[Below are excerpts from this 22 page paper. Alice Friedemann]
Davidsson, S., et al. (2014) Growth curves and sustained commissioning modelling of renewable energy: Investigating resource constraints for wind energy. Energy Policy.
Although the wind itself is a type of renewable energy, the wind turbines converting the kinetic energy in the wind into electrical energy are not renewable and are built using a wide range of non-renewable resources.
Several recent studies have proposed fast transitions to energy systems based on renewable energy technology. Many of them dismiss potential physical constraints and issues with natural resource supply, and do not consider the growth rates of the individual technologies needed or how the energy systems are to be sustained over longer time frames. A case study is presented modelling potential growth rates of the wind energy required to reach installed capacities proposed in other studies, taking into account the expected service life of wind turbines.
The annual installation and related resource requirements to reach proposed wind capacity are quantified and it is concluded that these factors should be considered when assessing the feasibility, and even the sustainability, of fast energy transitions. Even a sustained commissioning scenario would require significant resource flows, for the transition as well as for sustaining the system, indefinitely. Recent studies that claim there are no potential natural resource barriers or other physical constraints to fast transitions to renewable energy appear inadequate in ruling out these concerns.
A few recent peer reviewed studies stand out by proposing future energy systems almost completely based on energy from the wind and the sun, claimed to be achievable as soon as the year 2050, or even more rapidly by 2030 (García- Olivares et al., 2012; Jacobson and Delucchi, 2009; Kleijn and van der Voet, 2010).
Substituting the entire current energy system based on fossil fuels with renewable energy technologies involves up-scaling a disparate set of small scale industries, and the time-frame to do this within only a couple of decades, can appear optimistic. The implications of the fast growth of the renewable energy technologies needed to do this are often not adequately addressed in the studies proposing future energy systems based on renewable energy. The question of how these energy systems are then to be sustained over a longer time scale are usually not considered.
This study aims to add the perspectives of time and scale to evaluating the feasibility of fast energy transitions by taking account of annual growth rates needed to reach proposed future energy systems as well as investigating how an energy system based on renewable energy technologies could be sustained in the long run. This is mainly done by modelling growth patterns needed to reach the installed capacities of wind energy proposed in other studies, taking account of the life expectancies and need for replacement of technology, using wind energy as an example. The requirement of natural resources for the construction of wind energy is quantified on an annual basis to examine the impact on views of potential material constraints.
The growth of renewable energy technologies needed for an energy transition must inevitably come with the growth of an industry capable of manufacturing and installing that technology, capital to finance these investments, as well as an increased demand for certain natural resources.
Renewable energy technologies such as wind and solar energy are more metal intensive than current energy sources and a transition to renewable energy would increase demand for many different metals (Kleijn et al., 2011). Several different critical metals have been identified as potential bottlenecks in the deployment of “low-carbon energy technologies” (Moss et al., 2011). It has also been argued that a shift to an energy system based on renewable energy would inevitably be largely driven by fossil fuels, and a fast growth of renewables would actually add new fossil fuel demand to current demand during a transition period (Moriarty and Honnery, 2009).
The concept of “energy return on investment” (EROI) appears lower for renewable energy technologies than many conventional fossil fuels we currently rely on for our energy supply (Hall et al., 2013). Concerning solar photovoltaics (PV), it has been suggested that high energy input for the production of crystalline silicon solar cells could be a constraint for the growth of this technology, while current thin film technologies could never reach significant production levels due to the use of scarce materials (Tao et al., 2011). Dale and Benson (2013) even claim that the solar PV industry has not yet paid back any net energy to society, partly due to its high relative growth rates, and concludes that both the timing and magnitude of energy inputs and outputs are important factors in determining an energy balance for the solar industry.
Others raise issues with the variable production of electrical energy from wind and solar energy as well as the large amount of capital needed for investment in new energy production as potential constraints on this development (Trainer, 2013, 2012).
Installed wind capacity Jacobson and Delucchi (2009) describe an energy system consisting of 51% wind energy and 40% solar energy that is “technically possible” to achieve before 2030. This scenario is further elaborated on in Jacobson and Delucchi (2011) and Delucchi and Jacobson (2011), where the time frame is postponed due to difficulties in implementing the necessary policies by 2030, but it is still said to be technically feasible to achieve by 2030. Kleijn and Van der Voet (2010) present a similar scenario, with slightly more wind energy but many times more solar PV, since the total energy demand is assumed to be much larger. García-Olivares et al. (2012) propose an energy mix similar to the Jacobson and Delucchi (2009) scenario, but state that solar PV is unlikely to be able to reach these levels due to constraints induced by scarce materials used for solar PV technology and propose using concentrating solar power (CSP) instead. Table 1 summarizes the main features of these three studies as well as the current situation as of 2012.
The studies described in Table 1 all propose energy systems completely based on renewable energy technology, with wind and solar energy making up almost the entire global energy supply by 2030 or 2050. Although important differences occur between the different studies, some interesting similarities exist. While the solar energy contributions vary greatly both in size and technologies chosen, the assumed contribution from wind is very similar between the studies, with suggested installed capacities ranging from 18 to 24 TW. All three studies discuss potential constraints caused by natural resources and conclude that this factor will likely not constrain the development towards the proposed energy future. The growth patterns needed for the individual technologies is not given much attention, and when growth rates of technologies are mentioned it appears as if exponential growth rates are assumed, or at least deemed feasible.
This study investigates the implications of fulfilling these growth patterns by letting wind energy grow exponentially reaching 19 TW by 2030 and 24 TW by 2050. Although not specified in the studies, these capacities are then assumed to be sustained to the year 2100, to be able to investigate the implications of sustaining this capacity.
Laxson et al. (2006) describes a sustained manufacturing model, where installed capacity of wind energy grows to reach 1 %, 20% and 30% of U.S. electricity demand by 2020 or 2030. After 25 years the capacity installed 25 years earlier are replaced (repowered). The need to replace the capacity after the end of the service life of the wind turbines affects the desired manufacturing capacity of the wind industry. If the installed capacity of wind is to be sustained over a longer time frame, an industry capable of replacing the capacity taken out of use must exist. If the growth trajectory is too slow to reach a manufacturing capacity large enough to replace the old turbines in the future, the actual wind capacity in use can in fact see a drop after the initial goal is reached. On the other hand, if the manufacturing capacity is expanded too fast, the demand for new turbines will drop and leave manufacturing capacity idle.
The sustained commissioning model in this study builds upon the ideas proposed by Laxson et al. (2006), with some modifications. The use of the word commissioning instead of manufacturing is proposed to highlight the fact that taking wind capacity into use is not only about physically producing wind turbines, but requires an entire industry of getting the right materials, manufacturing parts, permission to install wind farms, assembling and installing turbines, as well as getting the wind farms connected to an electrical grid capable of transporting the power to consumers.
Höök et al. (2012) reviewed historical growth rates of energy output from the six energy resources considered as global energy systems, defined as energy sources contributing over 100 Mtoe, or supplying about 1% of global annual primary energy. These include oil, gas, coal, biomass, hydropower and nuclear power. Generic growth behavior for these six energy systems was found, with growth rates decreasing as the energy output increased. It is stated that none of the fossil fuels have grown at more than 10% over longer time periods, and not even the “oil boom” showed sustained growth rates of more than around 7%. The growth rates for nuclear and hydropower show similar behavior as those seen for fossil fuels, despite fundamental differences in technology, suggesting that similar growth patterns could be expected for other energy technologies as well.
Technology can be taken out of use for several different reasons, making the assumption of expected service life somewhat difficult to estimate. However, it must be considered certain that they will not last forever. In the case of wind turbines, the end-of-life can be reached due to technical failure or fatigue, or when the turbine no longer satisfies the need or expectations of the user, when a wind farm is either decommissioned or repowered, where the individual turbines are replaced with new ones (Ortegon et al., 2013). The assumed service life will have a significant impact on annual installations needed in the models in this study.
The question then is what a reasonable estimate of service life for a wind turbine is. Ortegon et al. (2013) state that the designed life expectancy for a wind turbine is 20-30 years, but assumes a service life of 20 years. Laxson et al. (2006) state that the design service life of a wind farm is 20 to 30 years but use a 25 year service life in the models. Within the life cycle assessment (LCA) community it appears to be somewhat of a standard to assume a 20 year service life. Kubiszewski et al. (2010) presents a meta-analysis of 119 different turbines from 50 different analyses between 1977 and 2006, where a vast majority assumed a 20-year life span. Davidsson et al. (2012) looked at ten more recent LCAs of wind turbines and found similar tendencies. Dolan and Heath (2012) reviewed and harmonized 72 LCAs on wind turbines and concluded that 20 years was the most commonly cited lifetime estimate as well as a common design life for modern wind turbines. Basically, a 20 year service life appears to be the most reasonable assumption based on current literature.
One of the first countries to build large quantities of wind energy was Denmark, and data on both commissioned and decommissioned facilities exist all the way back to 1977 (Energistyrelsen, 2014). Using the assumption that the wind turbines will be in use for 20 years it is then possible to compare how much capacity that should be decommissioned 20 years after its construction with the actual numbers on decommissioning. Figure 1 shows these theoretical numbers on decommissioning as well as actual historical decommissioned capacity. Although they do not correlate exactly, especially since a large amount of turbines was taken out of use in the year 2002, they appear to follow a similar pattern, and the total cumulative decommissioned capacity of 431 MW comes remarkably close to the theoretical number of 468 MW.
Including an assumption on service life for a technology can have large impacts on the annual installation need for the growth period, but also for the energy system in a longer time frame. Looking at a scenario for 2050, assuming a 20 year service live of wind turbines, only turbines built after 2030 will even be in use at that time. Turbines built between now and 2030 will only be in service during the transition and for scaling up the industry. After 2050 the old turbines will need to be replaced, so an industry capable of sustaining this level of production needs to be in place.
Wind turbines can roughly be divided into two categories: geared turbines and gearless turbines. The turbines can operate with either a fixed speed or limited variable speed concept, both cases using a three-stage gearbox. Turbines operating with variable speed can use either a gearbox or a direct drive train concept. Some concepts use significant amounts of scarce materials in their design. For instance, permanent magnet synchronous generators (PMSG), which is a widely used generator concept with a direct drive train, uses significant amounts of rare earth elements (REEs). These generators often operate without gears, which can be beneficial since the gearbox often needs maintenance. There are other direct drive concepts that do not use these materials, such as induction generators and exited synchronous generators (EESG). The need for rare earth elements is estimated to be 160-200 kg/MW for generators used in direct drive concepts, while PMSG designs used in combination with a gearbox the need for REE is reduced to about 30 kg/MW (Buchert, 2011).
As a constraint for a total expansion of wind energy on a global scale the significance of these materials are often dismissed since designs not relying on them would likely arise if the supply of these materials becomes increasingly limited.
Wind turbines require large amounts of other materials, such as steel and copper as well, and these materials are quantified in the case study as an example of resource requirements. This study uses the assumption that 1 MW of wind capacity requires 140 tons of iron and steel and 2 tons of copper, as described by Kleijn and Van der Voet (2010).
Figure 2a presents the cumulative growth curves of wind capacity enabling 19 TW by 2030 and 24 TW by 2050 with exponential growth profiles. Figure 2b shows the resulting annual commissioning required to reach 19 TW wind capacity by 2030, as well as what is required to sustain this capacity in the future. It can be seen that not only the cumulative installations, but also the annual installations grow exponentially, leading to quite extreme annual installations at the end of the growth period. Reaching 19 TW by 2030 with exponential growth means that 21 % of all installed capacity would be installed in the final year, and 68 % would be installed in the last 5 years. Reaching 24 TW by 2050 with exponential growth means that 11% of all the capacity would be installed in the final year, and 45% would be installed in the last 5 years (Figure 2c). Sustaining these capacities will require an annual commissioning growing exponentially in a kind of cyclic behavior.
Similar results were found by Honnery and Moriarty (2011) who used 3 different exponential growth rates reaching 2 different installed capacities of wind power and found that these growth rates leads to “boom and bust cycles” in equipment manufacture as well as net energy output from the system.
Assuming double digit exponential growth of energy technologies for decades after reaching significant contributions to the global energy system can simply not be considered realistic since the pure arithmetic of such growth patterns leads to unreasonable expectations on annual installation rates. Further discussions on the nature of exponential growth can be found in other studies (Bartlett, 1993; Meadows et al., 1972).
Figure 2. a) Cumulative installed capacity of wind power reaching 19 TW by 2030 and 24 TW by 2050 with exponential growth. b) Annual commissioning of wind capacity required for reaching 19 TW by 2030 and sustaining this capacity. c) Annual commissioning of wind capacity required for reaching 24 TW by 2050 and sustaining this capacity.
Reaching 24 TW by 2050 alone is modeled using a logistic function. Figure 3a describes a logistic growth curve fitted to the historic data and constrained at 24 TW wind capacity. This appears to be a more realistic growth pattern than exponential growth, but what is not always considered is that the annual additions needed will not only be installing new turbines, but also replacing old turbines at the end of their service life. Assuming a 20 year service life for a wind turbine, the annual requirements of replacing old turbines can be modeled with a second logistic curve with a 20 year time lag. Figure 3b shows the annual commissioned capacity needed both for the net growth as well as replacing old capacity taken out of use.
The maximum annual installations needed for logistic growth is much lower than the exponential case, but reaching 24 TW still requires significant numbers. Also, as can be seen in Figure 3b, assuming logistic growth of cumulative installed capacity in this case means that the total annual installations needed when taking account for replacing old turbines creates a dip in annual installation need before rising again. This type of pulsing behavior is commonly seen in nature (Odum, 2007), and might not be an unrealistic scenario. However, it might not be optimal, since this would create an industry capable of installing more wind capacity in a year than is needed to sustain this in the long run.
Less scarce materials are commonly ruled out as constraints based on quite simple arguments, but for a complete transition to a renewable energy system even common materials have been mentioned as potentially problematic. Kleijn and van der Voet (2010) suggest that the sheer size of the proposed transition would challenge production even for “bulk materials” such as steel and copper.
Constructing the wind capacity of 24 TW would only demand a few per cent of global iron ore and copper reserves. However, using the growth patterns from the case study, this total resource requirement can be spread out over the time period leading up to the proposed realization year and be translated into annual requirements for the different resources. These annual quantities can then be compared with projections for future production of these resources. It could also be useful to take account for competing demand from other uses for a more complete systems view.
The quantities presented in Table 2 could give an indication of the size of the annual resource requirements for building these quantities of wind capacity. Table 2 describes the resulting maximum annual installations
Even in the sustained commissioning model, the annual installation of 1.2 TW needed to sustain the 24 TW wind capacity leads to significant annual requirements for copper and steel.
Under these assumptions, only sustaining the 24 TW of wind energy, assumed to provide 15% of global energy demand by Kleijn and Van der Voet (2010), would need the equivalent of 11% of total global steel production and 14% of global copper production (based on 2012 rates of production).
This means that reaching and sustaining this installed wind capacity would require quantities of steel that is similar to the current automotive industry, that used 12% of the steel produced in 2011, while the entire sector of electrical equipment used only around 3% (World Steel Association, 2012). The amount of copper needed for the turbines is comparable to what is used for making electric motors, of around 12% of the global copper production, while the electric energy transmission sector use about 26% (Achzet et al., 2011).
This study makes no attempt to project what the future energy systems might look like, neither on the demand nor the supply side. Instead, the assumptions of future installed capacity of wind energy for the case study is taken directly from these other studies, and translated into possible growth patterns. It should be mentioned that the works used in this case study are quite extreme when it comes to proposed installed capacities of wind and solar energy compared to most other studies proposing similar energy transitions. However, they are still considered relevant since they are widely cited in peer reviewed scientific journal articles.
During the growth phase this demand would be additional to current demand and must be assumed to come from supplementary production, and even if the replacement of turbines in the future would be based on recycling old turbines, a similar sized commissioning industry would be needed, as well as an industry capable of recycling the materials and making them available for new turbines. The pure scale of creating and sustaining this type of energy system is simply massive.
In the case of wind energy, metals considered somewhat scarce, such as neodymium, are sometimes mentioned as a potential issue, but “bulk” materials such as steel and copper are usually dismissed as potential constraints. However, none of them pay much attention to assumed growth rates or what resource flows that would be needed to sustain the growth or to sustain the proposed energy system in the future.
Three common ways to evaluate natural resource constraints in other studies have been found. First, the “Reserve-to-production ratio” (R/P ratio), comparing the current annual production to reserve estimates is a very common method. Secondly, simply comparing the total demand incurred by the proposed energy system to reserve estimates is a frequently used method. Thirdly, simply stating that the materials used are theoretically recyclable is sometimes used as an argument that no natural resource constraints will occur. All three of these arguments have their merits and can be used to make fast and easy estimates of natural resource constraints, but using any of them to completely dismiss potential problems with natural resource supply appears questionable.
An example of R/P ratio being used to disregard natural resource constraints can be found in Jacobson and Delucchi (2011), where it is stated that the world have “somewhat limited reserves” of iron ore, which is claimed to last for 100-200 years at current production. However, this assumes that annual production remains constant and global steel production is currently increasing rapidly, and realizing the Jacobson and Delucchi (2009) scenario would mean a significant increase of an already expanding demand for steel. Comparing current production to reserve estimates could give a first indication of potential constraints, but it appears insufficient to motivate a total dismissal of problems that might occur. Bartlett (2006) describes several problems with using the R/P ratio for a resource under growing production, and states that it gives rise to unwarranted optimism.
The method of comparing the total requirements of a resource for reaching a future energy system to estimated reserves can be found in García-Olivares et al. (2012), where it is stated that the complete power system needed for the energy system described would need 40% of total estimated copper reserves. Adding assumptions of the copper needed from the demand side of the transport sector García-Olivares et al. (2012) reach a total of 60% of global copper reserves. This method has the potential of indicating if the quantities needed could be a problem. For instance, the claim that realizing the energy system proposed in García-Olivares et al. (2012) could demand 60% of the current copper reserves appear like extraordinary quantities, although reserve estimates can change with time.
This method does not say anything about what resource flows would be needed and how fast the materials could be brought to market.
The third common argument to dismiss potential resource constraints is using the simple fact that some materials are recyclable. Jacobson and Delucchi (2011) argue that some rare resources, such as neodymium for electric motors and generators, platinum for fuel cells and lithium will have to be recycled or replaced with less scarce materials to reach a 100% renewable energy system, unless additional resources are located. Jacobson and Delucchi (2009) claim that there are indications that there are not enough economically recoverable lithium to build “anyway near the number of batteries needed in a global electric- vehicle economy”, but at the same time state that recycling could change this equation. There is no doubt that recycling would be important for sustaining a “sustainable” energy system in the future, but this does not mean that recycling will change the total amount of materials needed in the system at a given moment in time. The same atoms simply cannot both be in use and recycled to build other technology at the same time. The minimum amount of a resource needed to sustain the system simply does not change because of recycling. A more comprehensive discussion on recycling using the case of lithium is available in Vikström et al. (2013).
The end of life recycling rate (EOL-RR) appears to be around 70-90% for iron and steel, but since the steel demand is growing and is commonly used for long lived uses, the recycled content (RC) in new material is lower at around 32-52%, while the same factors for copper has been estimated to be between 43-53% and 22-37% respectively (Graedel et al., 2011).
While some expect that the recycling rates for metals used in electricity generation technologies will be higher due to expected high collection rates (Elshkaki and Graedel, 2013), others mentions different situations that could lead to materials not being recycled (Davidsson et al., 2012).
For some materials, recycling can even be technically problematic. In the case of REEs, such as neodymium, recycling is commonly mentioned as being important for a sustainable energy system, but at the moment no infrastructure for recycling of REEs from the permanent magnets exists and the end-of-life recycling rate is estimated to be less than 1% (Buchert, 2011).
One important problem with recycling rare earth elements is the fact that the metals oxidize quickly and disappear in the slag (Buchert et al., 2009). However, it could be technically possible to reach recycling rates of more than 90% for both neodymium and dysprosium (Schüler et al., 2011). A sustainable energy system would have to recycle as much as possible of the materials after the end of the service life, but even if recycling rates would eventually come close to 100%, the industry for replacing old technologies would still demand large resource flows indefinitely. The case study culminating in 24 TW of installed wind capacity demands an equivalent of over 10% of current (2012) global annual demand of bulk materials such as copper and steel. Even if these turbines were to be recycled at the end of their life and built using only recycled materials, it would still mean large material flows.
Another important perspective is the fact that this study only includes the material demands for constructing wind energy,
An energy system completely based on renewable energy technology would likely need more of these technologies, but also energy storage and transmission capable of creating a functioning energy system. For instance, Barnhart and Benson (2013) investigates energy and material requirements for different energy storage technologies and concludes that building an energy storage capacity that could be required in the future require amounts of materials and energy that are comparable to current annual production values.
An industry growing too fast can mean that the industry consumes more energy than it produces on an annual basis (Honnery).
There are many other examples of potential constraints on the growth of renewable energy technology, many of which are discussed by others. IEA (2013b) mentions costs, grid integration issues and permit issues as obstacles to a goal of 18% of global electricity from wind energy by 2050.
For wind energy, constructing the wind turbine and the connected capital costs constitute the majority of the total cost, with 76 – 85% of the total cost being capital cost (Timilsina et al., 2013). Financing for this cost needs to be in place before the wind capacity can be commissioned. Jacobson and Delucchi (2009) state that the construction of the proposed energy system would cost around 100 trillion USD over 20 years (not including transmission), which will be paid back by the sale of electricity and energy. Trainer (2012) interprets this as an investment of 5 trillion USD annually would be needed, which is said to be around 11 times the early 2000s annual investments in energy of around 450 billion USD. However, as discussed in this paper, this type of growth pattern is not very realistic.
The variability of production and grid integration is commonly suggested as the main barriers for implementation of renewable energy and it has even been suggested that this factor limits penetration rates of wind energy to 2 % of electricity production (Lenzen, 2010). These factors are discussed in more detail in other studies (Trainer, 2013, 2012).
The fact that energy production from renewable energy technologies is intermittent and non-dispatchable can also be argued to add to the total costs due to the need for backup power (Larsson et al., 2014).
The grid improvements and backup power requirements have to be in place before the variable energy production is taken into use, so the estimated growth curves can prove important for these aspects as well.
Although these technologies are likely more sustainable than fossil fuels, they are not without environmental impacts and are built using non-renewable resources. They should therefore not automatically be considered sustainable. A rapid growth in these technologies will even increase demand for a variety of different resources. Suitable growth rates of energy technologies, as well as how an energy system can be sustained over a longer time frame, should be considered when discussing sustainable energy systems for the future.
Achzet B., Reller A., Zepf V., University of Augsburg, Rennie C., BP, Ashfield M. and Simmons J., ON Communication, 2011. Materials critical to the energy industry. An introduction. http://www.bp.com/content/dam/bp/pdf/sustainability/group- reports/Materials_March2012.pdf Ang, B.W., Ng, T.T., 1992. The use
Bartlett, A.A., 2006. A Depletion Protocol for Non-Renewable Natural Resources: Australia as an Example. Natural Resources Research 15, 151–164.
Buchert, M., 2011. Rare Earths – a Bottleneck for future Wind Turbine Technologies? Presented at Wind Turbine Supply Chain & Logistics, Berlin, 29 August 2011. Available from: http://www.oeko.de/oekodoc/1296/2011-421-en.pdf
Dale, M., Benson, S.M., 2013. Energy Balance of the Global Photovoltaic (PV) Industry – Is the PV Industry a Net Electricity Producer? Environ. Sci. Technol. 47, 3482–3489.
Elshkaki, A., et al., 2013. Dynamic analysis of the global metals flows and stocks in electricity generation technologies. Journal of Cleaner Production, 59, 260-273.
Graedel, T. E., et al., 2011. Recycling rates of metals: A status report. Journal of Industrial Ecology. 15(3), 355-366.
Larsson, S., et a;., 2014. Reviewing electricity production cost assessments. Renewable and Sustainable Energy Reviews 30, 170–183.
Laxson, A., Hand, M.M., Blair, N., 2006. High Wind Penetration Impact on U.S. Wind Manufacturing Capacity and Critical Resources. National Renewable Energy Laboratory. Report No. NREL/TP 500-40482.
Lenzen, M.,2010. Current State of Development of Electricity-Generating Technologies: A Literature Review. Energies 3, 462–591. Lund, H., 2007. Renewable energy strategies for sustainable development. Energy 32, 912–919.
Meadows, D.H., Meadows, D.L., Randers, J., Behrens III, W.W., 1972. The Limits to Growth. Earth Island Limited, London.
Tao, C.S., Jiang, J., Tao, M., 2011. Natural resource limitations to terawatt- scale solar cells. Solar Energy Materials and Solar Cells 95, 3176–3180.
Trainer, T., 2012. A critique of Jacobson and Delucchi’s proposals for a world renewable energy supply. Energy Policy 44, 476–481.
Trainer, T., 2013. 100% Renewable supply? Comments on the reply by Jacobson and Delucchi to the critique by Trainer. Energy Policy 57, 634–640.
USGS, 2013. 2011 Mineral Yearbook: Rare Earths [Advanced Release]. Available from: http://minerals.usgs.gov/minerals/pubs/commodity/rare_earths/myb1-2011-raree.pdf
Vikström, H., Davidsson, S., Höök, M., 2013. Lithium availability and future production outlooks. Applied Energy 110, 252–266. Wilson, C., Grubler, A., Bauer, N., Krey, V., Riahi, K., 2013. Future capacity growth of energy technologies: are scenarios consistent with historical evidence? Climatic Change 118, 381–395. 395. 67-8