Assessing vulnerabilities and limits in the transition to renewable energies: Land requirements under 100% solar energy scenarios

1 Assessing vulnerabilities and limits in the transition to renewable energies: Land requirements under 100% solar energy scenarios Iñigo Capellán-Pérez*,a,b, Carlos de Castrob,c, Iñaki Artod aInstitute of Marine Sciences, ICM-CSIC. Passeig Marítim de la Barceloneta, 37-49, 08003 Barcelona, Catalonia, Spain. bResearch Group on Energy, Economy and System Dynamics, University of Valladolid, Spain. cApplied Physics Department, Escuela de Arquitectura, Av Salamanca, 18, University of Valladolid, 47014 Valladolid, Spain. [email protected]a.es dBasque Centre for Climate Change (BC3), Sede Building 1, 1st floor, Scientific Campus of the University of the Basque Country, 48940 Leioa, Spain. [email protected] *Corresponding author: [email protected]. Authorized Author manuscript published in “Renewable & Sustainable Energy Reviews”: Iñigo Capellán-Pérez, Carlos de Castro & Iñaki Arto: “Assessing vulnerabilities and limits in the transition to renewable energies: Land requirements under 100% solar energy scenarios”. Renewable & Sustainable Energy Reviews (September 2017). https://doi.org/10.1016/j.rser.2017.03.137. http://www.sciencedirect.com/science/article/pii/S1364032117304720 © 2017. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 2 Abstract The transition to renewable energies will intensify the global competition for land. Nevertheless, most analyses to date have concluded that land will not pose significant constraints on this transition. Here, we estimate the land-use requirements to supply all currently consumed electricity and final energy with domestic solar energy for 40 countries considering two key issues that are usually not taken into account: (1) the need to cope with the variability of the solar resource, and (2) the real land occupation of solar technologies. We focus on solar since it has the highest power density and biophysical potential among renewables. The exercise performed shows that for many advanced capitalist economies the land requirements to cover their current electricity consumption would be substantial, the situation being especially challenging for those located in northern latitudes with high population densities and high electricity consumption per capita. Assessing the implications in terms of land availability (i.e., land not already used for human activities), the list of vulnerable countries enlarges substantially (the EU-27 requiring around 50% of its available land), few advanced capitalist economies requiring low shares of the estimated available land. Replication of the exercise to explore the land-use requirements associated with a transition to a 100% solar powered economy indicates this transition may be physically unfeasible for countries such as Japan and most of the EU-27 member states. Their vulnerability is aggravated when accounting for the electricity and final energy footprint, i.e., the net embodied energy in international trade. If current dynamics continue, emerging countries such as India might reach a similar situation in the future. Overall, our results indicate that the transition to renewable energies maintaining the current levels of energy consumption has the potential to create new vulnerabilities and/or reinforce existing ones in terms of energy and food security and biodiversity conservation. Key-words: Solar potential, Energy footprint, Land-use, Transition to renewable energies, Energy security. Table of contents 1. Introduction .......................................................................................................................................................... 3 2. Materials and methods ......................................................................................................................................... 6 2.1. Multi-regional input-output model ............................................................................................................... 7 2.1.1. Electricity and final energy consumption .................................................................................................. 7 2.2. Solar power density at country level ................................................................................................. 10 2.2.1. Solar irradiance (Ii) .................................................................................................................................. 10 2.2.2. Cell efficiency conversion (f1) .................................................................................................................. 12 2.2.3. Average performance ratio over the park’s life cycle (f2) ....................................................................... 13 2.2.4. Land-occupation ratio (f3) ....................................................................................................................... 13 2.2.5. Comparison with other values estimated in the literature .................................................................... 16 2.3. Overcapacity and storage requirements due to short-term and seasonal variations ................................ 17 3 2.4. PV potential on buildings and in urban areas ............................................................................................. 21 2.5. Land-use requirements for solar power: summary .................................................................................... 23 3. Results and discussion ........................................................................................................................................ 24 3.1. 100% solar electricity mix scenario ............................................................................................................. 24 3.2. 100% solar final energy mix scenario .......................................................................................................... 28 4. Assessment of assumptions and uncertainties in estimating the land-use requirements of the transition to RES 31 4.1. Country’s self-sufficiency in electricity/final energy and 100% solar share ............................................... 31 4.2. Static vs. dynamic projections ..................................................................................................................... 33 4.3. Estimation of solar power density at the country level .............................................................................. 34 4.4. Not accounting for EROI .............................................................................................................................. 34 4.5. PV potential in buildings ............................................................................................................................. 35 5. Conclusions ............................................................................................................................................................. 35 Acknowledgements ..................................................................................................................................................... 37 Appendix A: Electricity and final energy consumption by country for 2009 .................................................................. 37 Appendix B: CRC and SV values per country ................................................................................................................... 38 Appendix C: Potential electricity produced by rooftop PV ............................................................................................. 39 Appendix D: Share of electricity consumption covered by hydro generation ................................................................ 41 Appendix E: Total land and land per capita by scenario ................................................................................................. 41 Appendix D: Land availability at country level ................................................................................................................ 44 References .................................................................................................................................................................. 44 1. Introduction Most governments are developing policy frameworks to promote the penetration of renewable energy sources (RES) to improve energy security (increasingly threatened by the depletion of conventional fossil fuels) while mitigating emissions to limit anthropogenic climate change and other negative externalities of conventional energy sources (IPCC, 2014; Johansson, 2013; REN21, 2015; WEO, 2014). Among renewables, wind and solar are estimated to have the greatest potential (de Castro et al., 2013; IPCC, 2011; Smil, 2010), with projections often assuming that the resource base provides no practical limitation if adequate investments are forthcoming (e.g., IPCC (2011)). While fossil fuels represent concentrated deposits of energy and thus can be exploited at high power rates (200- 11,000 We/m2), the technologies harnessing renewable sources are characterized by power densities several orders of magnitude lower. Hence, for delivering the same power, RES are substantially more land intensive (Smil, 2015). For example, typical ranges of net power density found in the literature are: 2-10 We/m2 for solar power plants, 0.5-7 We/m2 for large hydroelectric, 0.5-2 We/m2 for wind; and ~0.1 We/m2 for biomass (de Castro et al., 2014; MacKay, 4 2013; Smil, 2015). While wind farms are partially compatible with other uses (e.g., agriculture) or can be located offshore, biomass plantations, hydroelectric reservoirs and solar farms tend not to allow double use, that is, in practice they monopolize the occupied land. In the case of solar power, the potential in urbanized areas is limited due to the fact that cities are currently not designed to maximize solar reception (Izquierdo et al., 2011; La Gennusa et al., 2011; Ordóñez et al., 2010; Sorensen, 1999). Hence, the transition to RES will add to the pressure in the global competition for land, which is already driven by many factors (Smith et al., 2010). In particular, the dedication of land to produce energy has been identified as a potential concern not only for preserving natural ecosystems, their services and biodiversity, but also because of its competition with land use to cover human needs (i.e., food, fiber, shelter and infrastructure). These concerns arise in parallel with the current rapid expansion of modern RES technologies and the steady decrease in their costs over recent years (Deutsche Bank, 2015; REN21, 2015). Thus, this transition could aggravate existing vulnerabilities and create new ones in terms of energy security, biodiversity loss, and food sovereignty, among others (Johansson, 2013; MacKay, 2013; Nonhebel, 2003; Rao and Sastri, 1987; Scheidel and Sorman, 2012; Smil, 1984). As a recent example, the occupation of just ~0.1% of Italian agricultural surface area by PV systems provoked an intense debate in the country that ultimately lead to the ban of incentives for this technology on agricultural soil (Squatrito et al., 2014). The relevance of the land requirements of renewables is the subject of ongoing debate, with most studies focusing on 100% RES scenarios having estimated that the additional land requirements will not be a compelling constraint for the transition (e.g., Jacobson and Delucchi (2011), WWF (2011), Jacobson et al., (2015), Teske et al., (Greenpeace et al., 2015) and García-Olivares (2016)), while a few have found land availability to be a relevant biophysical constraint that may limit the feasibility of the transition within the current socio-economic system (e.g., Mackay (2013)). With our work, we contribute to the debate by estimating a conservative, lower bound for the land-use requirements to supply all current consumed electricity and final energy domestically with solar energy for 40 countries, devoting special attention to uncertainties such as future efficiency improvements. We focus on solar energy since, among renewables, it has the highest power density and biophysical potential (de Castro et al., 2013; IPCC, 2011). First, we concentrate on the land-use requirements and biophysical feasibility of supplying all current consumed electricity with solar technologies in a given region as proposed by Denholm and Margolis (2008) for the states of the USA and Šúri et al. (2007) for 30 European countries. A few estimates of solar land-use requirements have been published to date by various authors for advanced capitalist economies such as the USA and European states (Denholm and Margolis, 2008; MacKay, 2013; MIT, 2015; Šúri et al., 2007; Turner, 1999), and by Jacobson and Delucchi (2011) at a global level, while other studies have focused on comparisons with other energy technologies (Fthenakis and Kim, 2009). In general, these analyses have come up with relatively low values of solar land-use requirements, thereby minimizing the importance of land to sustain high penetration levels of solar energy. For example, Šuri et al. (2007) found that just 0.6% of the land surface area of the EU25 and 5 EU-candidate countries, all 5 corresponding to rooftop photovoltaic (PV), would suffice to cover the total electricity demand, with a range of 0.1- 3.6% depending on the country. Denholm and Margolis (2008) found that the land required to supply the electricity consumed in the USA by solar plants (assuming 25% on rooftops) was between 0.3 and 0.7% of the total surface area (with a range of 0.1-8% depending on the state). However, these analyses have not considered two key issues included in our analysis, and these have the potential to substantially increase the land-requirements of solar power plants:  In a 100% solar-based energy system, a substantial redundant capacity should be deployed in combination with storage capacity to cope with the intermittence and seasonal variability of the solar resource (MacKay, 2013; Trainer, 2010, 2012, 2013a),  The real land occupation of solar technologies is five to ten times higher than the estimates usually considered, which are based on ideal conditions (de Castro et al., 2013; MacKay, 2013; Ong et al., 2013; Smil, 2015). Although a diversified supply combining different renewable resources as a function of their local availability would make it possible to reduce the overcapacity and storage requirements to cope with solar intermittency to some extent, this effect would be partially offset by the fact that for most countries solar has a power density three to five times higher than wind, and one to two orders of magnitude higher than bioenergy and is slightly better than large hydropower (de Castro et al., 2014; MacKay, 2013; Smil, 2015). The approach applied does not fully correspond to an “extreme scenario” for two additional reasons: (1) there is a positive relation between the electricity consumption per capita and income (i.e., most countries have been experiencing electrification of the energy system for decades), and (2) the future deployment of renewables will require that this trend be intensified since they mainly produce electricity (Armaroli and Balzani, 2011; Smil, 2008). In the period 1990 to 2007, the annual growth in the global net electricity production (+1.9%) outpaced the annual growth in total energy consumption (+1.3%), a trend which is expected to strengthen in the next few decades. For example, the International Energy Agency (IEA) in its New Policies Scenario expects the world electricity demand to grow by 2.1% per year on average between 2012 and 2040 (i.e., +80% cumulative growth in the period), its share of total energy use rising in all sectors and regions (WEO, 2014). Thus, the land occupation by solar/RES in the future is likely to be higher than estimated in our study for current electricity consumption. Additionally, a third factor, critical for assessing potential vulnerabilities, is considered: over recent years, advanced capitalist countries have specialized in economic activities with high added value (reducing their share of energy intensive sectors and manufacturing industries) while some emerging economies, like China and India, have undergone a process of rapid industrialization, increasing their share in the global economy, and are exporting enormous volumes of manufactured products to developed countries (Baiocchi and Minx, 2010; Weber, 2009). This shift of economic activities between countries has also had consequences in terms of energy use. Arto et al. (2016) showed that an increasingly large proportion of the energy used by emerging countries is being devoted to sustain 6 the welfare of advanced capitalist economies by means of international trade. Hence, together with data on the electricity use per country, we will consider the net electricity consumption after accounting for international trade for each country, i.e., its electricity footprint, estimated from the multi-regional input-output model (MRIO) WIOD (Dietzenbacher et al., 2013). In a second stage, we replicate the analysis to explore the land implications and biophysical feasibility, for each country, of supplying all current final energy consumption by solar systems. Again, this approach must not be seen as an extreme, e.g., the world primary energy demand is expected to increase by almost 40% by 2040 (WEO, 2014). Thus, the exercise performed will allow us to test MacKay’s affirmation that: “…in a world that is renewablepowered, the land area required to maintain today’s British energy consumption would have to be similar to the area of Britain. The same goes for Germany, Japan, the Republic of Korea, Belgium and the Netherlands” (MacKay, 2013). If these numbers were to be confirmed, far from enhancing their energy security as usually claimed, the transition to renewable energies in some countries in the current socio-economic context would instead increase their external dependence and vulnerability (Johansson, 2013; Lilliestam and Ellenbeck, 2011; Moriarty and Honnery, 2016; Trainer, 2013b). The paper is organized as follows: Section 2 includes the literature review related to the estimation of land requirements for solar technologies and describes the materials and methods used, Section 3 presents the results obtained and discusses them, Section 4 assesses the main assumptions and uncertainties considered in the analysis and Section 5 outlines our conclusions. 2. Materials and methods In order to assess the total land requirements of solar generation at the country level, we performed a literature review related to estimating land requirements for solar technologies which informed the choice of methods used in the analysis. These methods were implemented in the following steps:  Calculation of the electricity and final energy consumption by country for the year 2009 from a terrestrial-perspective (electricity/final energy use) and a consumption-based perspective (electricity/final energy footprint) (Section 2.1),  Estimation of a likely range for the solar power density by country considering future technological advances (Section 2.2),  Conservative estimation of the overcapacity needed by country to deal with the intermittence and seasonal variations in the solar resource (Section 2.3),  Estimation of the potential share of the electricity to be covered by rooftop PV on buildings by country (Section 2.4). Having estimated these factors, the land-use requirements per country to supply an amount of energy by solar power can be obtained by applying the following formula: 7 Equation (1) 2.1. Multi-regional input-output model Input-output tables display the interconnection between different sectors of production, making it possible to track the production and consumption in an economy. Traditionally, energy consumption has been described by the “energy use” indicator that refers to the amount of energy used within the borders of a country. However, in the last decade, the acceleration of economic processes linked to globalization (e.g., specialization and offshoring) has resulted in a shift of economic activities between countries and in a dramatic growth in international trade. Advanced capitalistic economies have specialized in economic activities with high added value, while reducing their share of energy intensive sectors and manufacturing industries (Baiocchi and Minx, 2010; Weber, 2009). In relation to this, MRIO tables allow us to track the global supply chains of products consumed by including the trade between different countries. In this paper, we combine the common “electricity use” (or territorial-based) indicator with the concept of an “electricity footprint” (or consumption-based) indicator which relates to the electricity consumed worldwide to produce the goods and services demanded by the people living in a given country. We apply the WIOD (Dietzenbacher et al., 2013), a set of MRIO tables that comprises information for 35 industries, 59 products for the 27 member states of the European Union (EU-27), and 13 non-EU countries (Australia, Brazil, Canada, China, India, Indonesia, Japan, South Korea, Mexico, Russia, Turkey, and the United States of America (USA)), as well as the Rest of the World (RoW) as an aggregated region. These 40 countries represent 65% of world's population and 90% of the GDP. Although the WIOD presents data from 1995 to 2009, here we concentrate on the last year of the series to perform a static analysis. The energy (electricity and final energy) use and footprint per country are obtained following the methodology described in Arto et al. (2016). Since the proposed analysis assumes that all the electricity production is substituted by solar power plants, we took the total final electricity consumption as well as the electric power transmission and distribution losses from the IEA Energy Balances for the year 2009 (IEA, 2016a). Consistency is ensured by considering own consumption by the solar power plants in the f2 factor (see Section 2.2). 2.1.1. Electricity and final energy consumption The electricity and final energy consumption (use and footprint) calculated for the target countries in 2009 is shown in Figure 1a and b (see Table A1). The countries are sorted in descending order according to the national means in terms of energy use. For countries with the highest electricity use per capita (Canada, Finland, Sweden and USA), the electricity use ranges from 13.5 to 16.5 MWh/person/year with differences between the electricity footprint and electricity use per capita of -13% (Sweden) to +9% (USA). At the other extreme, in the countries with the lowest electricity 8 consumption (India, Indonesia, Mexico and China), the electricity use ranges from 0.5 to 2.5 MWh/person/year, with differences between the electricity footprint and use use of -21% (China) to +15% (Indonesia). Most European countries are characterized by electricity footprints per capita higher than their electricity use, notably Greece (+36%), Denmark (+35%), Ireland (+29%) and the UK (+22%), whereas countries such as Russia (-24%), China (-21%) and South Korea (-16%) are net exporters of electricity embodied in trade (Figure 1a). In relation to per capita final energy consumption, similar trends can be observed (Figure 1b). The advanced capitalist economies consume over 20 MWh/person/year, while emerging and developing economies consume below the world average of 16 MWh/person/year. The per capita footprint of the advanced capitalist economies is almost +10% greater than their final energy use on average, with some countries showing notably greater differences such as France (+30%), the UK (+28%) and Italy (+25%). On the other hand, negative differences of -22%, -21%, and -14%, were found for Russia, Korea and China, respectively. These results are in accordance with the general assessment of energy footprint resulting from globalization and the increase in specialization and phenomena such as delocalizations, the advanced capitalist countries tending to be net importers of energy from the rest of the world (Arto et al., 2016). 9 a b 0 3 6 9 12 15 18 Canada Finland Sweden United States Luxembourg Australia Taiwan South Korea Japan Belgium France Austria Netherlands Germany Russia Estonia Slovenia Cyprus Czech Republic Ireland EU-27 Denmark Spain United Kingdom Greece Italy Malta Portugal Slovakia Bulgaria Hungary Poland Lithuania World China Turkey Brazil Romania Mexico Rest of the world India Indonesia MWh/year/person Per capita electricity consumption Electricity use Electricity footprint World mean 0 20 40 60 80 100 120 Luxembourg Canada Finland United States Denmark Belgium Netherlands Australia Sweden Taiwan South Korea Austria Russia Germany Ireland Japan Cyprus Czech Republic EU-27 France Malta United Kingdom Greece Estonia Slovenia Italy Spain Slovakia Portugal Hungary Poland Latvia Lithuania Bulgaria Mexico China Turkey Brazil Romania Rest of the world Indonesia India MWh/year/person Per capita final energy consumption Final energy use Final energy footprint World mean 16 Figure 3: Estimated average solar power density per country (We/m2/year) considering uncertainty in the efficiency of future PV modules and specific geographical characteristics. 2.2.5. Comparison with other values estimated in the literature Few studies have provided power density estimates of solar technologies that analyze real power plants. Smil (2015) reviewed the largest PV projects in operation, finding a range of 3-9 We/m2/year depending on the technology and geographical location of the site. De Castro et al. (2013) and Ong et al. (2013) highlighted the importance of assessing the entire land occupation of solar parks through the analysis of satellite images to identify plant configuration, direct land use and project area boundaries, since official project data are often unavailable or do not reflect the actual occupation of the infrastructure. Ong et al. (2013) analyzed 72% of installed and underconstruction utility-scale PV and CSP capacity in the USA, finding a generation-weighted average of the total land-use requirements 6 of 6.9 We/m2/year for small PV, 8.3 We/m2/year for large PV and 8.1 We/m2/year for CSP. These values are lower than those applied by Denholm and Margolis (2008), who took a PV ground-based array power density value for fixed panels which translated to a ~30% overestimation of the power density. Our estimated average power density for the USA is in the range of 3.6-6 We/m2/year, which is even lower than that found by Ong et al. (2013). This difference can be explained by several factors, in particular, the fact that they used a higher PR derived from solar manufacturers, and the fact that most of the current parks are installed in areas with very high irradiance levels (i.e., exceeding 250 We/m2/year, in California and Arizona; see also the CV obtained for the USA in Table 1). On 6 The total estimated area corresponds to all land enclosed by the site boundary, and the direct area comprises land directly occupied by solar arrays, access roads, substations, service buildings, and other infrastructure. 17 the other hand, only 15% of the projects analyzed by Ong et al. (2013) refer to completed projects; and hence, they relied to a large extent on manufacturers’ data, which have been shown to systematically overestimate the power density of the real parks (de Castro et al., 2013). 7 In terms of land use energy intensity (the inverse of ρe), our equivalent range is 9.4-99.8 m2yr/MWh. Specifically, the values for all countries, except for the Scandinavian and Baltic countries (Finland, Sweden, Estonia, and Latvia), lie within the range found in a literature review (Horner and Clark, 2013). However, this review only included one study at typical Scandinavian solar irradiance (see ref [5] in Horner and Clark (2013)) without taking into account the additional shadowing in these latitudes (f3=0.33). 2.3. Overcapacity and storage requirements due to short-term and seasonal variations For each location, the solar resource is variable over time, with both short-term variability (e.g., cloudiness, daynight) and seasonal variability (e.g., winter-summer), the latter completely uncorrelated with the demand. Usually, a grid can accommodate up to only 20% electricity from renewable sources without a need for dedicated storage facilities (Armaroli and Balzani, 2011; Lenzen, 2010). Thus, with the hypothesis that all the national electricity would be produced by solar power, a certain level of (1) storage, (2) overcapacity and (3) flexible demand should then be considered. In this work, we focus on the two first elements, distinguishing between short-term and seasonal variability. For the short-term variability, we focus on hydro pumping storage as proposed by other authors (Denholm and Margolis, 2008). Although electric batteries might also address the short-term variability, 8 hydroelectric pumping storage is currently the best solution due to its demonstrated functioning, competitive cost, high efficiency, long storage times (up to years) and fast response (Armaroli and Balzani, 2011). This solution would require the construction of a certain amount of additional capacity to compensate for the related losses, which for pumped storage are typically of the order of 25% (χ factor in Equation (8), i.e., a round-trip storage efficiency of 75%) (Denholm and Kulcinski, 2004; Denholm and Margolis, 2008; MacKay, 2013). These losses apply only to the fraction of demand passing through storage (fstor). The estimation of this fraction should ideally be done at the country level comparing hourly load to hourly PV supply (e.g., (Wagner, 2014)), which is far beyond the scope of this paper. Instead, as a reference value, we used the middle of the range (60-70%) found by Denholm and Margolis (2008) for a variety of regions in the USA. This approach based on hydro pumping is simplistic and conservative since it may be impossible to achieve the required storage volumes depending on the population density and the climate and geography of the country (Trainer, 2012). For example, MacKay (2013) estimated that summer/winter balancing for the UK would require lakes for pumped storage occupying 5% of the area of the country, which is physically unfeasible. Trainer (2013a) estimated for Europe that generation from pumped storage would have to be scaled up 7 The report also lacks of information to assess differences in the estimation of the f3 parameter. For example, the only table that would allow for a comparison (Table 5), reports only seven parks with a power density of 4-6.7 We/m2/year, which are all values lower than the reported average (8.3 We/m2/year). 8 In particular, electric cars may act as storage devices. The IEA (2016b) estimates that “125,000 cars could be equivalent to 300 MW of flexibility – a medium size pump storage plant or a successful stationary demand side response program”. 18 by a factor approaching 20. 9 CSP with storage could also help to mitigate the short-term variability, though unlike hydro it would not be a universal solution, since it requires high irradiance locations with low cloudiness to operate at profitable rates (typically desert areas). Additionally, CSP has a higher seasonal variability than PV. For instance, in Spain, the ratio of the highest/lowest monthly production is 9 for all CSP installations but 2.6 for the PV facilities (REE, 2016). In relation to seasonal variability, depending on latitude (e.g., winter-summer) and the regional climate characteristics (cloud cover, monsoon, etc.), there may be substantial differences in average monthly irradiance levels (Smil, 2015; Trainer, 2012). For example, although the minimum average monthly irradiance level represents over 90% of the annual average for cities such as Jakarta (Indonesia) or Rio de Janeiro (Brazil), in other cities such as London (UK), Paris (France) and Berlin (Germany), this ratio falls below 40% (NASA SSE, 2008). These differences are reflected in actual PV electricity generation: for example, in 2014, the electricity from solar in Germany was over 5 TWh in June and just 0.4 TWh in December, an order of magnitude difference. 10 This difference cannot be exclusively attributed to the difference in monthly irradiance since the minimum is only around 30% lower than the maximum (see Table B1), and is likely related to the increased shadowing in winter (especially critical on rooftops, see Section 2.4). Thus, previous studies considering only average annual irradiance levels without including the seasonal variability when estimating the solar potential of different countries and states (e.g., (Denholm and Margolis, 2008; Šúri et al., 2007)), underestimate the actual capacity (and land-use requirements) required to produce the electricity in months in which the irradiance is substantially lower than the annual average (Trainer, 2012, 2010). Apart from hydro pumping storage systems to compensate for the seasonal variations are not yet available and alternative technologies of large-scale storage are still in the R&D phase (Wagner, 2014). Thus, here we propose a novel approach to produce a conservative estimate of the additional capacity required to take into account seasonal variations at different geographical locations. We posit that, for each country, the total installed capacity should be able to cover the electricity consumption in the month with the lowest solar irradiance level. 11 This process consists of the following steps: a) Since within a given country (and especially those with large surface areas), there may be locations with very different solar irradiance potentials, we start by estimating a reference geographical coordinate center (the country-specific “central reference coordinate”, CRCi) with longitude CRClgi and latitude CRClati. b) CRClati was taken at approximately 12 +1/3 of the distance between the minimum and the maximum longitude relative to the country i, from its closest area to the equator. For example, the latitude in the case of Spain 9 However, the identified total technical potential for hydropower in Europe only doubles current installed capacity (IPCC, 2011). 10 http://www.solarwirtschaft.de/en/photovoltaic-market.html 11 In most countries, the winter consumption of electricity is higher than in summer (excepting those in tropical regions with a substantial use of air conditioning). Thus, the approach is internally consistent since by assigning a monthly annual average to winter, we are in fact underestimating the actual demand of electricity. 12 For some countries with very low irradiance and an elongated geographical shape such as Finland and Sweden the +1/3 criteria was softened to move the CRC to the south. 19 ranges between 36°N and 44°N and thus its CRClat would then be 36+(44-36)*1/3 ~ 39°N. The reason to take 1/3 instead of, for example, ½, was to consider that locations with better solar resources are more economically attractive and will tend to be occupied first (as is occurring for CSP plants in the USA (Ong et al., 2013), for example, although there are exceptions such as Australia where the majority of plants are close to the largest cities in the south of the country). c) CRClgi: starting from CRClati, we take the longitude values encompassed by the country. From this set of values, we select the longitude with the smallest difference between the value in the month with the lowest irradiance and the annual value, i.e., the most favorable longitude given the latitude. For example, Spain is characterized by the latitude 39°N and spans the longitude values 0, 1, 2, 3, 4, 5, 6 and 7°W (large variations can exist: for example, for Luxembourg there was only one longitude value while for Russia there were 90). The longitude level with the smallest difference between the value in the month with the lowest irradiance and the annual average was 0°W (thus, the CRC for Spain is 39°N 0°W). We call this ratio the seasonal variability (SV) and in the case of Spain it is estimated to be 0.7 (see Table B1). 13 Minor adjustments were made to ensure that the CRC annual average irradiance is greater than the country average (compare with Table 1). Hence, the overcapacity factor per country i to deal with short-term and seasonal variations can be expressed as (see the country values in Table B1): Equation (3) These overcapacity requirements can be as low as +30% (Australia) or 3 to 5-fold for those countries with a lower SV (typically northern European countries). In fact, as shown by Weitemeyer et al. (2015) with an hourly resolution study for Germany, while a 50-80% share of intermittent renewable sources may require relatively low levels of storage and overcapacity, a system 100% based on intermittent sources substantially increases these requirements, i.e., there is an asymptote when approaching the full intermittent energy mix. Our result for Germany (2.8-fold overcapacity) is in good agreement with their range. 14 Figure 4 illustrates the seasonal and geographical variations in the solar irradiance at the CRC for five representative countries. The CRC for the UK represents a typical northern European country characterized by a low irradiance and 13 The following case illustrates the conservative nature of the estimated SV in this analysis. In Spain, the PV electricity production in 2014 in two months (January and November) was less than 60% of the annual average of PV electricity production. In 2015, December was the worst case for PV generation. Considering other renewables in December 2015, wind electricity production was 88% of the annual average, hydroelectricity was 61% of its annual average and CSP only 20% of its annual average. Therefore, even an ideal renewable mix will likely require overcapacity/storage for some months, and our SV is likely optimistic (own calculations based on (REE, 2016)). 14 In fact, their study: (1) assumes no grid limitations, and (2) considers seasonal storage capacities and technologies (such as hydrogen) that are currently not commercially available on a gigantic scale. In their own words: “… the results derived from our approach for large-scale systems […] exhibit lower bounds for the actual storage demand” (Weitemeyer et al., 2015). 20 large seasonal variations (with winter values below 50% of the annual average, while reaching almost 150% in summer). The CRC for Greece is typical of that for Mediterranean countries, where significantly high irradiance values are reached during most of the year although still with a substantial seasonal variability. The Indonesian CRC represents a country with a high and stable solar irradiance over the year, while that for Australia shows an area with very high and stable solar irradiance. The influence of the monsoon is visible in the Indian CRC, provoking a decrease in the summer months from over 250 We/m2 to below 200 We/m2. 0 50 100 150 200 250 300 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec We/m²/year CRC latitude tilt radiation CRCC Germany Hungary India Turkey 0.25 0.5 0.75 1 1.25 1.5 CRC monthly/annual average ab Figure 4: Seasonal variations in solar irradiance (latitude tilt) at the estimated CRC for 5 countries: Australia, Greece, India, Indonesia and the UK. a) Monthly evolution (We/m2/year) and b) the ratio between each month and the annual average value (%). The SV is the key parameter to model the required overcapacity in order to deal with the seasonal variations in different geographical locations. It represents a rough estimate of the magnitude of the total solar PV capacity required to supply the electricity demand of the month with the lowest irradiance level in relation to the average annual level (see Equation (8)). Since neither daily peak demands nor variations within each month are taken into account (“good” sunny days vs. “bad” cloudy/rainy days), and the SV is the result of taking optimistic assumptions in relation to the CRC, the estimated overcapacity represents a lower bound (Trainer, 2013a). To finish this section, we remark that the estimated losses and overcapacity parameters are not independent of each other. For example, current parks are designed to optimize the yearly (average) output instead of maximizing the output for the period of the year with the least favorable climatic conditions (e.g., winter). In the second case, the distance between panels would then need to be increased, thereby reducing the actual f3 (that is, there is a trade-off between SV and f3). 21 2.4. PV potential on buildings and in urban areas Actual land requirements for solar plants are reduced by considering the potential of solar electricity to be produced on buildings and in urban areas. 15 Studies that evaluate this potential at a regional or country level are very common in the literature (e.g., (Bergamasco and Asinari, 2011; Byrne et al., 2015; Izquierdo et al., 2011; Jo and Otanicar, 2011; La Gennusa et al., 2011; Ordóñez et al., 2010; Paidipati et al., 2008; Wiginton et al., 2010)). Despite the potential being substantially reduced when considering shading, orientation, and other availability factors, it is generally found that rooftop PV could cover from a very low to moderate share of the electricity consumption. However, there is a substantial lack of standardization and no consensus method in the literature, with different methodologies achieving a different geographical coverage and different levels of spatial resolution (Melius et al., 2013). Hence, global estimates based on a consistent methodology across regions are scarce. For this reason, we have developed our own approach with the objective of making a rough assessment of the rooftop PV potential in each WIOD country. For this, we rely on GIS-based methods that represent a more objective and accurate approach for identifying rooftop availability than others based on constant values (Mainzer et al., 2014; Melius et al., 2013). It has been estimated that it would only be possible to cover a small percentage of today’s urban areas with solar panels (<2%), assuming acceptable efficiency (La Gennusa et al., 2011; Sorensen, 1999), since existing urban and architectural designs were not conceived to incorporate solar modules and are poorly compatible with them. Ordóñez et al. (2010) performed an extensive GIS-based analysis for all urbanized areas in Andalusia (Spain) taking into account the maximum occupation of roofs (9.4% of the urbanized area; Instituto de Estadísticas de Andalucía (2015)). Without taking into account non-usable buildings, e.g., those under heritage protection, or shadows between buildings (personal communication), they found a potential of 3% of surface area covered in relation to the urbanized land in that region. Thus, in current conditions, a plausible maximum range dedicated to PV systems would be in the order of 2-3% of urban areas. However, in practice, there are other uses for rooftops: daylighting, solar thermal, roof-top gardens or terraces, etc. Although some uses might be compatible with rooftop PV (and sometimes even complementary, e.g., green roofs, hybrid solar collectors, etc.), others will compete for the available roof space, some of these uses already being promoted as sustainable/green practices. For example, solar thermal is a promoted and competitive technology already occupying many suitable locations (Cansino et al., 2011; REN21, 2015) (including in high latitude regions (Hagos et al., 2014)), and needs to be close to consumers due to the technical difficulty of transporting heat over large distances without incurring in high losses, unlike electricity (IEA, 2006). Globally, solar thermal already accounts for about 1.2% of water and space heating in buildings (REN21, 2015). For example, a GIS study for Spain found that after satisfying up to 70% of the service hot water demand in every municipality, around 80% of the suitable roof area of the country was identified as available for rooftop PV (Izquierdo et al., 2011). 15 Other studies such as that of Šúri et al. (2007) did not consider the land-use requirements for solar plants by assuming a priori that all the PV power would be roof-mounted panels. 22 Thus, studies that do not take into account these competitive uses are likely overestimating the actual surface area available for rooftop PV. Hence, we derive the net power density for rooftops in each country (ρe,rti) from Equation (2) but with the land-occupation ratio (f3,rt) corresponding to the range 1-2%, 16 and with f2,rt less than the performance ratio in Equation (2) taking into consideration the lower relative efficiency of rooftop PV systems over ground-mounted systems (see Equation (4)). Although by deploying PV systems on buildings the system is situated at a potential point of use (eventually minimizing transmission and distribution requirements and losses), rooftops are less efficient than ground-mounted systems since (1) the orientation and tilt of the roof is given and will ordinarily be suboptimal, and (2) there is a correlation between the size of the plant and its capacity factor (Cp). Equation (4) How efficient is rooftop PV in relation to ground-mounted installations? To answer this question, we compare countries with a high proportion of rooftop systems like Italy and Germany (Smil, 2015) with countries with almost all ground-mounted systems like Spain. For the case of Italy, we could calculate the Cp as a function of the power of the installation (GSE, 2015) (see Table 5). The Cp for <20 KW plants (mainly rooftop) is around 80% of the Cp for >1,000 KW plants (mainly ground mounted). Therefore, for the Italian case, the f2,rt/f2 ratio would be around 0.80 under present conditions when the rooftop PV installations cover less than 1% of the built-up surface area of the country. Germany has an average irradiance of 66% of that of Spain, but only a Cp of 53% of that of Spain (Table 1, (Wirth, 2015)), and therefore the efficiency of the solar system of Germany is 0.53/0.66 ≈ 0.80 of that in Spain and this difference could be attributed to the poorer performance of rooftop systems. The case study for Andalusia, carried out by Ordóñez et al. (2010), a region with a very good average irradiance (~ 200 We/m2) and considering the maximum occupation of roofs and module panels with f1 = 20%, found a Cp of 0.096. If we compare this number with the present Cp for Spain as a whole of 0.197 (with less average irradiance than Andalusia and with more than 97% of PV ground–mounted systems (Prieto and Hall, 2013)), then the performance ratio of rooftop/ground mounted would be f2,rt/f2 < 0.5. The study of Ordóñez et al. (2010) maximizes the power produced and not the efficiency of the system, and therefore in current roof systems far from the maximum occupation potential, the ratio f2,rt/f2 is higher. Capacity of installed plants Cp (%) < 20 KW 12.2 < 200 KW 12.7 < 1,000 KW 14.0 > 1,000 KW 15.1 Table 5: Solar PV capacity factor depending on the capacity of the installed plant for the year 2013 in Italy (GSE, 2015). Hence, Equation (4) can be rewritten as: 16 Although in principle this ratio might improve in the future if urban norms were focused on maximizing PV rooftop output, its influence would be substantially reduced by the fact that buildings have a very long lifetime and in many countries the stock of buildings will not increase much in the future (especially in more industrialized countries) due to projected population stagnation. 23 Equation (5) The urbanized area for each country (BUi) was approximated by the category “Built-Up” from the Global Agroecological Zones Data Portal version 3.0 that identifies the estimated share of land cover/land use required for infrastructure and settlement (FAO/IIASA, 2011). We consider this database since it is the most up-to-date synthesis of global information sources. Since the urban area is included in the built-up area, by applying the factor of land occupation to the built-up area, we are in fact providing an upper boundary for the actual potential of rooftop PV. 17 2.5. Land-use requirements for solar power: summary Calling Erti the potential electricity output from rooftop PV and Etoti the total consumption of electricity of that country for a given year (here 2009), rti represents the share of the electricity covered by rooftop for each country in relation to the total consumption (the values for each country are shown in Appendix C). We subtracted the current production from hydropower from the electricity consumption (see Section 2.1.1) for the same year for each country in order to represent the results in terms of land use excluding water bodies (see Figure D1) (US EIA db, 2015). In any case, since the power density of hydropower is similar to solar (0.5-7 We/m2 (Smil, 2015)), the conclusions in terms of land use would not vary significantly under the assumption that solar would also produce the electricity currently supplied by hydro. However, we judge the first option to be more realistic since the dams and related existing hydro infrastructure have lifetimes of over 100 years, are consistent with a 100% renewable scenario and we are assuming the operation of hydro pumping storage. 18 Equation (6) Equation (7) Thus, with representing the energy density for country i, χ the storage losses in pumping hydro to compensate for the short-term variability and SVi the overcapacity required to address the seasonal variation in each country, the land-use requirements for solar power in each country (LUi) can be expressed as: Equation (8) 17 With this hypothesis, we expect to compensate for developments not taken into account such as PV potential on facades. 18 However, with this approach, we are not considering inter-annual rain variability and hence not addressing the question of, for example, the electricity supply in a dry year during a cloudy winter. 24 3. Results and discussion 3.1. 100% Solar electricity mix scenario Table E1 shows the total land surface area occupied by solar facilities to cover the electricity use and footprint per country (except for current hydro generation). As expected, the countries with the largest populations and highest levels of electricity consumption per capita lead the ranking: EU-27 (220 – 380 · 103 km2), the USA (95 – 170 · 103 km2), Russia (65 – 115 · 103 km2) and China (45 – 100 · 103 km2), and those with the smallest size and populations are at the bottom: Malta (50 – 85 km2), Cyprus (110 – 200 km2) and Luxembourg (490 – 830 km2). 19 Hence, to comprehend the implications in terms of land use for each country, a relative perspective must be taken. Figure 5a depicts the land associated with the solar electricity footprint and use for each WIOD country as a share of the total land. To facilitate the interpretation of these results, different land uses are depicted from GAEZ (FAO/IIASA, 2011). Specifically, the countries that would need to occupy, proportionally, a larger area to cover their current electricity consumption with solar would be (in decreasing order): the Netherlands, Malta, Belgium, the UK, Luxembourg, South Korea, Germany, Finland, Taiwan, Denmark and Japan. The absolute land cover share for these eleven countries ranges from 50-60% (the Netherlands) to 10-11% (Japan), these countries in most cases requiring a surface area similar to or larger than that of the land currently cultivated (range: electricity use – electricity footprint). Another useful indicator to comprehend the scale of these land requirements is to compare these land requirements with the land currently dedicated to infrastructure and settlement (built-up): for most of the advanced capitalist economies, the area required would be of the same or a higher order of magnitude (see Table E1). 19 Countries with a high share of hydroelectricity such as Slovenia and Brazil also have low land requirements due to the methodology applied (see section 2.5). 25 a bMean > 100% 0 10 20 30 40 50 60 70 80 90 100 Netherlands Malta Belgium United Kingdom Luxembourg South Korea Germany Finland Taiwan Denmark Japan EU-27 Ireland Czech Republic Sweden Estonia Poland Italy France Slovakia Hungary Slovenia Greece Lithuania Portugal Spain Cyprus Austria United States Bulgaria Latvia China Turkey Romania Russia India Canada Mexico Indonesia Australia Brazil Land cover (%) Cultivated land Grassland&Woodland Barren-sparsely vegetated land Forest land Built-up Solar electricity use (mean) Solar electricity footprint (mean) Figure 5: (a) Land associated with the solar electricity footprint and use (mean values) for each WIOD country as a share of the total land. Different land uses are depicted: cultivated, grassland and woodland and barren/sparsely vegetated, built-up and forest. The countries are ranked from the highest to the lowest share of total land required to cover the current electricity use. (b) Land occupation by solar power plants to cover the electricity use 32 Trainer, 2013a). Although several projects have been proposed, none is currently under way (Breyer et al., 2015; DESERTEC, 2003; Gulagi et al., 2017). For example, the DESERTEC project aiming to harness solar energy from deserts, producing large amounts of electric power by CSP plants based in Northern Africa and the Arabian Peninsula, and transmit it through high-voltage direct current (HVDC) transmission lines to Europe (DESERTEC, 2003). Its connection with the electricity generated by PV, wind, hydro, geothermal and biomass in European countries would lead to an integrated regional framework (Armaroli and Balzani, 2011). However, these kinds of projects are very sensitive to geopolitical circumstances and depend on gigantic investments. In fact, the DESERTEC consortium dissolved some years ago and there are no imminent prospects for its reactivation (Smil, 2015). On the other hand, although significant new construction of HVDC lines is currently under way (particularly in China and Brazil), very little of the new capacity actually crosses country borders except in the EU. The feasibility of large intercontinental grids operating within the next few decades critically depends on future global societal pathway uncertainty. Under pathways characterized by an intensification of globalization and market integration processes (including international trade growth), these forecasts may prove valid. However, under scenarios dominated by fragmentation and regionalization of the relationships between countries, the functioning of these systems would be unrealistic (MEA, 2005; Raskin et al., 2002). In fact, these regional and intercontinental interconnection processes require that the participants collaborate to share and commonly manage their resources. Alternatively, this could also be achieved through force, with the strongest countries seizing locations with abundant renewable resources from countries not willing to collaborate. On the other hand, large intercontinental grids of thousands of kilometers and large areas dedicated to RES generation systems in foreign territories might also be an “easy” target to paralyze the economic activity of the importing regions. Moreover, given the lower power densities of RES in relation to fossil fuels, the surface required for RES generation systems is much higher, and hence, the geographical area to be secured by importing regions substantially increases, thereby substantially increasing their vulnerability (Lilliestam and Ellenbeck, 2011) A renewable mix portfolio would mitigate the variability of the solar resource, eventually reducing the need for storage and overcapacity considered in this analysis. For example, in Europe, the annual cycles of wind and PV are partially complementary since the lower solar irradiance in winter is generally balanced by increased wind (and vice versa in the summer). There is also generally a trade-off between the installation of additional generation capacities and storage capacities to balance the intermittence of resources that is not captured by our approach (Armaroli and Balzani, 2011; François et al., 2016; Wagner, 2014; Weitemeyer et al., 2015). However, this complementarity is far from perfect. In any region there is a (low) probability of extreme combinations in the availability of natural resources, such as no wind over large parts of Europe during the winter. Moreover, there can be extreme annual variations in the availability of natural resources; for instance, the output of wind turbines in any given area can vary by up to 30% from one year to the next (Brower et al., 2013; Li et al., 2010). With hydroelectricity covering currently a mere 15% of the current global electricity consumption and already being close to its technical potential in many 33 countries (IPCC, 2011), only wind could cover a substantial share of the remainder in a sustainable way. Moreover, wind generation could moderate the impact in terms of land use since, despite its lower power density, both offshore deployment and double land use are possible. On the other hand, technical (Lenzen, 2010) and biophysical limitations (de Castro et al., 2011; Miller et al., 2011) hinder large-scale deployment of wind energy. We have chosen to give priority to overcapacity (following a conservative approach, see Section 2.3) rather than storage, since economic options for large-scale and seasonal storage are currently neither technically available nor foreseen at the required scale. Moreover, storage losses would be higher than considered here, as hydro pumping cannot be generalized (MacKay, 2013; Trainer, 2012). Other sources of variability such as low rain years have not been considered; and, by taking the monthly irradiance average, we are not accounting for short-term solar variability (i.e., over hours, days, weeks). Finally, we have not allowed for the fact that demand peaks at certain times of day at levels much higher than the average (Trainer, 2013a), conservative estimates of these peaks being +30%, while other studies have yielded estimates several times higher. Coping with these factors would require greater levels of storage and/or overcapacity (higher SV). Moreover, the system should be designed for the minimum of all the energies in the mix, that might well be close to the minimum in a full 100% solar system (Trainer, 2012). 4.2. Static vs. dynamic projections The proposed analysis follows a hybrid approach by combining static data (i.e., electricity consumption values for 2009) with dynamic factors (e.g., the uncertainty range for the future evolution of cell efficiency). The objective of this approach is double: on the one hand, to replicate the methodology followed in previous studies to allow comparability (Denholm and Margolis, 2008; Šúri et al., 2007), and on the other, to perform a vulnerability study in order to identify countries for which the objective of high solar power deployment might be unfeasible due to land constraints. This approach does not fully correspond with an “extreme scenario” when considering the expected increases in electricity and energy production globally for the coming decades (Armaroli and Balzani, 2011; WEO, 2014). In particular, emerging economies are expected to substantially increase their income and population levels in the coming decades. Let us take the case of India. Our estimated land requirements in terms of total land use seem relatively low for both covering electricity (0.4% ± 0.2) and final energy (3.8% ± 1.1) consumption. However, its population density is very high (over 400 people per km2 of terrestrial land), a large share of their land being required to produce feedstock (almost 60%, see Figure 5a) or built up (~7%). Hence, in terms of land availability the situation worsens substantially, especially in terms of final energy (26.1% ± 7.5). Population growth (projected to be +30% to 2050 (UN, 2015)) combined with the expected increase in energy consumption (currently around a third of the global average, see Figure 1) are likely to drive land requirements to unfeasible levels. Similar reasoning could be applied to other emerging countries such as China and Nigeria (Deng et al., 2015). In relation to the estimated “land availability”, it could be argued that improvements in agricultural yield might to some extent compensate for increases in food demand due to population growth and diet shift in the future. 34 However, after decades of improvement due to agricultural intensification (based on inputs such as fertilizers, energy, water and capital equipment), yields might be approaching biophysical limits in some developed countries, and could stagnate for socio-economic reasons in developing countries (Alston et al., 2009; Grassini et al., 2013; Sands et al., 2014). 4.3. Estimation of solar power density at the country level Various factors affect the estimation of the solar density for each country (see Section 2.2.1). The consideration of country average irradiance levels might be seen as a conservative assumption since areas with more irradiance are likely to be occupied more intensively. On the other hand, there are other important variables that influence the location of new power plants such as proximity to the consumer demand and existing electricity transport infrastructure. This is the case of, for example, Australia and Germany, where the locations used are mainly suboptimal in terms of irradiance. However, we tried to compensate for this effect by selecting favorable parameters for the overcapacity factors (e.g., SV, CRC, etc.). In some cases, these parameters are even unrealistic since they correspond to locations of high mountains (i.e., less cloud cover, e.g., China CRC) or protected areas. As stated above, large solar parks usually have higher power density levels than smaller ones. Hence, factors such as existing promotion policies and development frameworks will determine the number of each to be installed (i.e., planned economy, investment of large companies or small cooperatives or producers, self-production, etc.). Moreover, if land competition is to play a relevant role in future solar deployment, the factor f3 might even decrease due to the need to adapt to the accessible fields. 4.4. Not accounting for EROI The EROI is defined as the ratio of the amount of usable energy acquired from a particular energy resource (Eout) to the amount of energy spent obtaining that energy resource (Ein). Thus, accounting for the EROI of the solar PV systems would increase the required installed capacity to deliver the same output of net energy to the society. Defining r as the relative losses (Ein/Eout, see eq. 10), the total capacity to install in order to compensate for the energy invested would be given by the series 1+r2+r3+…, which is the Taylor series of 1/(1-r) (see eq. 11). Equation (9) Equation (10) Equation (101) 35 An examination of the EROI literature on solar PV energy generation shows differences in the assumptions, parameters and methodologies employed (Hall et al., 2014). Assuming a conventional value of 10:1 and applying the Eq. 10, an overcapacity of 10/(10-1)-1≈11% would be then required in an hypothetical 100% solar energy system. The extended EROI (EROIext) broadens the conventional analysis by considering all the energy used to create and run a PV facility, including the fabrication of the PV modules (more than 95% of the power used for the manufacture of solar PV cells being electricity (Briner, 2009)), but also for the energy required for transportation, installation, maintenance, and all the other required energy inputs. The estimation of the EROIext of PV systems range 2 to 3:1 (Hall et al., 2014), which would translate into overcapacities of 50% to 100%. Hence, accounting for the EROI factor would substantially increase the land requirements obtained in this analysis. 4.5. PV potential in buildings As pointed out in Section 2.4, there is substantial heterogeneity in the literature in relation to the methods applied (achieving a different geographical coverage and levels of spatial resolution) and the results obtained. Our study incorporates assumptions about storage, seasonal variability and competing uses that are rarely taken into account in research in this field. It is therefore challenging to compare our results (see Appendix C) with those of other studies. The estimated PV rooftop potential seems to be in accordance with the GIS-based studies considered (e.g., for Spain, Izquierdo et al. (2011) having found that rooftop PV could cover around 4% of the total electricity demand, which is within our range, 2.2-7.4%, including hydro), while our figures are substantially lower than those obtained by constant-value methods (e.g., Defaix et al. (2012) and Paidipati et al. (2008) having reported potential shares of over 20% of the annual electricity demand for the EU-27 and USA). Further research is required to characterize the variability in these methods to reduce the uncertainties in current estimates, as has already been done, for example, for PV and CSP technologies (Horner and Clark, 2013). Further, the method applied does not escape the fact that there is a large uncertainty in the estimation of urban/built-up areas, with differences of an order of magnitude (Schneider et al., 2009). In relation to this, the GAEZ database (FAO/IIASA, 2011) used is the most recent attempt to provide a consistent estimate across different countries, and the global built-up area is in the upper range of the literature (around 1.5 million km2) (Schneider et al., 2009). 5. Conclusions In this work, we have analyzed the biophysical feasibility and potential vulnerabilities in the transition to renewable energies focusing on the land requirements for 100% solar energy scenarios for 40 countries considering two issues that are not usually considered in the literature: (1) the need to cope with the variability of the solar resource, and (2) the real land occupation of solar technologies. The exercise performed shows that, for many advanced capitalist economies, the land requirements over the total terrestrial surface area to supply current electricity consumption would be substantial, the situation being especially challenging for those located in northern latitudes with high population densities and high electricity consumption per capita such as the Netherlands, Belgium, the UK, Luxembourg, South Korea, Germany, Finland, Taiwan, Denmark and Japan (10-50%). Moreover, accounting for the 36 electricity footprint, i.e., for the net energy embodied in international trade, tends to worsen the situation of these countries, increasing the land requirements to 11-60%. To assess the implications in terms of land competition, we have defined a land availability factor based on current land use and including a biodiversity buffer. With this indicator, the list of vulnerable countries enlarges substantially (e.g., the EU-27 would require around 50% of their available land), while few advanced capitalist economies would require low shares of the estimated available land (e.g., Canada and Australia, < 1%). Specifically, the consideration of circumstances not usually taken into account in the literature such as the need for redundant capacity to ensure sufficient winter supply and the real land-occupation factor of solar power plants produces results more challenging in terms of land requirements than previously found (Denholm and Margolis, 2008; Jacobson and Delucchi, 2011; Šúri et al., 2007). In particular, assertions about the practical unlimited nature of the solar potential should be reconsidered (e.g., (IPCC, 2011; Rogner et al., 2012)). In fact, our results help to explain why, when solar is scaled up, parks are often being located in agricultural areas and biodiversity hotspots, namely, that this is a palpable example of the finiteness of the biosphere, i.e., of the “full world” to use the popular expression coined by Daly (2005). The explorative exercise to examine the implications of a 100% solar energy mix for the same set of countries reveals the marked implications of the transition from concentrated fossil energies (i.e., “mines”) to distributed (over the Earth’s surface) RES. The results, that must be interpreted as order of magnitude estimates, due to the limitations and uncertainties of the analysis, indicate that the transition to domestically produced RES maintaining the current levels of energy consumption could be physically unfeasible for many countries: in particular, the Netherlands, Luxembourg, Belgium, the UK, Denmark, Germany, South Korea, Taiwan, Finland, Japan, Ireland, Czech Republic, Sweden, Poland, Estonia and Italy would require over 30% of their total land area (over 50% for the whole EU-27). Due to the biophysical restrictions on solar, these countries will have to rely on other domestic renewable sources (mainly wind and eventually biomass) or imports in their path to a 100% RES system. Although concerns over the potential land scarcity provoked by a full transition to RES have been already pointed out from a theoretical point of view (Johansson, 2013; Rao and Sastri, 1987; Scheidel and Sorman, 2012), currently available energy models are unable to properly capture the trade-offs between energy and food security, as well as biodiversity conservation, in the context of 100% RES scenarios. We attempt to contribute to the discussion by a rough quantification of potential implications. Specifically, by providing a ranking, we identify countries with particularly high vulnerabilities. In this context, emerging countries might seem less vulnerable. However, if the projections of electricity and energy consumption growth, on the one hand, and the population increase, on the other, prove to be correct, such countries may suffer similar land pressures in the future to those currently on the advanced capitalist economies. Indeed, the pressures might be even greater, if the current structure of international trade, where they export substantial amounts of embodied energy, is not modified. 37 Thus, the transition to RES maintaining the current levels of energy consumption might create new vulnerabilities and/or reinforce existing ones in terms of biodiversity conservation, energy (and food) security and sovereignty, with the potential to intensify imperialist geopolitics to grab land and seize resources from other countries. It seems likely that, without profound changes in the level and management of energy demand (adaptation to the natural fluctuations in the availability of renewables, consumption reduction and equity at the global level) and the socioeconomic system (i.e., the growth paradigm), the transition to renewables will substantially increase the competition for land globally. Finally, we recall that the analysis performed must be understood as a vulnerability study of the unavoidable path to a 100% RES system for all countries. Future work could be focused on (1) dynamic developments, (2) expanding the potential constraints that may further reduce PV feasibility (e.g., material availability (García-Olivares et al., 2012; Lo Piano and Mayumi, 2016)), (3) analyzing the feasibility and land requirements of different 100% RES mixes, which has to be approached at a country level and with a different methodology, 21 and (4) improving the methods to represent the land availability and competition between different uses in existing energy models. In particular, GIS- based methods or integrated assessment models that include the interaction between the energy system and landuse changes could substantially contribute to advance in these directions. Acknowledgements The authors gratefully acknowledge Ignacio Cazcarro for performing the calculations with GIS. This work has been partially developed under the MEDEAS project, funded by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 691287. This work was partially developed within the ESPON Study on “Possible European Territorial Futures” (Contract No EE/SO1/007/2016). Appendix A: Electricity and final energy consumption by country for 2009 Electricity use Electricity footprint Final energy use Final electricity footprint MWh/year/person MWh/year/person MWh/year/person MWh/year/person Australia 11.9 12.5 51.0 59.1 Austria 7.6 9.0 41.3 50.7 Belgium 8.2 8.9 52.0 48.7 Brazil 2.6 2.6 13.5 13.6 Bulgaria 4.9 4.3 15.1 15.2 Canada 16.9 16.3 79.2 75.6 China 2.8 2.2 14.6 12.5 Cyprus 6.6 9.1 32.7 42.9 Czech Republic 6.5 6.1 31.8 30.0 Denmark 6.2 8.4 53.7 47.8 21 The methodology applied here cannot straightforwardly be extended for wind since bottom-up analyses violate the Principle of Conservation of Energy (de Castro et al., 2011; Miller et al., 2011). 38 Estonia 6.6 6.0 28.8 29.0 Finland 15.8 14.2 62.4 59.9 France 7.9 9.1 31.0 40.2 Germany 7.0 7.7 37.6 44.0 Greece 5.8 7.9 28.9 35.1 Hungary 4.1 4.5 22.4 24.2 India 0.8 0.8 4.6 4.8 Indonesia 0.7 0.8 7.6 7.7 Ireland 6.4 8.3 36.4 46.3 Italy 5.6 6.6 27.3 35.0 Japan 8.2 8.9 35.0 38.8 South Korea 9.4 7.9 43.7 34.5 Latvia 3.2 4.0 20.9 25.5 Lithuania 3.5 4.2 20.4 23.9 Luxembourg 13.3 13.0 121.9 91.6 Malta 5.2 6.3 30.4 24.3 Mexico 2.3 2.5 14.7 15.8 Netherlands 7.0 8.0 51.8 44.7 Poland 3.9 4.0 21.6 21.7 Portugal 5.1 5.8 23.5 26.3 Romania 2.5 2.7 13.4 15.4 Russia 6.7 5.1 41.0 32.1 Slovakia 5.0 5.2 26.5 27.4 Slovenia 6.6 6.8 28.3 35.4 Spain 6.2 6.9 26.6 32.6 Sweden 15.2 13.3 47.2 47.8 Taiwan 9.7 6.6 44.3 27.1 Turkey 2.7 2.8 13.7 15.7 United Kingdom 6.1 7.4 30.1 37.6 USA 13.7 14.9 60.1 66.7 Rest of the world 1.4 1.5 10.1 9.9 WORLD 2.9 2.9 15.9 15.9 EU-27 6.4 7.1 31.2 36.0 BRIIC 2.0 1.7 11.3 10.1 EAS 8.7 8.4 38.2 36.4 Developed 9.4 10.0 42.8 46.7 Table A1: Electricity and final energy consumption (use and footprint) by country for the year 2009. Appendix B: CRC and SV values per country Lat Lon Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann Min/Ann (SV) fovercapacity CRC (º) (º) We/m²/year % % Australia -20 133 282 251 262 267 255 239 257 274 275 268 265 286 265 90% 129% Austria 47 14 104 142 166 176 187 178 184 181 155 122 93 81 148 55% 212% Belgium 50 5 59 92 125 157 183 180 182 178 140 103 66 48 126 38% 308% Brazil -10 -41 240 248 230 230 220 220 228 255 271 265 253 246 243 91% 128% Bulgaria 42 23 130 148 173 178 195 217 230 227 201 160 117 108 174 62% 188% Canada 52 -74 97 155 205 239 205 193 175 166 126 102 93 81 153 53% 219% 39 China 30 96 233 242 225 220 228 220 214 213 217 219 229 233 224 95% 123% Cyprus 35 27 163 193 244 268 288 303 309 313 300 258 185 151 248 61% 191% Czech Republic 49 18 75 115 136 158 180 171 178 179 141 100 67 62 130 48% 243% Denmark 55 9 51 91 129 175 205 193 193 186 141 92 58 48 130 37% 315% Estonia 58 25 51 93 147 177 210 206 201 183 141 88 55 36 133 27% 430% Finland 61 25 37 88 147 181 207 198 197 172 136 78 49 22 126 17% 675% France 45 6 120 150 184 187 203 218 232 221 198 154 117 98 174 57% 205% Germany 50 12 66 105 126 157 178 170 177 177 133 100 61 52 125 42% 279% Greece 38 21 143 155 194 209 230 262 264 257 229 188 130 114 198 58% 202% Hungary 47 17 88 131 154 180 199 193 197 197 161 119 83 67 148 45% 256% India 17 76 255 272 274 258 261 216 188 178 205 226 241 242 235 76% 153% Indonesia -8 112 198 198 203 210 214 208 220 238 252 244 220 208 218 91% 128% Ireland 52 -7 47 77 105 148 168 163 162 152 128 89 60 42 112 37% 312% Italy 41 14 121 144 178 188 210 232 247 240 204 164 114 102 179 57% 204% Japan 35 138 173 193 203 213 197 173 178 194 160 177 171 173 184 87% 133% South Korea 36 129 185 194 205 225 215 191 163 167 167 187 170 175 187 87% 134% Latvia 56 27 58 103 150 173 200 192 188 182 134 91 58 46 131 35% 330% Lithuania 55 26 65 106 148 173 201 191 190 185 139 94 60 55 134 41% 283% Luxembourg 50 6 60 98 123 160 180 176 181 176 137 96 63 49 125 39% 296% Malta 36 13 164 200 248 265 277 285 296 293 265 231 179 151 238 64% 183% Mexico 20 -101 248 282 300 284 261 265 253 238 231 243 257 230 258 89% 130% Netherlands 52 6 54 93 119 155 180 171 171 169 133 88 58 43 120 36% 321% Poland 51 24 71 111 145 169 193 181 180 185 138 104 66 59 134 44% 263% Portugal 38 -8 162 183 223 231 237 265 277 275 241 188 155 138 215 64% 181% Romania 45 24 130 150 173 166 181 184 193 195 170 144 114 100 158 63% 184% Russia 55 112 95 143 208 234 219 213 202 189 165 143 100 78 166 47% 247% Slovakia 48 20 83 124 151 169 183 183 185 188 155 115 78 66 140 47% 247% Slovenia 46 14 109 150 178 184 197 193 200 196 167 123 96 82 156 52% 222% Spain 39 0 187 208 246 266 260 275 285 278 254 219 178 164 235 70% 167% Sweden 58 15 46 88 141 175 200 198 195 180 147 88 52 34 129 26% 443% Taiwan 23 120 153 150 172 187 200 228 228 214 207 205 177 156 190 79% 147% Turkey 38 37 155 180 206 213 239 264 279 274 259 200 160 137 214 64% 182% United Kingdom 53 -1 47 84 110 144 168 162 165 156 125 92 57 40 113 36% 324% USA 35 -107 208 224 260 274 272 261 239 232 244 242 214 201 239 84% 138% Table B1: Latitude tilt radiation values (We/m²/year) for each country central reference coordinate (CRC). The last column shows the ratio between the value for the month with the lowest irradiance and the annual average, i.e., the seasonal variability (SV) factor. Minor adjustments due to cloudiness influence for Brazil and Indonesia were made to ensure that the CRC annual average irradiance is higher than the country average: we selected a different longitude associated with the CRClat with a SV difference lower than 3% compared to the most favorable longitude. fovercapacity refers to the overcapacity factor considered in order to deal with the intermittence of the solar resource and storage (see Section 2.3). Appendix C: Potential electricity produced by rooftop PV Ert_min Ert_max %total Australia 5.2% 17.4% Austria 4.2% 13.9% Belgium 1.2% 3.9% Brazil 45.0% 150.2% Bulgaria 4.4% 14.7% Canada 2.2% 7.3% China 8.6% 28.7% 40 Cyprus 3.1% 10.4% Czech Republic 1.6% 5.5% Denmark 1.6% 5.5% Estonia 1.2% 4.1% EU-27 1.9% 6.3% Finland 0.3% 1.0% France 1.9% 6.2% Germany 1.4% 4.5% Greece 2.1% 7.1% Hungary 3.7% 12.3% India 20.6% 68.5% Indonesia 27.8% 92.7% Ireland 1.1% 3.6% Italy 2.5% 8.2% Japan 1.4% 4.7% South Korea 0.9% 3.1% Latvia 5.7% 19.1% Lithuania 3.6% 12.1% Luxembourg 0.6% 1.9% Malta 2.5% 8.4% Mexico 8.9% 29.6% Netherlands 1.0% 3.2% Poland 2.4% 8.0% Portugal 3.6% 12.0% Romania 9.8% 32.5% Russia 1.9% 6.3% Slovakia 2.4% 7.9% Slovenia 2.7% 9.1% Spain 2.4% 8.1% Sweden 1.0% 3.4% Taiwan 2.5% 8.4% Turkey 5.9% 19.7% United Kingdom 1.0% 3.4% USA 2.6% 8.8% Table C1: Rooftop PV production per country as a share of the electricity use minus the hydroelectricity production. 41 Appendix D: Share of electricity consumption covered by hydro generation 0% 20% 40% 60% 80% 100% Brazil Canada Austria Latvia Sweden Slovenia Romania Turkey Russia China Slovakia Portugal Finland Italy India France Mexico Bulgaria Spain Greece Japan Indonesia United States Australia Lithuania Czech Republic Germany Ireland Taiwan Luxembourg Poland United Kingdom Korea, South Hungary Belgium Estonia Netherlands Denmark Malta Cyprus Share of electricity consumption covered by hydro generation (2009) Figure D1: Share of electricity consumption covered by hydro generation by country in 2009 (US EIA db, 2015). Appendix E: Total land and land per capita by scenario Land area (km²) Land per capita (m²/person) Elec-use Elec-F Total land Built-up area Elec-use Elec-F Total land Built-up area Australia 3,638 ± 1,140 3,837 ± 1,202 7,609,834 11,299 168 ± 53 177 ± 55 354,159 521 Austria 1,402 ± 420 2,078 ± 622 82,161 2,268 168 ± 50 249 ± 75 9,877 272 Belgium 9,054 ± 2,382 9,835 ± 2,587 30,061 3,879 839 ± 221 911 ± 240 2,805 359 Brazil 700 ± 700 705 ± 705 8,222,763 49,313 4 ± 4 4 ± 4 43,197 255 Bulgaria 1,531 ± 463 1,326 ± 401 107,975 2,635 206 ± 62 178 ± 54 14,583 354 Canada 19,998 ± 5,497 18,052 ± 4,962 8,032,228 12,221 595 ± 163 537 ± 148 270,410 363 China 72,209 ± 26,146 54,287 ± 19,656 9,338,610 265,005 54 ± 20 41 ± 15 7,080 199 Cyprus 155 ± 44 215 ± 62 8,318 184 142 ± 41 197 ± 57 8,473 169 Czech Republic 5,241 ± 1,407 4,963 ± 1,332 77,223 3,166 502 ± 135 475 ± 128 7,397 303 Denmark 4,698 ± 1,261 6,324 ± 1,698 37,223 2,128 851 ± 228 1,145 ± 307 7,682 385 Estonia 2,575 ± 679 2,327 ± 613 38,591 573 1,929 ± 509 1,743 ± 460 31,764 430 EU-27 302,221 ± 80,518 327,999 ± 87,547 4,039,667 141,887 605 ± 161 656 ± 175 8,373 284 48 Global Energy Assessment - Toward a Sustainable Future. Cambridge University Press, Cambridge, UK and New York, NY, USA and the International Institute for Applied Systems Analysis, Laxenburg, Austria, pp. 423– 512. Sands, R.D., Jones, C.A., Marshall, E., others, 2014. Global drivers of agricultural demand and supply. Econ. Res. Rep. 174. Scheidel, A., Sorman, A.H., 2012. Energy transitions and the global land rush: Ultimate drivers and persistent consequences. Glob. Environ. Change, Global transformations, social metabolism and the dynamics of socioenvironmental conflicts 22, 588–595. doi:10.1016/j.gloenvcha.2011.12.005 Schneider, A., Friedl, M.A., Potere, D., 2009. A new map of global urban extent from MODIS satellite data. Environ. Res. Lett. 4, 044003. doi:10.1088/1748-9326/4/4/044003 Smil, V., 2015. Power Density: A Key to Understanding Energy Sources and Uses. The MIT Press, Cambridge, Massachusetts. Smil, V., 2010. Energy Transitions: History, Requirements, Prospects. Praeger, Santa Barbara, California, USA. Smil, V., 2008. Energy in nature and society: general energetics of complex systems. MIT Press, Cambridge, Massachusetts, USA. Smil, V., 1984. Views: On Energy and Land: Switching from fossil fuels to renewable energy will change our patterns of land use. Am. Sci. 72, 15–21. Smith, P., Gregory, P.J., Vuuren, D. van, Obersteiner, M., Havlík, P., Rounsevell, M., Woods, J., Stehfest, E., Bellarby, J., 2010. Competition for land. Philos. Trans. R. Soc. B Biol. Sci. 365, 2941–2957. doi:10.1098/rstb.2010.0127 Sorensen, B., 1999. Long-term scenarios for global energy demand and supply: four global greenhouse mitigation scenarios (No. 359). Roskilde Universitet, Roskilde, Denmark. Soulé, M.E., Sanjayan, M.A., 1998. ECOLOGY: Conservation Targets: Do They Help? Science 279, 2060–2061. doi:10.1126/science.279.5359.2060 Squatrito, R., Sgroi, F., Tudisca, S., Trapani, A.M.D., Testa, R., 2014. Post feed-in scheme photovoltaic system feasibility evaluation in Italy: Sicilian case studies. Energies 7, 7147–7165. Šúri, M., Huld, T.A., Dunlop, E.D., Ossenbrink, H.A., 2007. Potential of solar electricity generation in the European Union member states and candidate countries. Sol. Energy 81, 1295–1305. doi:10.1016/j.solener.2006.12.007 Trainer, T., 2013a. Can Europe run on renewable energy? A negative case. Energy Policy 63, 845–850. doi:10.1016/j.enpol.2013.09.027 Trainer, T., 2013b. Can the world run on renewable energy? A revised negative case. Humanomics 29, 88–104. doi:10.1108/08288661311319166 Trainer, T., 2012. A critique of Jacobson and Delucchi’s proposals for a world renewable energy supply. Energy Policy 44, 476–481. doi:10.1016/j.enpol.2011.09.037 Trainer, T., 2010. Can renewables etc. solve the greenhouse problem? The negative case. Energy Policy 38, 4107– 4114. doi:10.1016/j.enpol.2010.03.037 Turner, J.A., 1999. A Realizable Renewable Energy Future. Science 285, 687–689. doi:10.1126/science.285.5428.687 UN, 2015. World Population Prospects: The 2015 Revision. United Nations, Department of Economic and Social Affairs Population Division. ST/ESA/SER.A/313. UN Comtrade, 2015. UN Comtrade Database. US ACT, 2007. Energy Independence and Security Act of 2007 (No. PUBLIC LAW 110–140—DEC. 19, 2007). US EIA db, 2015. International Energy Statistics (Database). US Energy Information Administration, http://www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm. Vačkář, D., 2012. Ecological Footprint, environmental performance and biodiversity: A cross-national comparison. Ecol. Indic., The State of the Art in Ecological Footprint: Theory and Applications 16, 40–46. doi:10.1016/j.ecolind.2011.08.008 Valin, H., Peters, D., van den Berg, M., Frank, S., Havlik, P., Forsell, N., Hamelinck, C., Pirker, J., Mosnier, A., Balkovic, J., others, 2015. The land use change impact of biofuels consumed in the EU: Quantification of area and greenhouse gas impacts. Wackernagel, M., Schulz, N.B., Deumling, D., Linares, A.C., Jenkins, M., Kapos, V., Monfreda, C., Loh, J., Myers, N., Norgaard, R., Randers, J., 2002. Tracking the ecological overshoot of the human economy. Proc. Natl. Acad. Sci. 99, 9266–9271. doi:10.1073/pnas.142033699 49 Wagner, F., 2014. Considerations for an EU-wide use of renewable energies for electricity generation. Eur. Phys. J. Plus 129, 1–14. doi:10.1140/epjp/i2014-14219-7 Ward, D.J., Inderwildi, O.R., 2012. Global and local impacts of UK renewable energy policy. Energy Environ. Sci. 6, 18– 24. doi:10.1039/C2EE22342B WCED, 1987. Our common future (Report of the World Commission on Environment and Development). United Nations. Weber, C.L., 2009. Measuring structural change and energy use: Decomposition of the US economy from 1997 to 2002. Energy Policy 37, 1561–1570. doi:10.1016/j.enpol.2008.12.027 Weitemeyer, S., Kleinhans, D., Vogt, T., Agert, C., 2015. Integration of Renewable Energy Sources in future power systems: The role of storage. Renew. Energy 75, 14–20. doi:10.1016/j.renene.2014.09.028 WEO, 2014. World Energy Outlook 2014. OECD / IEA, Paris. Wiginton, L.K., Nguyen, H.T., Pearce, J.M., 2010. Quantifying rooftop solar photovoltaic potential for regional renewable energy policy. Comput. Environ. Urban Syst., Geospatial Cyberinfrastructure 34, 345–357. doi:10.1016/j.compenvurbsys.2010.01.001 Wirth, H., 2015. Recent facts about photovoltaics in Germany. Fraunhofer Institute for Solar Energy Systems, Germany. WWF, 2011. The energy report: 100% renewable energy by 2050. WWF, Ecofys, OMA.