforward to electrify (e.g. aviation and shipping). Using the conversion
factors of table 3, which enable to calculate the amounts of electricity required to replace the use of fossil fuels in
all main activities, weighed average conversion factors can be evaluated that determine the total amount of electricity
required in order to replace each type of hydrocarbons. Given the respective current use of fossil fuels, the total
amount of electricity required in order to keep all economic activities in place would be of 219 EJ/y. Thus, for
complete electrification of the global economy excluding aviation and shipping, renewable and nuclear sources would
be required to produce at least an additional 219 EJ of electricity, compared to the current production of 73 EJ/y.
This leaves 34 EJ of oil to be replaced by biofuels for aviation and shipping, which converts to approximately equal
amounts.
These values are however only lower bounds since they correspond to current energy use. As expressed in figure 14, the TPED in 2100 is likely to be within the range of 600 to 1200 EJ/y. This scales up all values by a factor
between 1.1 ad 2.3. Therefore, the additional load onto renewable and nuclear energy sources from phasing out fossil
fuels is likely to lie between 241 and 504 EJ/y, in the case where most of the transport is electrified. Since the current
amount of oil used in all forms of transport, about 126 EJ, is larger than the technical potential of biofuels in at least
two SRES scenarios, at least a certain fraction of the global land transport fleet must become electric.
As seen in figures 13 and 9 d), these values are large in comparison to the technical potentials of all low carbon
energy sources with the notable exception of solar energy. In particular, in the case where biofuels only are used in
25 Note
that this arbitrary IEA value of 30 EJ of nuclear primary fuel produces 10 EJ of electricity.
large efficiency losses could occur.
26 Otherwise
transport, the load of energy demand from transport onto the bioenergy sector is likely to be large, thus restricting the
amount available for electricity. Without subsidies, the amounts of energy available from renewables at LCOEs below
200 USD/MWh are of about 170 EJ/y for wind, 60-150 EJ/y for electricity from biomass (less the amount used for
transport), 60 EJ/y for geothermal, 40 EJ/y for hydroelectricity and none for solar or ocean energy.
11.5. The effect decarbonising the current global energy system and the associated strain on the environment
In this section, a decarbonisation scenario is created, in which all energy demand is met by renewables, in order
to explore whether such a plan is at all possible, and the effect it would have on the global environment. Powering
global land transport is much more efficient when it is electrified, and therefore, the scenario created here assumes
that it becomes so before 2050. Therefore, some of the biomass resource is available for electricity production. In
this scenario, other activities involving direct use of fossil fuels (excluding power production) are assumed to become
based onto electricity, in order to replace such fuel use by renewable energy.
Using such a scenario, it becomes possible to evaluate from figures 3 and 13 reasonable estimates for energy flows
from each renewable resource. By simply adding them up, it can be observed to first order whether the energy flows
required to achieve decarbonisation in this way, according to the last section based on the projected energy demand,
can be produced by the sum of global renewable resources.
Wind
Solar
Biomass
Total
Density
MW/km2
2.2 4
64
CF
%
2535
< 50
Density
TJ/km2
1744
6395
6
Use
EJ/y
100
20260
50
Area
Mkm2
2.35.9
0.24.1
8.3
10.818.3
of global
%
1.54.0
0.1 2.8
5.6
7.212.3
Table 4: Ranges of global land use by wind, solar and biomass energy given energy demand values. The global land area is of about 149 Mkm2 .
Following the lowest estimate of the biomass technical potential in the A2 scenario, a flow of around 50 EJ/y can
probably be realistically expected in all scenarios. Due to cost considerations, flows of no more than 40 EJ/y are likely
to be obtained from hydropower. Geothermal energy may generate a flow of about 20 EJ/y, and wind power can be
expected to produce a large amount, possibly around 100 EJ/y. Marine energy can be expected to provide at most
10 EJ/y at relatively high costs. Excluding solar energy, this results in a flow of renewable energy of 220 EJ/y. Given
the likely requirement of between around 240 and 500 EJ/y, the possible remaining energy demand must be provided
for by costly solar energy, by a value between 20 and 260 EJ/y, possibly using government support to help it bridge
the valley of death through learning-by-doing and become competitive [102].
Thus decarbonising the current global energy system appears possible. However, this is not done without a cost
to the natural environment. These approximate large numbers signify important underlying changes to the local
environment where these energy flows are produced. As stated in sections 4.1, 4.2 and 4.5 of Part I, values of the
energy productivity of the land for wind, solar and biomass energy are low, therefore producing large energy flows
from these sources involves large areas of land devoted to energy production activities. The replacement of a 1 GW
coal power station by wind energy is likely to require between about 700 to 1100 turbines covering an area of between
500 to 1500 km2 . Meanwhile, in the case of solar energy, this would require an area of between 50 and 500 km2
covered in solar panels, and in the case of biomass this becomes of the order of 5000 km2 . In order to decarbonise the
current level of energy use without the involvement of fossil fuel use in conjunction with carbon capture and storage
(CCS) technology, a global capacity of between 3600 and 7600 GW of coal, gas and oil power plants (scaling up data
from [42] by between 1.1 and 2.3) are likely to be required to be replaced by renewable energy systems.27
Assuming a future where 100 EJ/y are produced by wind power, 20 to 260 EJ/y by solar energy and 50 EJ/y by
biomass, the respective land requirements would be of about 2.3 to 5.9 Mkm2 , 0.2 to 4.1 Mkm2 and 8.3 Mkm2 , for a
27 This does not take consideration of the problem of introducing massive amounts of variable renewables into the current energy system, which
requires available flexibility which currently can only be provided by gas turbines. An appropriate approach to treat this problem was published
elsewhere [4]. Large amounts of energy storage or demand management is therefore assumed possible here, of which admittedly, the nature is not
known. The result is decreasing capacity factors, requiring additional capacity. This will be explored elsewhere.
total between 10.8 and 18.3 Mkm2 (see table 4). Since the total onshore land area covers approximately 149 Mkm2 ,
this represents 7 to 12% of global land area to be dedicated purely to energy production. According to the Global Land
Cover 2000 assessment [103], 15.7% of the global land area was devoted to agriculture in 2000. Although energy
production activities need not be restricted to such areas but may be done in uninhabited regions such as deserts,
transmission cost considerations will often generate the incentive to use areas in the vicinity to urban centres, located
predominantly within agricultural areas. Thus agriculture would not be in competition uniquely with biomass energy
crop production, but also with wind and solar power.
Thus, without significant reductions in energy demand before 2100, the use of land for energy purposes is likely
to increase to such an extent that a significant fraction of the global inhabited land area would become dedicated to
such activities. Thus, reducing energy consumption below the current supply of 500 EJ/y with an increasing global
population is a challenging but necessary task, especially given that in developing regions with high population growth,
energy efficiency (energy intensity) tends to be much lower (higher) than in developed economies. CCS technology
may also be considered as a useful tool to restrict land use from renewables. However, the key to energy use reductions
lies most probably in large parts in improved energy efficiency, in reduced energy use per unit of service produced,
an area that is currently poorly covered in the current literature, where often energy efficiency is represented using
simple extrapolations of past global averages such as the Autonomous Energy Efficiency Index (AEEI). Since climate
change mitigation policy does not have a historical precedent, and changes of the magnitude required for limited
anthropogenic climate change have never been attempted in history, it is hardly justified to use a historical trend to
project future energy efficiencies. In contrast, these should be evaluated using detailed engineering energy models
with feedback to a global macroeconomic model.
The extent of land use for energy production may threaten the ability to preserve natural areas in order to protect
ecosystems and biodiversity. Thus, significant additional modelling work should be done in order to address the feasibility of maintaining both land intensive energy production activities from renewables without significantly affecting
agriculture and food production or ecosystems and biodiversity, in terms of global land use. This could be achieved
with a land use model where the allocation of the land responds to the global energy demand as well as the food
demand.
12. Conclusion
The second part of this work provides tools for scientists and policy makers to evaluate secondary energy flows
from all major types of natural resources and compare them to one another. Part I of this work presented the calculation
of cost-supply curves for all these resources, in terms of energy flows, for renewables, or fixed amounts, for stock
resources such as fossil and nuclear fuels. Since these are not directly comparable, this paper provides a theoretical
framework for converting stock resources into flows and depletion timescales. This model is used to determine both
energy flows given exogenous prices, or to determine prices given exogenous energy demand values. It is designed,
however, to be used dynamically into an E3 model that features a strong feedback between the economy and the energy
system, and enables to track the progression of the depletion of stock resources, where price increases are avoided
through technology substitution. Calculations are performed using this model in both limiting cases of exogenous
demand or exogenous prices for oil, coal, gas and uranium. In both approaches, uranium resources are found to
become depleted within this century, while coal resources are large enough to last far beyond 2100. Meanwhile,
the accession to unconventional oil and gas resources at certain price ranges can enables massive expansions in their
exploitation, or conversely, can damp out price increases. Reality however lies between these limiting situations of
either rigid prices or demand, where rising or falling prices are avoided by technological change. This provides a
mechanistic equivalence to elasticities of substitution.
The feasibility of global complete decarbonisation is explored on grounds of available resources. In particular, the
question of whether there is sufficient renewable electricity to replace fossil fuel use globally is examined. Projections
of energy demand up to 2100 are given, derived from a study involving five existing E3 models. Conversion factors
are given, with which an approximate range is derived for the amount of electricity likely to be required for replacing
fossil fuels in the case of complete decarbonisation by 2100. This amount is found to be large in comparison to the
technical potentials of most renewable resources, with the exception of solar energy. For a particular scenario of
renewable energy use in 2100 in a decarbonised world without significant effort for energy demand reductions, the
scale of land use by wind, solar and biomass energy lies in the range of 7 to 12% of global land area. This emphasises
the requirement for large energy demand reductions in order to achieve global decarbonisation without significantly
disturbing ecosystems and agricultural activities.
Part III
Appendix
13. Introduction and use of the Appendix
This part of the work aims to complement the main paper by providing all details that could be required by
researchers interested in either:
Verifying the methodology or reproducing the results
Rebuilding this database for use in a particular model of energy systems
A high amount of care was put into summarising compactly all relevant information in this part of the work in order to
make this possible. Mathematical details underlying the calculations given in the main text are provided. Additionally,
tables of data and parameters are given for a chosen set of world regions, which may not necessarily correspond to
the particular divisions used by other research groups. It is however impractical to provide larger tables involving all
countries of the world, even though such tables exist underlying this work (190 countries exist in our database). For
more information, the authors may be contacted at the address provided.
Natural resource assessments are performed continuously and what is known of global natural resources changes
continuously. Therefore the cost-supply curves in this work may become outdated. This will happen due to three
processes: firstly the total amount of resources might change (parameter A), secondly the scaling of costs may change
through inflation (parameter B) and the costs of technology may reduce due to learning-by-doing (parameter C0 ).
However, the structure of the cost-supply curves, or shape, will not change. These parameters should be simply scaled
to new values and the results will still be valid.
14. Derivation of distribution functions and cost-supply curves
14.1. Distribution function for the hierarchical type of resources
Hierarchical resources have an exponential energy distribution in productivity space:
( A
e d >
,
f ()d =
0
(8)
where A is the technical potential and is the half width of the function. This function is required in cost space, and
the equation connecting cost C to productivity is
C=
Cvar
+ C0 .
(9)
where Cvar corresponds to costs per unit of effort or per unit of resource producing items such as the rent of the land
(in $/km2 ), bore hole depth ($/km), dam size and type, mine depth, etc. The ratio of Cvar to has units of $/GJ. C0 is
the sum of fixed costs, such as capital investments, transport or transformation costs, etc, in $/GJ.
The density productivity interval must be transformed into a cost interval:
dC =
Cvar
d,
2
d =
Cvar
dC.
(C C0 )2
Using the cost scaling parameter B = Cvar /, the distribution becomes:
B
CC
AB
0 dC
(CC
C > C f ixed
2e
0)
f (C)dC =
.
0
C C0
(10)
(11)
14.2. Distribution function for nearly identical resources

In the case of nearly identical resources, there is no convenient exact analytical form that can be derived from eq. 2
of the main paper. However, the form given in eq. 4 works extremely well when compared to data (see section 14.5),
and can be derived from eq. 2 through a simple approximation.
Nearly identical energy producing resources, such as land plots, are assumed truly identical, and therefore have a
potential situated at a single value of productivity = ,
n()d = N( )d,
(12)
where n is a density of energy producing land area, while N is the total energy producing land area (in km2 ) and
the function () is the Dirac delta function.28 Without any additional reductions in productivity, the total amount of
energy that can be obtained from these land resources would be their area times the productivity:
Z
A=
N( )d = N.
(13)
0
Unit land areas have a suitability factor, however, that reduces their productivity below the maximum value of by
a small amount with a probability P. The probability for the reduction in productivity is assumed to be normally
distributed around zero but positive , with standard deviation much smaller than the average productivity << :
2 e 22 d 0
2
,
(14)
P()d =
0
otherwise
where the reduction in productivity must be less than the maximum value . The distribution of resources must be
calculated by summing over all reduction values given their probability P():
Z
Z
2N 22
(15)
n ()d =
n()P()dd =
e 2 ( + )dd.
0
2
This is a convolution and can be seen as a sum of several Dirac Delta functions centred at slightly reduced values of
productivity, , with probability P(), instead of one Dirac Delta function centred at with probability 1. The total
amount of energy that can be obtained from each plot of land corresponds to its area times its productivity. Thus, the
productivity distribution of energy production potential leads to eq. 2 of the main paper:
()2
2N e 22 d
.
(16)
g()d = n ()d =
2
0
>
This function is required in cost space, and the equation connecting cost to productivity is
C=
Cvar
+ C f ixed ,
(17)
where Cvar corresponds to the rent of the land (in $/km2 ), while C f ixed is the sum of fixed costs (in $/GJ). The
productivity is situated very near the value of , since the variations of productivity are small and << . can be
rewritten as a small variation around , i.e. :
!
Cvar
Cvar /
Cvar
Cvar
1+
+ C f ixed = 2 (2 ) + C f ixed ,
+ C f ixed =
+ C f ixed
C=
(18)
1 /
which is the crucial approximation, and using C0 = Cvar / + C f ixed ,

= (C0 C)
28 The
Dirac delta function is defined such that
2
+ ,
Cvar
(19)
(x a) f (x)dx = f (a).
where C0 is defined as the sum of fixed costs plus Cvar /, the total cost at the maximum productivity value. Since
dC = Cvar /d, the density can be rewritten as

(CC0 )2
2N (C0 C) C + e 2B2 dC C > C0

var
2B
,
(20)
g(C)dC =
0
C C0
where the cost scaling parameter B is defined as Cvar /2 . This can be rewritten further as
(CC0 )2

2N C0BC + 1 e 2B2 dC C > C0

.
g(C)dC =
2B
0
C C0
(21)
The value of C cannot be below C0 by definition, but also, the factor exp[(C C0 )2 /2B2] decreases rapidly to zero
as C C0 becomes larger than B. Therefore, the value of the term (C0 C)/B is less than one wherever a significant
potential of energy exists. Since is much smaller than , this results with
C C0
<< 1,
B
(22)
and therefore the distribution becomes
g(C)dC =
2A e
2B
(CC0 )2
2B2
dC
C > C0 ,
C C0
(23)
where now A = N is approximately the total energy potential.

This is the strict region of validity of the expression given in 4 of the paper. In numerical terms, it is empirically
found that the rigidity of these rules can be relaxed and the validity extended. For example, the distribution in
productivity space can actually have a tail towards higher values or have a large value for , and this does not
significantly alter the goodness of fit of the function in cost space.
14.3. Cost-supply curve expressions
From the distributions f (C)dC and g(C)dC, cumulative distributions N(c) can be derived. For hierarchical resources, this results in

N(C) = A e
B
CC
(24)
while for nearly identical resources this is

!
(C C0 )
,
N(C) = A erf
2B
(25)
where erf is the error function.

The cost-supply curves are the inverse of these functions, which respectively give
C(N) =
and
C(N) =
B
+ C0
ln NA
N
2B inverf
+ C0 ,
A
(26)
(27)
where inverf is the inverse error function.
14.4. Parameterisation formulas

In order to define a particular distribution or cost-supply curve, the parameters A, B and C0 are required, while the
data available usually involves the total technical potential of the resource and a fraction of this considered to exist
at costs situated between two values, as well as the current level of use of the resource. Assuming that two points of
the curve are known, this may be expressed as two quantities Q1 and Q2 which occur at two cost values C1 and C2 .
These quantities can be expressed as fractions of the total technical potential, 1 and 2 , the latter corresponding to
the parameter A. In the case of the distribution for hierarchical resources, the values of B and C0 are the following:
C0 =
C2 ln 2 C1 ln 1
,
ln 2 ln 1
B = (C1 C0 ) log 1 .
(28)
(29)
In the case of nearly identical resources, this becomes

C2 C1
,
B=
2(inverf 1 inverf 2 )
C0 = 2B inverf 1 + C1
(30)
(31)
14.5. Demonstrating the validity of the functional forms using IMAGE data
Examples of the use of the analytical forms of the distributions are presented in figure 16.1 for biomass, solar
and wind energy. The data are taken from land use simulations performed using IMAGE by Hoogwijk et al., which
provide the only sources of cost-supply curves calculated outside of this project that do not already use assumptions
on the analytical form of the resource distribution [18, 19, 22]. Since IMAGE simulates the use of the land on each
point of a global grid, and since these cost-supply curves were calculated by building histograms of the number of
grid points with productivities situated within various ranges, their form stems purely from the statistical nature of the
data. These are thus appropriate for testing the functions given above.
Non-linear least-squares fits were performed with both analytical forms for each data set. In every case, only
one of the two functions given above is appropriate, while the other is not. Fits are moreover of exceptional quality.
For instance, wind resources are the best example of resources of the hierarchically ordered type, which stems from
the exponentially increasing number of simultaneous factors required to produce ever higher productivities. The
data is found to follow almost exactly the distribution for hierarchical resources (note that the deviation at low cost
values stems from the aggregation of a region with a different cost structure into the region for Canada). Meanwhile,
solar resources represent the best example of nearly identical resources, since in regions of similar irradiation, all
sun-facing areas are equivalent. The data is found to follow closely the distribution for nearly identical resources.
Biomass resources from abandoned agricultural land are nearly identical. This stems from the similar nature of local
areas of agricultural land (i.e. large plains, deltas, similar irradiation, etc). Land plots with lower productivity are used
for other activities. Rest land, however, is the category of land which would not be used for agriculture, and can be
of various nature, but includes mainly savannah, shrubland and grassland or steppe. These can be ordered, and can be
seen to follow the distribution for hierarchical resources.
15. Cost-supply curve calculation methodology per resource type
15.1. Definition of world regions
Cost-supply curves were calculated in this work for every E3MG world regions from aggregations of data defined
for 190 countries. However, the region definition in E3MG is very specific and does not correspond closely to that of
most other global models, and tables provided here for E3MG regions would be of limited use to the global modelling
community. For accuracy, data for 190 countries would be required to be provided here, but is not possible for space
considerations. For the convenience of potential users, the results are provided in tables with a definition of regions
resembling that of other models such as IMAGE, AIM, etc. Any other aggregation of data, in table or curve form, can
be supplied by the authors upon request. Table 16.1 gives the list of regions used here with most countries that belong
to them.
15.2. Wind and solar energy

Following the justification of section 14.5, wind resources were modelled using a distribution of hierarchical type,
while solar resources were modelled using a distribution for nearly identical resources. In both cases simulations
performed by Hoogwijk et al. (wind energy), Hoogwijk (PhD thesis, wind, solar and biomass energy) and de Vries
et al. (both) using IMAGE 2.2 were used, published in the form of data tables featuring both technical potentials and
interpolations through cost supply curves at specific cost values for a list of 17 world regions [18, 19, 21]. These
values were used to find the distribution parameters A, B and C0 for every one of their regions. In the case of wind, A
values were thus obtained without additional processing. In the case of solar, A values were obtained from de Vries
et al. while B and C0 values from Hoogwijk [18, 21]. In both cases, B and C0 values were obtained using equations
28 to 31.
However, the regional aggregation in the aforementioned work does not match exactly the one chosen in the
current study (or the one used in E3MG), detailed in section 15.1. In order to obtain curves for this set of world
regions, energy potentials from IMAGE 2.2 regions were disaggregated into 190 countries, and subsequently reaggregated. This required additional assumptions in particular cases where A values were required to be divided
between underlying countries29 . In the case of wind energy, the division of A values was done proportionally to the
cube of the yearly averated wind speed30 , times the amount of land suitable in each country for these energy production
activities (the land area times the suitability factor provided by Hoogwijk et al., assumed the same for all countries
member of a region) [104106]. In the case of solar energy, A values were divided proportionally to the insolation
averaged over countries times the amount of land suitable in each country [104, 106, 107]. Assuming an identical
shape for the cost supply curves (identical values of B and C0 ) for every country within a particular IMAGE region,
and using A values thus divided, cost supply curves for the 190 countries were built. Given this set of curves, the
re-aggregation of curves into new world regions was performed by summing the energy potential values at each cost
(i.e. a sum along the horizontal axis of the cost-supply curve, called a horizontal sum henceforth). These aggregated
curves do not correspond anymore to pure distributions of either type, but do not differ significantly from pure forms
in any of the regions chosen for this work. Thus, new values for A, B and C0 for this works regional definition were
re-estimated using equations 28 to 31, for the sake of simple presentation in this work (avoiding listing parameters for
190 countries, or providing aggregate curves defined on large numbers of cost data points). For E3MG, data curves
evaluated on 1000 cost data points are used directly instead.31
Cost values with which cost-supply curves were calculated using equations 28 to 31 were also obtained from
Hoogwijk et al., but were rescaled to 2008 prices [18, 19]. This procedure, however, generates costs of energy
production slightly different (wind) or higher (solar) than recent estimates available from the International Energy
Agency, due to small errors (wind) or significant learning-by-doing cost reductions (solar) stemming from economies
of scale with large expansion of electricity generation capacities that occurred between 2004 and 2008 [49]. The
curves were therefore recalibrated with a constant offset to match recent values. The results are provided in table 16.3
for this works list of world regions.
15.3. Hydropower
Hydroelectric resources, highly site dependent, were modelled using the distribution for hierarchical resources.
Hydroelectric potentials and current annual electricity generation values were obtained from the last available technical report of the International Journal on Hydropower and Dams (IJHD), while the costs were obtained using an
extensive study of 250 recent projects by Lako et al. from which statistics were derived [47, 48]. These statistics were
performed for the countries studied in Lako et al., and were used as proxies for regions not studied in their work, or
where no information on recent hydroelectric developments was found. Some countries do not have recently reported
hydroelectric projects onto which to base cost values.
Recent developments have hardly followed an order of cost, since they were scattered between 500 and 4000
2003USD/kW. In order to use a cost-supply curve, it can only be assumed that future developments actually will
29 Note
that this is mostly true for E3MG regions; the regions used for this paper are very similar to those used by Hoogwijk et al. [19]
energy scales with the cube of the average wind speed averaged over time (see for instance [27]).
31 Exact analytical forms for cost-supply curves correspond to the inverse the cumulative distribution. When the cumulative distribution involves
the sum of several distributions, an analytical form for the cost-supply curve does not exist.
30 Wind
approximately follow a cost order. Although only approximately true, this is reasonable, since development costs
will significantly increase when more and more usable sites are depleted, irrespective of the particular order in which
they were built, and only difficult or distant river basins remain. This is important since the costs of hydroelectricity
are currently not high in comparison to alternatives, but the resources are limited, and therefore the development of
hydroelectric resources must be limited through an increasing cost in models of power systems such as FTT:Power.
As can be seen in the current hydroelectricity generation data compared to the data for hydroelectric potentials
from IJHD, a significant portion of the technical potential of every region is already developed [15, 47]. The cost
values delimiting the technical and economic potentials amongst remaining potential hydroelectric sites are not given
by IJHD. Since the distribution of costs is not symmetrical, the assumption was taken that the amount of resources
considered economical lies at costs between the local average cost minus its standard deviation , , and plus
twice its standard deviation + 2. This puts the upper cost limit to around 5000 2008USD/kW. Thus, sites within
the technical potential with costs higher than this are considered currently uneconomical. Given this definition, a
cost-supply curve for each region was calculated. Parameters for each regional cost-supply curve are given in table
16.3. The global cost-supply curve of figure 3 of the main text is an aggregation (a horizontal sum) of these regional
curves.
15.4. Geothermal energy
Geothermal resources were divided into two groups, occurring in either in belt or out of belt land areas, referring to the so-called volcanic belt. In belt areas are located in volcanically active zones with high geothermal
gradients (temperature gradients with bore depth from the surface of the ground). Given the particular characteristics
of geothermal active areas in terms of their heat storage and underground temperature variation, the extraction of
geothermal resources in those places are highly site-specific, and were therefore modelled using a hierarchical distribution. Out of belt areas corresponds to the rest of the continental masses, with sites that are characterised by
smaller geothermal gradients, and that are almost identical to one another within large geographical areas. Out of
belt resources were thus modelled using a distribution for nearly identical resources. The ratio of in belt to out
of belt land area values were obtained from the 1978 report of the EPRI [57], enabling to divide reported technical
potentials into two A values for each distribution type. Geothermal resources were moreover calculated for both hydrothermal and EGS dry rock technologies, yielding four sets of parameters. Each cost-supply curve in each region
was obtained by aggregating four curves.
Technical potentials for different world regions were obtained from Bertani [63]. Given the differences between
their regional aggregation and this work, the same methodology was used as for wind and solar energy in order to
disaggregate the regional technical potentials between the same 190 countries. The proportion of the regional technical
potentials assigned to every country within a particular region was assumed to be proportional to the total amount of
geothermal energy stored up to five kilometres of depth in each country [56].
Cost values for geothermal electricity production were taken from the IEA [64]. It was assumed that 90% of the
resources in belt were situated within these ranges of costs. However, resources out of belt follow the distribution
for nearly identical resources, but face higher costs due to lower geothermal gradients and less productivity per unit
investment. Since no additional cost information was available in this regard, these resources were assumed to lie in
the upper half of the cost range given by [64].32 Table 16.5 gives the parameters that can be used to reproduce these
cost-supply curves using both types of distributions.
The lower boundary curve of the uncertainty range assumes a technical potential of 4 EJ/y based on differing
assumptions for both technologies. In the case of hydrothermal technology, a conservative potential estimate of
70 GW (2 EJ/y) was derived by limiting the calculation to well known sites that have been already characterised by
direct involvement or informed calculations [62, 63]. Meanwhile, the limited amount of accumulated experience with
EGS technology creates uncertainties in the evaluation of the technical potential through variations in the efficiency
of extraction [55], leading Bertani to estimate a lower limit of 70 GW (2 EJ/y) [63]. The upper boundary of the
uncertainty range involves yet another set of assumptions for both hydrothermal and EGS technologies. In the case of
32 In out of belt areas, the same technologies are involved, either for hydrothermal or EGS, as for in belt areas. However, the resources are
nearly identical over large areas and of equally low quality in comparison to in belt areas. Significantly higher productivities are found in volcanic
areas.
hydrothermal, according to Stefansson, undiscovered or additional resources could exist which would be five to ten
times higher than identified resources, increasing the potential to 1000-2000 GW (57 EJ/y) [60]. In the case of EGS,
the technical potential is calculated by an extrapolation of resources in the United States to the global level using the
proportion between the heat stored at depths of less than 10 km in the United Stated with the known EGS primary
energy potential in the same region, estimated as 1106 EJ of heat stored per 2.61 EJ/yr of EGS primary energy
potential [55]. Using the estimation of the heat stored at depths less than 10 km on the global scale of 403106 EJ
[108], this estimation results in a global technical potential of 54 EJ/yr, yielding a total of 111 EJ/y.
15.5. Bioenergy
Four bioenergy cost-supply curves are given in figure 3 of the main text, for each of the SRES scenarios A1, A2,
B1 and B2, based on simulations performed using IMAGE 2.2 [18, 20, 22] (see [72] for information on SRES scenarios). The primary biomass energy sources considered in the cost supply curve are abandoned agricultural land, rest
land and bagasse, where the first is the largest source in all scenarios. Following the justification of section 14.5, abandoned agricultural land was modelled using distributions for nearly identical resources, while rest land was modelled
using distributions for hierarchical resources. The remaining type of primary biomass resources, bagasse (from [15]),
contributes very small fractions of the total potentials, and its technical potentials were simply added to the potentials
from abandoned agricultural land and rest land, for every region in every scenario. Cost values, however, are only
given by Hoogwijk et al. for the total amount of biomass resources in each region, not individually for abandoned
agricultural and rest land [22]. Therefore, the right distribution to use had to be determined, by deciding which of the
two represented best the data. This corresponds to finding the dominant distribution. Therefore, the appropriate type
of distribution was determined by visual inspection for each region. Potentials for abandoned agricultural land are
for most regions much larger than those for rest land, and therefore most regions were modelled using distributions
for nearly identical resources. These distributions were disaggregated into 190 countries, following the methodology described in section 15.2, proportionally to country land areas times their suitability factor. Table 16.4 provides
values that can be used to parameterise biomass cost-supply curves for the world regions used in this work, with the
appropriate type of distribution used indicated in the last column.
15.6. Ocean energy
Given the vast extent of oceans, the calculation of theoretical potentials for ocean energy sources produces large
values. For instance, using a global wind-wave model, Mork et al. estimated a potential for wave energy between
2986 and 3703 GWe (94 to 117 EJ/yr), while Charlier and Justus estimated a global theoretical tidal power potential
between 1000 and 3000 GWe (32 to 95 EJ/yr) using a capacity factor of 100% [77, 109]. In the case of ocean thermal
energy, Pelc and Fujita estimated a theoretical potential of approximately 10 TW (315 EJ/yr) using a capacity factor
of 100%, while for salinity gradient energy, Cavanagh et al. calculated a value of 2.6 TW (82 EJ/yr) using a capacity
factor of 100% [80, 110]. Using values from these particular studies, the total theoretical potential for ocean energy
would be as high 523 and 609 EJ/yr. However, more reliable and conservative potentials have also been evaluated,
given below. These values are modest in comparison. As indicated in the main text, cost-supply curves were calculated
for wave and tidal systems only. For presentation in this work only, these two cost-supply curves were combined into
a single one for ocean energy. Parameters for regional ocean cost-supply curves are given in table 16.3.
Wave Energy
In the case of wave energy, WEC estimated a maximum global installable capacity of 2 TW by limiting developments to technically favourable locations near coastlines [34]. Using this value, and assuming a single capacity factor
value of 32%, Krewitt et al. estimated a technical potential for wave energy of 20 EJ/yr, while UNDP estimated a technical potential of 65 EJ/yr using the same value but assuming a capacity factor of 100% instead [11, 16]. Following
a more conservative approach restricted to shorelines exceeding a power production of 30 kW/m (resulting in around
2% of global coastlines), Sims et al. estimated a technical potential of 500 GW (6.3 EJ/y using a capacity factor of
40%) [75]. It is clear however that using single capacity factor values is not appropriate. It is likely that, as it is the
case for wind, the cost variation of a wave energy cost-supply curve should stem from capacity factor variations which
stem from the local quality of the resource. Such data is however not currently available as it is for wind resources.
Capacity factor distributions are likely to follow roughly those of wind energy, which vary between 15 and 35%, since
both resources are closely related (wind capacity factor distributions were obtained by extracting the capacity factor
from Hoogwijk et al. data [19]). The assumption was therefore taken in this work that wave energy resources are
captured using a single technology, with an investment cost given by ETSAP of 6600 USD/kW, with a capacity factor
that varies between 35% (where the resource quality is highest, and the cost of electricity production is lowest per unit
energy produced) to a low value of 15% (below which sites are not economically useable) [73]. Using a maximum
global capacity of 2 TW, the cost-supply curves were calculated with a hierarchical distribution, assuming that 90% of
the wave resources are available at capacity factors within the range 15-35%. The disaggregation into 190 countries
was performed according to the lengths of their respective coastlines, using data from the Central Intelligence Agency
(CIA) [106].
Tidal Energy
Accounting for most of the global installed capacity of ocean energy systems, tidal energy is the only technology
that has reached a commercial scale, with approximately 523 MW installed at the end of 2010 [111]. WEC made
a rough estimation of the technical potential of tidal energy of about 2000 TWh/y (7.2 EJ/y), 10% considered economical [34, 112, 113]. In a more detailed study, Hammons presented a global but non-exhaustive list of potential
tidal sites that could be considered for development, including projected installed capacities and approximate annual
outputs [76]. The total output from these sites would be of almost 400 TWh/y (1.4 EJ/y). Hammons furthermore
extrapolated that the inclusion of additional sites around the world not studied specifically in his work would result
in a global technical potential for tidal energy likely to range between 500 and 1000 TWh/yr (1.8 3.6 EJ/y). The
cost-supply curve for tidal energy was calculated using the range of cost values given in ETSAP of 5000 to 6500
USD/kW [73]. Existing capacity was assumed to have been built at costs below that range, while the sites reviewed
by Hammons (400 TWh/y) were assumed to be associated with costs within the range [76]. Additional sites were
assumed to have costs above the range.
Ocean Thermal and Salinity
The state of development of ocean thermal and salinity gradient energy technologies is currently experimental and
therefore large uncertainties accompany calculations of associated energy potentials [75]. Upper limits in the form
of theoretical potentials have been calculated. Nihous estimated a theoretical potential for ocean thermal energy of
2.7 TW (85 EJ/yr or 23 652 TWh/yr), which corresponds to the maximum amount of energy resources that could be
extracted without disrupting significantly the temperature of the upper layers of the ocean in an steady state regime,
using a one-dimensional model of oceanic temperature gradients [78]. Using a similar method, Charlier and Justus
produced a more conservative estimation for the theoretical potential of 1000 GWe (32 EJ/yr), assuming a capacity
factor of 100% [77]. However, according to von Arx, such a level of heat extraction would imply a decrease in
the ocean surface layer temperature of approximately 1 C [114]. In order to avoid such a decrease, Charlier and
Justus recommend a reduced estimate based on 10 TW of usable heat replenishment rate, corresponding to 100 GWe
(3.2 EJ/y) [77].
In the case of Salinity Gradient, based on average discharge and low flow discharge values, Skramesto estimated
the theoretical potential in the range of 1600 - 1700 TWh/yr (5.8 - 6.1 EJ/yr) [79]. Using a global discharge rate of
fresh water to seas of 44 500 km3 per year, Krewitt et al. estimated a theoretical potential of 2000 TWh (7.2 EJ/yr),
value very similar to the estimate of Skramesto [16, 79].
15.7. Oil
Oil resources (Table 16.6) were considered in four types of occurrences, crude oil, oil shales, extra-heavy oil and
oil sands [13, 15]. Cost information was obtained from the IEA [82]. The data were aggregated into this works
world regions. For each type of occurrence for each region, a hierarchical distribution was parameterised by assuming
that 1% of the resources have extraction cost below the lower bound, while 90% have a cost of extraction below the
upper bound. The distributions were summed for each region in order to calculate regional cost-supply curves, and all
distributions were summed in order to determine the global cost-supply curves given in figure 4 of the main text.
The curve for the lower boundary of the uncertainty range was defined by assuming that only crude oil, extra-heavy
oil and oil sands reserves are available, and that the rest is either unusable or does not exist. The most probable costsupply curve was calculated assuming that crude oil, oil sands and extra-heavy reserves and resources are available,
as well as oil shale resources, but no additional amounts. The curve for the upper boundary of the uncertainty range
assumes that all reserves, resources and additional amounts are available, and that an additional amount of oil shales
is discovered, evaluated at 50% of the current resources. This was done in order to compensate for the absence of
speculative resources and lack of detailed information available for oil shale resources, which are likely to become
larger if additional exploration is carried out, and will occur if (but only if) interest in oil shales intensifies.33
15.8. Natural gas
Gas occurrences were considered in five forms, of which four unconventional: conventional gas, shale gas, tight
gas, coalbed methane and methane hydrates [11, 13, 8385]. The associated cost ranges were obtained from the ETSAP [86]. Of the unconventional forms, only shale gas has seen exploitation larger than experimental. An additional
source of methane exists, which is thought very large, aquifer gas [11]. However, its potential being very speculative,
no reliable information over costs of extraction was found, and thus these were not considered in the present study.
Similarly, methane hydrates provide a very large source of natural gas, however, these resources occurring under the
sea, and the methods of extraction very experimental, the costs of exploitation are very large, and due to large amounts
of shale gas available at lower costs, it is unclear whether the world will see wide-scale exploitation of methane hydrates. All resources were distributed into this works world regions, except for the methane hydrates, for which it
is not clear whether they are situated within territorial waters or not, and thus were assigned to an international category. Regional cost-supply curves were calculated with the same method as oil, and the global cost-supply curve is
an aggregation of all regions.
The curve for the lower boundary of the uncertainty range includes conventional gas reserves only. The most
probable cost-supply curve includes conventional, shale and tight gas reserves, along with half the conventional, shale
and tight gas resources. It moreover includes half of the coalbed methane reserves. The curve for the upper boundary
of the uncertainty range includes all reserves and resources, including methane hydrates.
15.9. Coal
Although coal is a very common commodity and well known resource, information over its natural occurrences is
not very detailed. Coal information was available from two sources, BGE and WEC, and table 16.8 was constructed
using a mixture of both [13, 15]. Where information was inconsistent, the larger amounts were kept (such inconsistencies were not frequent nor very large). Since BGR does not report coal resources in the complete classification (proven,
probable and possible reserves or resources) for all countries, some elements of the table are nil [13]. This situation
is likely to be due to the large amounts of coal available with conventional mining techniques, and therefore most of
the resources are considered reserves, and occurrences with lower productivity or higher costs are not reported. Coal
formations occur in different forms which have different calorific contents. For similar mining and transport costs,
the costs of coal in terms of energy produced are higher for lower grade coals. Coal resources were divided into two
categories, hard coal, including anthracite and bituminous coal which posseses higher calorific contents of between
16 500 and 35 000 kJ/t, and soft coal, including sub-bituminous coal and lignite, with calorific contents between about
11 000 and 16 500 kJ/t [13]. Note that these classifications are not strictly well defined in geological terms, and that
coal occurrences exist that have intermediate properties. This stems from different geophysical processes taking place
during the slow formation of these hydrocarbons.
The curves for the lower boundary of the uncertainty ranges for soft and hard coal include proven reserves only.
The most probable curves include proven and probable reserves, and half of the proven and probable resources. The
curves for the upper boundary of the uncertainty ranges include all proven, probable and possible amounts for both
reserves and resources.
15.10. Uranium
Information for uranium occurrences is available from a survey of the IAEA [26], which compiles data provided
by all member countries. They are highly detailed with four classifications and four cost ranges, as seen in table
16.9. As the amounts in each row of the table are cumulative (i.e. the data in one cell is inclusive of the sum of
33 For instance, if strong decarbonisation policies are implemented globally, oil shales are not likely to be explored much further, since they
currently involve large processing costs and only small scale exploitation.
the cells to the left), with the associated cost values they correspond to cumulative distributions. Assuming that they
should follow hierarchical distributions, the associated cumulative distribution may be fitted using a non-linear least
squares method. Although a fit of a function with three parameters over four data points is hardly a reliable method
to determine a best fit with any level of certainty, it nevertheless produced the best curves that could be interpolated
between points, as was determined by close inspection of each fit. Note that additional data points were be defined
in order to constrain the fits better, such as additional values at higher costs by repeating the last data point, assumed
equal to the technical potential in the saturation region, and at (0,0). The resulting fits were found to follow the data
very closely. Values for military stocks of U are uncertain, since the information that is publicly available is scarce
and incomplete, and were omitted. These are very small in comparison to natural stocks [26].
The curve for the lower boundary of the uncertainty range includes only RAR (reasonably assured reserves) in
all cost ranges. The most probable curve includes RAR and inferred reserves. The curve for the upper boundary of
the uncertainty range includes all four classifications of resources in all cost ranges. The speculative resources in the
unassigned cost range were not included in the non-linear fitting procedure, since their distribution into existing or
higher cost ranges is ambiguous. They were therefore added to the technical potential of the upper boundary of the
uncertainty range. Finally, it was assumed that sea water U is too costly and uncertain to include in the present study.
Given the large amount of sea water on the planet, this resources is thought very large even though the concentration
of U in sea water is very low. However, due to the very slow mixing process of sea water, it is misleading to consider
the global body of sea water as a usable source of U, as it could be rapidly depleted locally, providing small amounts,
without access to the remaining resources situated far offshore [87].
15.11. Thorium
Thorium (Th) deposits in the Earths crust around the world are expected to be three times larger than those of
U, as determined from isotope lifetimes and the composition of the accreted material which formed the planet [87
90]34 . However, known resources of Th are much smaller and less detailed than those of U, a situation which is the
result of the relatively small interest that has been given to Th in comparison to U. Therefore, it is to be expected that
Th resources increase in size significantly if at some time in the future interest grows in Th based nuclear reactors.
Although the Th nuclear fuel cycle has been demonstrated several decades ago, it has not been used commercially
as it involves more safety hazards related to radiation than the U fuel cycle [87, 88]. The Th nuclear fuel cycle is
more efficient than that of U and therefore, involves less mass of Th per unit of electricity produced. For similar
mining costs, Th resources are less expensive per unit of energy, however, the processing of Th into 233 U for fuel
preparation has not been performed at an industrial scale. Therefore, the cost variable in the Th cost-supply curve is
highly uncertain.
Data for Th resources were obtained from the IAEA [26]. These are provided with much less detail than for U,
with four uncertainty ranges but only one cost category. Consequently, the strategy of curve-fitting cannot be applied
here, and one distribution of the hierarchical type per uncertainty category per world region was parametrised in the
same way as for fossil resources. The cost axis was transformed from a cost per unit of mass to a cost per unit of
energy using the efficiency of the Indian experimental model reported by Sinha and Kakodkar of 2100 TJ/t [88]. The
curve for the lower boundary of the uncertainty range includes RAR only. The most probable curve includes RAR
and inferred reserves. The curve for the upper boundary of the uncertainty range includes all four categories.
34 Such arguments, which originates from astrophysics and the study of stars and nebulae, along with nuclear lifetimes of isotopes which were
formed through nuclear reactions within stars or during supernova events, enable the prediction of relative amounts on Earth for all isotopes of
elements of the periodic table (see for instance [90]).
16. Data tables and figures

IMAGE Canada Wind Potential
IMAGE North Africa Solar Potential

200
60
50
40
30
A = 70.96 EJ/y
B = 9.3 $/MWh
C0 = 58.2 $/MWh
20
A = 63.32 EJ/y
B = 22.1 $/MWh
C0 = 57.0 $/MWh
Energy potential ( EJ/y )
180
160
140
120
100
A = 219.6 EJ/y
B = 57 $/MWh
C0 = 457 $/MWh
80
A = 181.5 EJ/y
B = 87 $/MWh
C0 = 466 $/MWh
60
40
10
Dist. Hierarchical Resources

Dist. Nearly identical Resources
40
60
80
100 120 140

Cost ( USD/MWh )
160
180

20
0
300
200
400
500
600
700
800
Cost ( USD/MWh )
900
1000
IMAGE A1 2050 Rest Land
IMAGE A1 2050 Abandoned agricultural Land
70
350
60
250
200
A = 442.4 EJ/y
B = 0.718 $/GJ
C = 1.07 $/GJ
150
100
50
A = 345.1 EJ/y
B = 0.995 $/GJ
C0 = 1.19 $/GJ

300
50
40
30
A = 64.45 EJ/y
B = 1.46 $/GJ
C = 0.670 $/GJ
20
10
A = 82.86 EJ/y
B = 0.978 $/GJ
C = 0.541 $/GJ
0

0
0
2
3
Cost ( USD/GJ )
Cost ( USD/GJ )
Figure 16.1: Curve fits using non-linear least-squares of the two types of cumulative distribution with data from various studies of renewable energy
potentials previously reported, calculated using the model IMAGE (reproduced from [18, 19, 22]). The goodness of these fits are a good indication
for which type of distribution represents best each type of resource. It can observed that data for abandoned agricultural land is well described by
the cumulative distribution for nearly identical resources (top le f t), while the data for rest land is described by the cumulative distribution of the
hierarchical type (top right). Data for wind energy is well described by hierarchical resource distribution (bottom le f t), while the data for solar
energy is well described by a distribution for nearly identical resources (bottom right).
1- USA
2- Canada
3- EU-15
4- Rest-EU
5- Russia
6- China
7- Japan
8- India
9- Rest Asia
10- Oceania
Region
USA
Canada
EU-15
Rest Europe
Russia
China
Japan
India
Rest Asia
Oceania
Brazil
Rest America
Africa
Middle East
11121314-
Brazil
Rest America
Africa
Middle East
Member countries
USA
Canada
Austria, Belgium, Danemark, Finland, France, Germany, Greece, Ireland, Italy, Luxemburg, Netherlands, Portugal, Spain,
Sweden, United Kingdom
Albania, Belarus, Bosnia-Herzegovina, Bulgaria, Croatia, Cyprus, Czech Rep., Estonia, Hungary, Iceland, Latvia, Lithuania,
Macedonia, Malta, Moldova, Montenegro, Norway, Poland, Romania, Serbia, Slovakia, Slovenia, Switzerland, Turkey, Ukraine
Russia
China
Japan
India
Afghanistan, Armenia, Azerbaijan, Bangladesh, Bhutan, Brunei, Cambodia, Georgia, Hong Kong, Indonesia, Kazakhtan, Korea,
Kyrghizstan, Laos, Malaysia, Mongolia, Myanmar, Nepal, Pakistan, Philippines, Singapore, Taiwan, Tajikistan, Thailand,
Turkmenistan, Uzbeksitan, Viet Nam
Australia, New Zealand, Papua New Guinea, pacific islands
Brazil
Mexico, Central America, South America excluding Brazil
Africa
Barhain, Iran, Iraq, Israel, Jordan, Kuwait, Lebanon, Oman, Palestine, Qatar, Saudi Arabia, Syria, UAE, Yemen
Table 16.1: Definition of world regions for this paper with member countries for each.
Renewable cost-supply curve parameters

Hydro
Hierarchical
Region
A
B
PJ/y
$/MWh
Name
USA
5 746
198.40
4 217
38.82
Canada
EU-15
3 066
22.01
4 283
54.13
R. Eur.
6 595
82.84
Russia
11 502
26.37
China
Japan
846
25.24
2 788
103.23
India
7 227
160.05
R. Asia
779
71.64
Oceania
4 738
18.45
Brazil
R. Amer.
7 414
127.92
5 767
69.16
Africa
1 094
304.10
Mid. East
Total
66 061
C0
$/MWh
1.40
78.06
88.25
38.85
75.89
35.65
82.31
0.00
0.00
0.00
12.00
0.00
64.42
0.00
Wind
Hierarchical
A
B
PJ/y
$/MWh
75 600
30.19
43 505
10.17
12 009
65.77
32 183
19.38
38 774
65.10
6 057
175.01
360
109.24
2 018
109.62
18 125
83.64
50 410
59.23
13 248
23.33
22 752
30.14
23 106
143.77
7 200
109.62
345 348
C0
$/MWh
145.53
149.28
113.60
142.14
129.73
161.04
161.04
208.48
123.77
138.60
127.38
132.96
161.04
208.48
Solar
Nearly identical
A
B
PJ/y
$/MWh
262 800
350.03
51 571
1542.58
61 471
817.28
57 355
593.61
384 448
428.97
175 509
241.71
3 600
185.69
120 525
140.01
282 441
369.67
430 421
141.75
113 386
310.45
160 214
268.27
974 259
157.16
306 000
172.80
3 384 000
C0
$/MWh
620.44
839.26
616.21
718.00
611.15
695.79
840.10
483.33
498.17
521.29
527.52
527.44
456.31
437.87
Table 16.2: Table of cost-supply curve parameters for each region for hydro, wind and solar power.
Renewable cost-supply curve parameters
Wave
Hierarchical
A
Region
PJ/y
Name
USA
496
5030
Canada
1525
EU-15
R. Eur.
2442
937
Russia
361
China
741
Japan
174
India
R. Asia
2843
1536
Oceania
186
Brazil
1303
R. Amer.
Africa
1027
309
Mid. East
Total
18 910
B
$/MWh
32.46
32.46
32.46
32.46
32.46
32.46
32.46
32.46
32.46
32.46
32.46
32.46
32.46
32.46
C0
$/MWh
199.44
199.44
199.44
199.44
199.44
199.44
199.44
199.44
199.44
199.44
199.44
199.44
199.44
199.44
Tidal
Hierarchical
A
PJ/y
145
757
287
333
743
49
101
89
396
238
25
253
140
42
3600
B
$/MWh
89.18
89.18
89.18
89.18
89.18
89.18
89.18
89.18
89.18
89.18
89.18
89.18
89.18
89.18
C0
$/MWh
303.33
303.33
303.33
303.33
303.33
303.33
303.33
303.33
303.33
303.33
303.33
303.33
303.33
303.33
Table 16.3: Table of cost-supply curve parameters for each region for wave and tidal energy.

Primary biomass cost-supply curve parameters
A1
Region
A
B
PJ/y
$/MWh
Name
USA
53 082
3.93
18 000
1.64
Canada
11 305
1.62
EU-15
11 855
0.43
Rest Europe
Russia
126 017
1.73
107 100
11.26
China
115
0.69
Japan
25 890
3.37
India
Rest Asia
12 921
0.91
55 625
2.01
Oceania
77 751
1.33
Brazil
27 629
2.06
Rest America
Africa
139 246
4.32
13 011
25.38
Middle East
Total
679 548
B1
Region
A
B
PJ/y
$/MWh
Name
USA
36 082
0.85
14 000
0.84
Canada
7 268
3.29
EU-15
9 844
2.42
Rest Europe
Russia
87 319
0.99
77 179
1.57
China
115
0.69
Japan
13 756
3.35
India
5 075
0.72
Rest Asia
Oceania
35 279
0.96
56 539
4.76
Brazil
18 841
3.86
Rest America
81 240
3.36
Africa
4 011
4.99
Middle East
Total
446 548
A2
C0
$/MWh
3.60
3.60
6.15
6.28
3.60
5.04
7.20
3.00
3.48
3.14
6.54
5.45
1.68
6.96
Type
C0
$/MWh
5.00
3.60
5.93
3.59
3.60
3.60
7.20
2.92
3.49
2.81
4.44
3.00
2.50
7.20
Type
2
2
2
2
2
1
2
2
2
2
2
2
2
1
2
2
1
1
2
2
2
2
2
2
1
2
2
2
A
PJ/y
33 082
12 000
11 305
10 781
67474
23 322
44
13 756
8 605
34 477
22 070
7 310
53 240
8 011
305 477
B2
A
PJ/y
49 082
13 000
12 921
12 178
77 396
46 261
225
6 289
5 349
30 329
38 863
10 517
15 237
3 011
320 658
B
$/MWh
6.42
1.50
0.93
1.45
1.96
9.92
9.98
2.32
0.35
3.31
5.12
3.58
6.78
13.08
C0
$/MWh
3.10
3.60
6.18
3.44
3.16
7.20
14.40
2.86
3.52
2.52
3.81
3.44
-0.86
7.20
Type
B
$/MWh
2.07
0.94
4.77
0.45
0.94
5.51
0.69
5.24
1.80
0.87
2.29
12.20
4.47
5.49
C0
$/MWh
3.60
3.60
4.48
5.70
3.60
7.20
7.20
3.60
3.85
3.09
6.03
3.60
1.17
7.20
Type
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
1
2
2
2
2
2
2
2
2
1
2
2
Table 16.4: Table of cost-supply curve parameters for biomass primary energy resources for four SRES scenarios A1, A2, B1 and B2.
Geothermal energy cost-supply curve parameters
Direct Use
In belt
Out of belt
Hierarchical
Nearly identical
A
B
C0
A
B
Region
PJ/y
$/MWh
$/MWh
PJ/y
$/MWh
Name
USA
74
7.22
94.34
138
81.44
9
7.22
94.34
83
81.44
Canada
1
7.22
94.34
89
81.44
EU-15
12
7.22
94.34
27
81.44
R. Eur.
Russia
9
7.22
94.34
177
81.44
44
7.22
94.34
103
81.44
China
15
7.22
94.34
0
81.44
Japan
2
7.22
94.34
37
81.44
India
115
7.22
94.34
90
81.44
R. Asia
Oceania
20
7.22
94.34
80
81.44
5
7.22
94.34
96
81.44
Brazil
143
7.22
94.34
91
81.44
R. Amer.
54
7.22
94.34
210
81.44
Africa
Mid East.
11
7.22
94.34
49
81.44
Total
514
1 269
C0
$/MWh
118.44
118.44
118.44
118.44
118.44
118.44
118.44
118.44
118.44
118.44
118.44
118.44
118.44
118.44
Electricity
In belt
Hierarchical
A
B
PJ/y
$/MWh
1 361
20.62
253
20.62
14
20.62
123
20.62
275
20.62
809
20.62
144
20.62
39
20.62
1 498
20.62
222
20.62
102
20.62
2 030
20.62
761
20.62
188
20.62
7 818
C0
$/MWh
144.96
144.96
144.96
144.96
144.96
144.96
144.96
144.96
144.96
144.96
144.96
144.96
144.96
144.96
Out of belt
Nearly identical
A
B
PJ/y
$/MWh
2 528
63.62
2 280
63.62
1 458
63.62
524
63.62
5 226
63.62
1 887
63.62
0
63.62
743
63.62
1 482
63.62
1 647
63.62
1 942
63.62
1 552
63.62
4 179
63.62
901
63.62
26 349
C0
$/MWh
255.71
255.71
255.71
255.71
255.71
255.71
255.71
255.71
255.71
255.71
255.71
255.71
255.71
255.71
Table 16.5: Table of cost-supply parameters for geothermal energy, for both direct use of heat and electricity production.

Oil
Region
Name
USA
Canada
EU-15
Rest Europe
Russia
China
Japan
India
Rest Asia
Oceania
Brazil
Rest America
Africa
Middle East
Total
Costs
Upper
Lower
Mtoe
Crude Oil
[13]
Reserves
3 863
667
1 193
1 155
10 436
2 018
6
792
9 218
595
2 450
8 996
17 277
102 366
161 031
USD2008 / boe
Crude Oil
Reserves
10
40
Resources
10 000
2 400
1 545
3 530
16 400
2 300
10
400
10 535
1 100
5 000
9 588
15 485
21 170
99 463
Oil Shales
[15]
Resources
536 931
2 192
13 248
4 411
35 470
47 600
0
0
3 988
4 534
11 734
60
23 317
5 792
689 277
Oil Sands
[15]
Reserves
0
24 909
31
0
4 147
0
0
0
6 203
0
0
0
263
0
35 552
Resources
5 429
227 189
276
1
39 034
233
0
0
55 945
0
0
136
2 364
0
330 607
Oil Shales
Resources
50
100
Oil Sands
Reserves
40
50
Resources
40
70
Additional
2 388
355 828
0
0
7 505
0
0
0
0
0
0
0
6 778
0
372 500
Extra Heavy Oil

[15]
Reserves
Resources
3
379
0
0
24
1 928
5
51
1
25
110
1 168
0
0
0
0
18
1 163
0
0
0
0
8 476
270 637
7
66
0
0
8 644
275 416
Additional
4
0
0
0
0
0
0
0
0
0
0
27 704
0
0
27 707
Additional
40
70
Extra Heavy Oil

Reserves
Resources
10
40
50
70
Additional
10
70
[82]
Resources
10
100
Table 16.6: Oil resources by world region in units of Mtoe (million tonnes of oil).
Gas
Region
Name
USA
Canada
EU-15
R. Eur.
Russia
China
Japan
India
Rest Asia
Oceania
Brazil
R. Amer.
Africa
Mid. East
International
Total
Costs
Upper
Lower
109 m3
Conv. gas
[13]
Reserves
7 080
1 754
2 338
3 889
47 578
2 455
21
1 115
23 951
3 553
365
7 704
14 753
75 358
0
191 914
Resources
20 000
7 000
2 530
7 510
105 000
10 000
5
900
22 805
2 450
2 000
8 858
16 155
35 370
0
240 583
Shale gas
[83]
Reserves
17 000
10 988
7 024
9 374
538
36 109
0
1 784
1 444
11 215
7 533
47 579
29 482
0
0
180 070
USD2008 / GJ
Conv. gas
Reserves
Resources
0.5
0.5
5.7
5.7
[86]
Shale gas
Reserves
3.8
8.6
Resources
45 600
42 198
28 547
39 734
2 152
144 463
0
8 213
5 834
39 111
25 658
170 745
112 206
0
0
1 0748 62
Tight gas
[13]
Reserves
1000
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1 000
Resources
3.8
8.6
Tight gas
Reserves
2.6
7.6
Resources
210 000
7 000
7 000
0
45 000
9 000
0
1 000
0
1 000
6 000
0
0
0
38 000
324 000
Coalbed Methane
[84]
Reserves
Resources
9 700
2 000
5 700
70 800
2 802
0
1 908
0
17 000
96 300
30 000
5 100
0
0
800
0
1 100
0
8 500
5 700
0
0
0
0
800
0
0
0
0
0
78 310
179 900
Methane Hydrates
[85]
Reserves
Resources
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
300 000
300 000
300 000
300 000
Resources
2.6
7.6
Coalbed Methane
Reserves
Resources
3.8
3.8
7.6
7.6
Methane Hydrates
Reserves
Resources
4.4
4.4
8.6
8.6
Table 16.7: Natural gas resources by world region in units of Gm3 (billion cube meters).
Coal
Region
Name
USA
Canada
EU-15
Rest Europe
Russia
China
Japan
India
Rest Asia
Oceania
Brazil
Rest America
Africa
Middle East
Total
Region
Name
USA
Canada
EU-15
Rest Europe
Russia
China
Japan
India
Rest Asia
Oceania
Brazil
Rest America
Africa
Middle East
Total
Costs
Mt
Hard coal
Reserves
Proven
[15]
226 694
4 346
84 721
24 534
68 655
180 600
340
56 100
54 678
44 627
1 547
9 960
32 546
1 203
790 551
Soft coal
Reserves
Proven
[15]
30 851
3 108
44 214
40 456
91 350
52 300
10
4 895
30 762
37 738
4 559
5 633
180
0
346 056
USD2008 / t
Reserves
Proven
20
50
Resources
Proven
[15]
6 691 942
187 606
278 420
254 658
2 730 810
681 600
4 603
105 820
289 048
1 620 675
6 212
20 496
58 150
41 203
12 971 243
Probable
[13]
0
0
0
7 428
0
0
1 988
123 470
0
0
0
0
0
0
132 886
Possible
Probable
[13]
0
40 055
0
14 961
0
0
1 132
0
11 871
73 102
10 799
0
0
0
151 920
Possible
0
0
0
3 124
0
0
0
0
34 070
101 100
4 575
790
0
0
143 659
Resources
Proven
[15]
1 398 669
17 371
89 158
275 185
1 371 030
318 000
160
38 647
387 263
46 973
6 513
7 524
338
0
3 956 831
Possible
20
50
Resources
Proven
20
100
Probable
20
100
Possible
20
100
Probable
[13]
0
0
0
752
0
0
0
0
0
0
0
4572
0
0
5 324
Possible
Probable
[13]
0
0
0
1 996
0
0
0
0
7 086
62 840
7 559
527
0
0
80 008
Possible
0
0
0
1 862
0
0
0
0
0
0
0
4 237
0
0
6 099
0
0
0
11 422
0
0
7 375
37 920
0
0
0
0
0
0
56 717
0
108 995
0
11 581
0
0
4 074
0
57 198
112 300
6 535
0
0
0
300 683
[82]
Probable
20
50
Table 16.8: Coal resources by world region in units of Mt (million tonnes of coal). Hard coal includes anthracite and bituminous coal, while soft
coal includes sub-bituminous coal and lignite. Since there is no clear demarcation between ranks of coal, the limit is put onto the calorific content,
and thus coal resources with a calorific content higher than 16 500 kj/t belong to the hard coal category (as defined in [13]), while coal resources
with a lower calorific content belong to soft coal. Anthracite can have calorific contents of up to 35 000 kJ/t while lignite can have calorific values
as low as 11 000 kJ/t.

Uranium
Region
Name
USA
Canada
EU-15
Rest Europe
Russia
China
Japan
India
Rest Asia
Oceania
Brazil
Rest America
Africa
Middle East
Total
Region
Name
USA
Canada
EU-15
Rest Europe
Russia
China
Japan
India
Rest Asia
Oceania
Brazil
Rest America
Africa
Middle East
Total
t
[26]
Reasonably Assured Reserves (RAR)
<40$/kg
<80$/kg
<130$/kg
0
39 000
207 400
267 100
336 800
361 100
0
7 000
20 800
2 500
39 100
88 500
0
100 400
181 400
52 000
100 900
115 900
0
0
6 600
0
0
55 200
14 600
326 600
454 500
0
1 163 000
1 176 000
139 900
157 700
157 700
0
7 000
11 700
93 800
194 600
644 100
0
44 000
44 000
569 900
2 516 100
3 524 900
Prognosticated
<80$/kg
<130$/kg
819 500
1 169 300
50 000
150 000
7 000
7 600
200
41 650
0
182 000
3600
3 600
0
3 300
0
0
377 900
591 400
300 000
300 000
73 600
121 000
6 600
23 500
49 400
156 900
67 800
89 000
1 755 600
2 839 250
<260$/kg
472 100
387 400
33 800
160 000
181 400
115 900
6 600
55 200
533 500
1 179 000
157 700
13 800
663 400
44 700
4 004 500
<260$/kg
1 036 950
150 000
7 600
66 650
182 000
3 600
3 300
63 600
592 900
300 000
121 000
23 500
156 900
89 000
2 797 000
Inferred
<40$/kg
0
99 700
0
3 200
0
15 400
0
0
29 800
0
0
0
78 500
0
226 600
Speculative
<80$/kg
19 500
110 600
0
15 000
57 700
49 100
0
0
276 800
449 000
73 600
4 400
121 800
67 800
1 245 300
<130$/kg
103 700
124 200
13 500
43 850
298 900
55 500
3 300
23 900
366 000
497 000
121 000
10 100
260 000
67 800
1 989 750
<260$/kg
236 050
157 200
110 400
109 850
384 900
55 500
3 300
24 900
474 900
500 000
121 000
11 300
286 300
69 200
2 544 800
<130$/kg
858 000
700 000
50 100
6 650
0
4 100
3 300
0
1 776 600
0
121 000
236 700
25 000
84 800
3 866 250
<260$/kg
858 000
700 000
50 100
126 650
0
4 100
3 300
0
1 806 100
0
121 000
236 700
25 500
98 800
4 029 750
Unassigned
482 000
0
94 000
314 000
633 000
0
0
17 000
264 700
500 000
0
176 200
1 112 900
0
3 593 800
Table 16.9: Uranium resources (in natural concentration) by world region in units of tonnes.
Thorium
Region
Name
USA
Canada
EU-15
Rest Europe
Russia
China
Japan
India
Rest Asia
Oceania
Brazil
Rest America
Africa
Middle East
Unassigned
Total
Costs
Lower
Upper
[26]
RAR
< 80 USD/kg
122 000
0
0
54 000
75 000
0
0
319 000
0
46 000
172 000
0
18 000
0
23 000
829 000
Inferred
< 80 USD/kg
278 000
44 000
0
213 000
112 500
0
0
478 500
0
406 000
130 000
300 000
127 000
0
10 000
2 099 000
Indentified
< 80 USD/kg
400 000
44 000
0
186 000
75 000
0
0
319 000
0
452 000
302 000
300 000
118 000
0
33 000
2 229 000
Prognosticated
N/A
274 000
128 000
0
164 000
0
0
0
0
0
0
330 000
0
410 000
0
81 000
1 387 000
Inferred
40
80
Indentified
40
80
Prognosticated
80
260
USD2008 / kg
RAR
40
80
Table 16.10: Thorium resources (in natural concentration) by world region in units of tonnes.
Acknowledgements
The authors would like to acknowledge particularly Dr T. S. Barker and T. Hanaoka for guidance and support,
as well as A. Anger, H. Pollitt, P. Summerton and P. Bruseghini for highly informative discussions. This work
was supported by the Three Guineas Trust, Conicyt (Comision Nacional de Investigacion Cientfica y Tecnologica,
Gobierno de Chile) and the Ministerio de Energa, Gobierno de Chile.
References
[1] O. Edenhofer, B. Knopf, T. Barker, L. Baumstark, E. Bellevrat, B. Chateau, P. Criqui, M. Isaac, A. Kitous, C. Kypreos, M. Leimbach,
K. Lessmann, B. Magne, S. Scrieciu, H. Turton, D. P. van Vuuren, The Economics of Low Stabilization: Model Comparison of Mitigation
Strategies and Costs, Energy Journal 31 (1) (2010) 1148.
[2] A. Grubler, N. Nakicenovic, D. Victor, Dynamics of energy technologies and global change, Energy Policy 27 (5) (1999) 247280.
[3] C. Marchetti, N. Nakicenovic, The dynamics of energy systems and the logistic substitution model, Tech. rep., IIASA (1978).
URL http://www.iiasa.ac.at/Research/TNT/WEB/PUB/RR/rr-79-13.pdf
[4] J.-F. Mercure, FTT:Power : A global model of the power sector with induced technological change and natural resource depletion, Energy
Policy 48 (0) (2012) 799 811.
URL http://dx.doi.org/10.1016/j.enpol.2012.06.025
[5] Cambridge Econometrics, Energy-Economy-Environment Model at the Global level, Cambridge Econometrics, Cambridge, UK, 2013.
URL www.e3mgmodel.com
[6] T. Barker, H. Pan, J. Koehler, R. Warren, S. Winne, Decarbonizing the global economy with induced technological change: Scenarios to
2100 using E3MG, Energy Journal (Sp. Iss. 1) (2006) 241258.
[7] J. Koehler, M. Grubb, D. Popp, O. Edenhofer, The transition to endogenous technical change in climate-economy models: A technical
overview to the innovation modeling comparison project, Energy Journal (Sp. Iss. 1) (2006) 1755.
[8] T. Barker, S. S. Scrieciu, Modeling Low Climate Stabilization with E3MG: Towards a New Economics Approach to Simulating EnergyEnvironment-Economy System Dynamics, Energy Journal 31 (Sp. Iss. 1) (2010) 137164.
[9] A. S. Dagoumas, T. S. Barker, Pathways to a low-carbon economy for the UK with the macro-econometric E3MG model, Energy Policy
38 (6) (2010) 30673077.
[10] J.-F. Mercure, Global electricity technology substitution model with induced technological change, Tyndall Working Paper (148).
[11] UNDP, World Energy Assessment, UNDP, 2000.
[12] IPCC, Energy supply, in: Climate Change 2007: Mitigation of Climate Change. Contribution of Working Group III to the Fourth Assessment
Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, 2007.
[13] BGR, Reserves, Resources and Availability of Energy Resources, BGR, 2010.
[14] IPCC, Special Report on Renewable Energy Sources and Climate Change Mitigation, Cambridge University Press, 2011.
[15] WEC, 2010 Survey of Energy Resources, WEC, 2010.
[16] W. Krewitt, K. Nienhaus, C. Klessmann, C. Capone, E. Stricker, W. Graus, M. Hoogwijk, N. Supersberger, U. von Winterfeld, S. Samadi,
Role and potential of renewable energy and energy efficiency for global energy supply, Tech. rep., German Federal Environment Agency
(2009).
[17] WWF, The energy report, 100% renewable energy by 2050, Tech. rep., World Wildlife Fund (2011).
[18] M. Hoogwijk, On the global and regional potential of renewable energy sources, Ph.D. thesis, Universiteit Utrecht (2004).
[19] M. Hoogwijk, B. de Vries, W. Turkenburg, Assessment of the global and regional geographical, technical and economic potential of onshore
wind energy, Energy Economics 26 (5) (2004) 889919.
[20] M. Hoogwijk, A. Faaij, B. Eickhout, B. de Vries, W. Turkenburg, Potential of biomass energy out to 2100, for four ipccsres land-use
scenarios, Biomass & Bioenergy 29 (4) (2005) 225257.
[21] B. J. de Vries, D. P. van Vuuren, M. M. Hoogwijk, Renewable energy sources: Their global potential for the first-half of the 21st century at
a global level: An integrated approach, Energy Policy 35 (2007) 2590 2610.
[22] M. Hoogwijk, A. Faaij, B. de Vries, W. Turkenburg, Exploration of regional and global cost-supply curves of biomass energy from shortrotation crops at abandoned cropland and rest land under four ipcc sres land-use scenarios, Biomass & Bioenergy 33 (1) (2009) 2643.
[23] H. Rogner, An assessment of world hydrocarbon resources, Annual review of Energy and the Environment 22 (1997) 217262.
[24] J. Wolf, P. S. Bindraban, J. C. Luijten, L. M. Vleeshouwers, Exploratory study on the land area required for global food supply and the
potential global production of bioenergy, Agricultural Systems 76 (3) (2003) 841861.
[25] E. M. W. Smeets, A. P. C. Faaij, I. M. Lewandowski, W. C. Turkenburg, A bottom-up assessment and review of global bio-energy potentials
to 2050, Progress in Energy and Combustion Science 33 (1) (2007) 56106.
[26] IAEA, Uranium 2009: Resources, Production and Demand, IAEA/OECD/NEA, 2009.
[27] B. Srensen, Renewable energy : physics, engineering, environmental impacts, economics & planning, Academic Press, 2011.
[28] M. J. Grubb, N. I. Meyer, Wind energy: Resources, systems and regional strategies, in: T. B. Johansson, H. Kelly, A. K. Reddy, R. H.
Williams (Eds.), Renewable Energy: Sources for Fuels and Electricity., Island Press, 1993, pp. 157212.
[29] K. A. Abed, A. A. El-Mallah, Capacity factor of wind turbines, Energy 22 (5) (1997) 487491.
[30] D. J. Mackay, Sustainable energy without the hot air, UIT Cambridge, http://www.withouthotair.com, 2008.
[31] X. Lu, M. B. McElroy, J. Kiviluoma, Global potential for wind-generated electricity, Proceedings of the National Academy of Sciences of
the United States of America 106 (27) (2009) 1093310938.
[32] A. Fellows, G. Gow, The potential of wind energy to reduce carbon dioxide emissions, Tech. rep., GL Garrad Hassan (2000).
[33] C. L. Archer, M. Z. Jacobson, Evaluation of global wind power, Journal of Geophysical Research-Atmospheres 110 (D12).

[34] WEC, New Renewable Energy Resources: A Guide to the Future, WEC, 1994.
[35] A. F. Bouwman, T. Kram, K. K., Integrated modelling of global environmental change. An overview of IMAGE 2.4, Tech. rep., Netherlands
Environmental Assessment Agency (2006).
URL http://www.rivm.nl/bibliotheek/rapporten/500110002.pdf
[36] Troen, E. L. Petersen, European Wind Atlas, Ris National Laboratory, 1989.
[37] J. Weinzettel, M. Reenaas, C. Solli, E. G. Hertwich, Life cycle assessment of a floating offshore wind turbine, RENEWABLE ENERGY
34 (3) (2009) 742747.
[38] M. Hoogwijk, W. Graus, Global potential of renewable energy sources: A literature assessment, Tech. rep., Ecofys (2008).
[39] EWEA, The economics of wind energy, Tech. rep., European Wind Energy Association (2009).
[40] G. W. Crabtree, N. S. Lewis, Solar Energy Conversion, in: Hafemeister, D. and Levi, B. G. and Levine, M. D. and Schwartz, P. (Ed.),
Physics of Sustainable Energy: Using Energy Efficiently and Producing it Renewably, Vol. 1044 of AIP Conference Proceedings, 2008, pp.
309321.
[41] Y. Hofman, D. de Jager, E. Molenbroek, F. Schillig, M. Voogt, The potential of solar electricity to reduce co2 emissions, Tech. rep., Ecofys
(2002).
[42] IEA, World Energy Outlook 2010, IEA/OECD, 2010.
[43] V. Avrutin, N. Izyumskaya, H. Morkoc, Semiconductor solar cells: Recent progress in terrestrial applications, Superlattices and Microstructures 49 (4) (2011) 337364.
[44] ETSAP, Concentrating solar power, Tech. rep., IEA (2010).
[45] M. A. Green, K. Emery, Y. Hishikawa, W. Warta, E. D. Dunlop, Solar cell efficiency tables (Version 38), Progress in Photovoltaics 19 (5)
(2011) 565572.
[46] IPCC, Hydropower, in: Special Report on Renewable Energy Sources and Climate Change Mitigation, Cambridge University Press, 2011.
[47] IJHD, 2011 world atlas and industry guide, Tech. rep., The International Journal on Hydropower and Dams, Wallington, Surrey, UK (2011).
[48] P. Lako, H. Eder, M. de Noord, H. Reisinger, Hydropower development with a focus on asia and western europe, overview in the framework
of vleem 2, Tech. rep., ECN Policy Studies and Verbundplan (2003).
[49] IEA, Projected Costs of Generating Electricity 2010, IEA/OECD, 2010.
[50] G. J. F. Macdonald, Calculations on the thermal history of the earth, Journal of Geophysical Research 64 (11) (1959) 19672000.
[51] G. J. Wasserburg, W. A. Fowler, G. J. F. Macdonald, F. Hoyle, Relative contributions of uranium thorium + potassium to heat production in
earth, Science 143 (360) (1964) 465&.
[52] R. Cataldi, Review of Historiographic Aspects of Geothermal Energy in the Mediterranean and Mesoamerican Areas Prior to the Modern
Age, Geo-Heat Centre Quarterly Bulletin 18 (1) (1993) 1316.
[53] J. W. Lund, 100 Years of Geothermal Power Production, in: Thirtieth Workshop on Geothermal Reservoir Engineering, 2005.
[54] J. Mock, J. Tester, P. Wright, Geothermal energy from the earth: Its potential impact as an environmentally sustainable resource, Annual
review of Energy and the Environment 22 (1997) 305356.
[55] J. W. Tester, B. J. Anderson, Impact of enhanced geothermal systems (egs) on the united states in the 21st century, in: The Future of
Geothermal Energy, Massachusetts Institute of Technology, 2006.
[56] M. J. Aldrich, A. W. Laughlin, D. T. Gambill, Geothermal resource base of the world: A revision of the electrical power research institutes
estimate, Tech. rep., Los Alamos Scientific Laboratory, University of California (1981).
[57] EPRI, Geothermal Energy Prospects for the Next 50 Years, EPRI, 1978.
[58] H. N. Pollack, S. J. Hurter, J. R. Johnson, Heat-flow from the earths interior - analysis of the global data set, Reviews of Geophysics 31 (3)
(1993) 267280.
[59] M. J. Pasqualetti, The site specific nature of geothermal-energy - the primary role of land-use planning in non-electric development, Natural
Resources Journal 23 (4) (1983) 795814.
[60] V. Stefansson, World Geothermal Assessment, in: World Geothermal Congress, Antalya, Turkey, 2005.
[61] B. Goldstein, A. L. T. Hill, M. Malavazos, A. Budd, B. Ayling, Hot Rocks Downunder - Evolution of a New Energy Industry, in: GRC
Annual Meeting, 2009.
[62] R. Bertani, Geothermal power generation in the world. 2005-2010 update report, in: World Geothermal Congress 2010, Bali, Indonesia,
2010.
[63] R. Bertani, Geothermal power generation in the world 2005 - 2010 update report, Geothermics 41 (0) (2012) 1 29.
[64] IEA, Renewable energy essentials: Geothermal, Tech. rep., IEA (2010).
[65] J. S. Rhodes, D. W. Keith, Engineering economic analysis of biomass IGCC with carbon capture and storage, Biomass & Bioenergy 29 (6)
(2005) 440450.
[66] C. Gough, P. Upham, Biomass energy with carbon capture and storage (BECCS or Bio-CCS), GHG Sci. and Tech. 1 (4) (2011) 324334.
[67] IPCC, Bioenergy, in: Special Report on Renewable Energy Sources and Climate Change Mitigation, Cambridge University Press, 2011.
[68] M. Hoogwijk, A. Faaija, R. van den Broek, G. Berndes, D. Gielen, W. Turkenburg, Exploration of the ranges of the global potential of
biomass for energy, Biomass & Bioenergy 25 (2) (2003) 119133.
[69] D. van Vuuren, J. van Vliet, E. Stehfest, Future bio-energy potential under various natural constraints, Energy Policy 37 (11) (2009) 4220
4230.
[70] IEA Bioenergy, Bioenergy - A sustainable and reliable energy source, Main report, IEA/OECD, 2009.
[71] V. Dornburg, F. van Vuuren, G. van de Ven, H. Langeveld, M. Meeusen, M. Banse, M. van Oorschot, J. Ros, G. J. van den Born, H. Aiking,
M. Londo, H. Mozaffarian, P. Verweij, E. Lysen, A. Faaij, Bioenergy revisited: Key factors in global potentials of bioenergy, ENERGY &
ENVIRONMENTAL SCIENCE 3 (3) (2010) 258267.
[72] IPCC, Emission Scenarios, Cambridge University Press, 2000.
[73] ETSAP, Marine energy, Tech. rep., IEA (2010).
[74] IPCC, Ocean energy, in: Special Report on Renewable Energy Sources and Climate Change Mitigation, Cambridge University Press, 2011.
[75] R. E. H. Sims, R. N. Schock, A. Adegbululgbe, J. Fenhann, I. Konstantinaviciute, W. Moomaw, H. B. Nimir, B. Schlamadinger, J. Torres-
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84]
[85]
[86]
[87]
[88]
[89]
[90]
[91]
[92]
[93]
[94]
[95]
[96]
[97]
[98]
[99]
[100]
[101]
[102]
[103]
[104]
[105]
[106]
[107]
[108]
[109]
[110]
[111]
[112]
[113]
[114]
Martinez, C. Turner, Y. Uchiyama, S. J. V. Vuori, N. Wamukonya, X. Zhang, Energy supply, in: B. Metz, O. Davidson, P. Bosch, T. Dave,
L. Meyer (Eds.), Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, 2007.
T. J. Hammons, Tidal Power, Proceedings of the IEEE 81 (3) (1993) 419433.
R. H. Charlier, J. R. Justus, Ocean Energies: Environmental, Economic and Technological Aspects of Alternative Power Sources, Elsevier
Oceanography Series, 1993.
G. C. Nihous, A preliminary assessment of ocean thermal energy conversion resources, Journal of Energy Resources TechnologyTransactions of the ASME 129 (1) (2007) 1017.
O. S. Skramesto, S. E. Skilhagen, W. K. Nielsen, Power Production Based on osmotic Pressure, in: Waterpower XVI, Spokane, WA, USA,
2009.
J. E. Cavanagh, J. H. Clarke, T. Price, Ocean Energy Systems, in: T. B. Johansson, H. Kelly, A. K. N. Reddy, R. H. Williams (Eds.),
Renewable Energy. Sources for Fuels and Electricity, Island Press, 1993.
V. E. Mckelvey, Mineral Resource Estimates and Public Policy, American Scientist 60 (1) (1972) 3240.
IEA, World Energy Outlook 2008, IEA/OECD, 2008.
EIA, World Shale Gas Resources: An Initial Assessment of 14 Regions Outside the United States, EIA, 2011.
C. M. Boyer, Q. Z. Bai, Methodology of coalbed methane resource assessment, International Journal of Coal Geology 35 (1-4) (1998)
349368.
R. Boswell, T. S. Collett, Current perspectives on gas hydrate resources, Energy & Environmental Science 4 (4) (2011) 12061215.
ETSAP, Unconventional oil & gas production, Tech. rep., IEA (2010).
P. Bonche, Le Nucleaire Explique par des Physiciens, EDP Sciences, 2002.
R. K. Sinha, A. Kakodkar, Design and development of the AHWR - the Indian thorium fuelled innovative nuclear reactor, Nuclear Engineering and Design 236 (7-8) (2006) 683700.
M. M. Abu-Khader, Recent advances in nuclear power: A review, PROGRESS IN NUCLEAR ENERGY 51 (2) (2009) 225235.
H. E. Suess, H. C. Urey, Abundances of the elements, Rev. Mod. Phys. 28 (1956) 5374.
D. W. Jones, P. N. Leiby, I. K. Paik, Oil price shocks and the macroeconomy: What has been learned since 1996, Energy Journal 25 (2)
(2004) 132.
IEA, Energy Technology Perspectives 2010, IEA/OECD, 2010.
IPCC, Climate Change 2007: Mitigation of Climate Change. Contribution of Working Group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change, Cambridge University Press, 2007.
B. J. de Vries, D. P. van Vuuren, M. G. den Elzen, M. A. Janssen, The Targets IMage Energy Regional (TIMER) Model, Tech. rep., National
Institute of Public Health and the Environment (RIVM) (2001).
S. Messner, M. Strubegger, Users guide for MESSAGE III, Tech. rep., IIASA (1995).
URL http://www.iiasa.ac.at/Admin/PUB/Documents/WP-95-069.pdf
BP, BP Statistical Review of World Energy 2009 Workbook (2009).
URL http://www.bp.com/statisticalreview
W. J. Nuttall, Nuclear Renaissance, IOP Publishing, 2005.
IEA, World Energy Balances dataset 2012, IEA/OECD, 2012.
IEA, Oil Information 2011, IEA/OECD, 2011.
IEA, Transport, Energy and CO2, IEA/OECD, 2009.
O. P. R. van Vliet, T. Kruithof, W. C. Turkenburg, A. P. C. Faaij, Techno-economic comparison of series hybrid, plug-in hybrid, fuel cell and
regular cars, Journal of Power Sources 195 (19) (2010) 65706585.
L. M. Murphy, P. L. Edwards, Bridging the valley of death: Transitioning from public to private sector financing, Tech. rep., NREL (2003).
URL http://www.nrel.gov/docs/gen/fy03/34036.pdf
Global Land Cover, Global land cover 2000, Tech. rep., European Commission, Joint Research Centre (2000).
URL http://bioval.jrc.ec.europa.eu/products/glc2000/glc2000.php
UNEP, Solar and wind energy resource assessment (SWERA): Renewable energy resource explorer (RREX) (2011).
URL http://swera.unep.net/
3TIER, Global wind speed map (2011).
URL http://www.3tier.com
CIA, The World Factbook, Central Intelligence Agency, 2011.
3TIER, Global solar irradiance map (2011).
URL http://www.3tier.com
J. C. Rowley, Worldwide Geothermal Resources, in: L. M. Edwards, G. V. Chilingar, H. H. Rieke, W. H. Fertl (Eds.), Handbook of
Geothermal Energy, Houston, 1982, pp. 44174.
G. Mork, S. Barstow, A. Kabuth, M. T. Pontes, Assessing the Global Wave Energy Potential, in: OMAE2010 - 29th International Conference
on Ocean, Offshore Mechanics and Arctic Engineering, 2010.
R. Pelc, R. Fujita, Renewable energy from the ocean, MARINE POLICY 26 (6) (2002) 471479.
IEA, Annual report 2011. implementing agreement on ocean energy systems, Tech. rep., IEA (2011).
M. Rodier, Tidal Power: Trends and Developments, in: Proceedings of the 4th Conference on Tidal Power, 1992.
WPC, WEC, Survey of Energy Resources, Central Office, World Power Conference, 1986.
W. S. von Arx, Energy - Natural Limits and Abundances, Transactions-American Geophysical Union 55 (9) (1974) 828832.

An Assessment of Energy Resources For Global Decarbonisation PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

An Assessment of Energy Resources For Global Decarbonisation PDF

Uploaded by

Copyright:

Available Formats

4CMR Working Paper Series

An assessment of energy resources for global decarbonisation

author: Jean-Franc ois Mercure, email address: jm801@cam.ac.uk

4CMR Working Paper no 2

September 25, 2013

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4 Renewable energy resources

6 Summary of energy resources

9 Methodology and definitions

10 Global overview of stock energy resources

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

13 Introduction and use of the Appendix

14 Derivation of distribution functions and cost-supply curves

15 Cost-supply curve calculation methodology per resource type

16 Data tables and figures

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

Cambridge Centre for Climate Change Mitigation Research (4CMR)

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cost Distribution [ EJ/(USD/MWhy ) ] Productivity Distribution ( km2 )

Cambridge Centre for Climate Change Mitigation Research (4CMR)

Energy Flow ( MWhy

Cambridge Centre for Climate Change Mitigation Research (4CMR)

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

LCOE (USD / MWh)

LCOE (USD / MWh)

LCOE (USD / MWh)

Cost (USD / GJ)

A2: 305 EJ/y

LCOE (USD / MWh)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

6. Summary of energy resources

Table 2: Summary table for all energy resources.

Cambridge Centre for Climate Change Mitigation Research (4CMR)

4CMR Working Paper no 2

Cambridge Centre for Climate Change Mitigation Research (4CMR)

Cambridge Centre for Climate Change Mitigation Research (4CMR)

9.1. The economic potential of natural resources