Costs of meeting international climate targets without nuclear power

The impact of a global phase-out of nuclear energy is assessed for the costs of meeting international climate policy targets for 2020. The analysis is based on simulations with the Prospective Outlook on Long-term Energy Systems (POLES) global energy systems model. The phase-out of nuclear power increases GHG emissions by 2% globally and 7% for Annex I countries. The price of certificates increases by 24% and total compliance costs of Annex I countries rise by 28%. Compliance costs increase most for Japan (+58%) and the US (+28%). China, India, and Russia benefit from a global nuclear phase-out because revenues from higher trading volumes of certificates outweigh the costs of losing nuclear power as a mitigation option. Even for countries that face a relatively large increase in compliance costs, such as Japan, the nuclear phase-out implies a relatively small overall economic burden. When trading of certificates is available only to countries that committed to a second Kyoto period, the nuclear phase-out results in a larger increase in the compliance costs for the group of Annex I countries (but not for the EU and Australia). Results from sensitivity analyses suggest that the findings are fairly robust to alternative burden-sharing schemes and emission target levels.Policy relevanceNew calculations show that the impact of a global phase-out of nuclear energy on global mitigation costs is quite modest, but that there are substantial differences for countries. Total compliance costs increase the most for Japan and the US, but these are rather marginal if measured in terms of GDP. China, India, and Russia benefit from a nuclear phase-out because their additional revenues from selling certificates outweigh the additional costs of losing nuclear power as a mitigation option. The findings also highlight the importance of certificate trading to achieving climate targets in a cost-efficient way. If Japan or the US were to be banned from certificate trading, along with other countries, because of their non-participation in a second Kyoto period, then their compliance costs would increase substantially under a nuclear phase-out. The EU, however, would benefit because certificate prices would be lower.


Introduction
For many countries, nuclear power has long been considered a key ingredient to meeting electricity demand and achieving GHG emissions targets. According to Joskow and Parsons (2012), the main non-Annex I countries, den Elzen and Höhne (2008) have advocated reductions of 15 -30% below baseline emissions in 2020. According to a concrete proposal by the European Commission (EC, 2009a), Annex I countries should collectively reduce emissions by 30% in 2020 compared to 1990 levels, and economically more advanced non-Annex I countries should decrease emissions by 15 -30% below business as usual.
Although the climate summits in Copenhagen and Cancun in 2009 and 2010, respectively, did not lead to an international agreement involving binding GHG emissions targets for the post-Kyoto era, most Annex I countries pledged to meet quantifiable emissions targets under the Copenhagen Accord (UNFCCC, 2009) and the Cancun Agreement (UNFCCC, 2010). In addition, several non-Annex I countries submitted Nationally Appropriate Mitigation Actions (NAMAs). The implied emissions targets, however, are unlikely to be consistent with a path towards reaching the 28C target (den Elzen, van Vuuren, & van Vliet, 2010;Höhne et al., 2012;Rogelj et al., 2010).
At the UNFCCC climate change conference in Doha in 2012, some Annex I countries committed to a second commitment period under the Kyoto Protocol, transforming their pledges for 2020 into binding reduction targets under an international agreement. Because large Annex I emitters such as Japan and Russia have refused to sign such an agreement, the amendment to the Kyoto Protocol regulates only about 15% of global GHG emissions. Currently, no country-specific targets are being debated at the international level for beyond 2020.
The impact of a potential global nuclear phase-out on the costs of meeting international climate policy targets for 2020 is assessed here. Methodologically, the analysis relies on simulations by a global partial equilibrium model, which allows for a wide range of electricity generation technologies and for a differentiated assessment of impacts for numerous countries. The simulations take into account that a phase out of nuclear may alter countries' baseline emissions and restrict their options to mitigate GHG emissions. The climate policy scenario used involves a uniform 30% reduction target for Annex I countries. For non-Annex I countries the targets are derived from the NAMAs submitted under the Copenhagen Accord and Cancun Agreements. The trading of emissions certificates across countries is also allowed to assess the impact of the phase-out of nuclear power on certificate prices, countries' revenues from certificate trading, and domestic mitigation efforts.
In Section 2 the current literature is briefly reviewed. In Section 3 the baselines for the reference and the nuclear phase-out scenarios are described. In Section 4 the climate policy scenario used in the analysis is introduced. In Section 5 the results of the modelling analysis for the climate policy scenario is presented. In Section 6 the main findings from additional policy scenarios involving alternative trading rules, burden-sharing rules among Annex I countries, and a more ambitious reduction target for the group of Annex I countries are summarized. The main findings and conclusions are presented in Section 7. den Elzen et al. (2011), McKibbin, Morris, andWilcoxen (2011), Peterson, Schleich, and Duscha (2011), Saveyn, van Regemorter, and Ciscar (2011), Dellink, Briner, and Clapp (2011), and Ciscar et al. (2013) have analysed the economic implications of the Copenhagen and Cancun pledges prior to the Fukushima accident. Their findings suggest that the economic costs, in terms of lower gross domestic product (GDP), consumption, or welfare compared to baseline levels, are rather low at the global level and for most individual countries. Economic costs are typically below 1%, particularly if the trading of emissions certificates is allowed. McKibbin et al. (2011) found significantly higher costs, mainly because emissions were assumed to grow rather strongly in the baseline. Methodologically, these studies typically rely on 'top-down' dynamic computable general equilibrium (CGE) models, which account for macroeconomic effects resulting from changes in prices, income, or exports and imports. Thus, topdown modelling typically does not allow for a specific treatment of generation processes such as nuclear energy technology. Only den  has used a 'bottom-up' partial equilibrium model. Although such models typically include a rather detailed representation of technologies, they can hardly capture macroeconomic effects.

Background
Only a few studies have focused on the role of nuclear power in global emissions mitigation scenarios (Bauer, Brecha, & Luderer, 2012;Kurosawa, 2000;Rafaj & Kypreos, 2008;Remme & Blesl, 2008;Vaillancourt, Labriet, Loulou, & Waaub, 2008). Assuming rather modest targets in Annex I countries for 2030 of 92% and 108% of 1990 emission levels, Kurosawa (2000) found that the cost of a global phase-out of nuclear energy amounts to 0.36% lower consumption. Vaillancourt et al. (2008) found nuclear power to be the dominant power technology, having a share in the power mix of more than 50% in 2100 under various emissions reduction scenarios. Rafaj and Kypreos (2008) concluded that, as a result of a nuclear phase-out, global CO 2 e emissions in 2050 would be 15% higher (a reduction below 2000 levels of 42% instead of 49%). According to Remme and Blesl (2008), annual costs of reaching the 28C targets may be lowered by 9% if nuclear energy is allowed to increase by two-thirds compared to the base case, and account for a share of 35% of global electricity generation compared to 21% in the base case. Comparing harmonized long-term, low-carbon stabilization scenarios with five models, 1 Edenhofer et al. (2010) concluded that, when investment in nuclear power stops after 2000, the additional aggregated global mitigation costs (measured as a percentage of GDP in 2100) will increase by up to 0.7 percentage points. Only the study by Bauer et al. (2012) was motivated by the Fukushima accident. Linking a long-term, top-down growth model with a bottom-up model, they analysed the global impact of decommissioning extant nuclear power plants and restricting future investments in new nuclear power capacity under long-term emissions caps that are consistent with the 28C target. The near-term effect on GDP of a nuclear phase-out will be rather small (a loss of less than 0.1% in 2020), and somewhat larger in the long term (loss of 0.2% in 2050). For 2020, ambitious climate policy will lead to a loss in global GDP of around 1.2%.
Overall, most studies have relied on bottom-up models to explore the role of nuclear power in meeting climate policy targets. Most (but not all) studies have found that the additional costs of a nuclear phaseout will be low, and will amount to less than 1% of global GDP. 2 Furthermore, the extant studies have demonstrated only a weak link to actual climate policy and have not allowed for certificate trading. Finally, because most modelling analyses tend to be aggregated at the country and regional level, they often have not allowed for a country-specific representation of technologies or policy impacts.

Methods and baselines
For the baseline and policy simulations the model POLES (Prospective Outlook on Long-term Energy Systems), a world simulation model for the energy sector, was used. POLES is a techno-economic model with endogenous projection of energy prices, and takes complete account of the demand and supply of energy carriers and associated technologies. A more detailed description of POLES is presented in the online Supplementary Material Appendix A, and for further details see also Criqui (2009), IPTS (2010), Kitous, Criqui, Bellevrat, and Château (2010), and DECC (2012).
As POLES is a partial equilibrium model, the GDP for each region is given exogenously (together with population), unlike in CGE models, for example. Hence, POLES does not model all economic linkages within an economy, such as income effects or price effects, and does not allow energy or climate policy to affect GDP, employment, consumption, or other measures of economic welfare. 3 In this study, POLES was used to generate a reference baseline and a nuclear phase-out baseline. Although the policy analysis focuses on 2020, the assumed phasing out of nuclear energy in the nuclear phase-out baseline was embedded in a longer-run phase-out path until 2050. 4 The baselines abstract from the fact that climate change may affect economic development or energy demand (e.g. heating and cooling needs) and energy supply (e.g. availability of hydropower, biomass).
Both baselines rely on the same macroeconomic assumptions: world population is expected to reach 7.6 billion in 2020 (UN, 2009) and global GDP growth is expected to evolve at an average rate of 4% between 2010 and 2020.
The reference baseline was calibrated on the energy balances of the 'Current Policies' scenario in the World Energy Outlook (WEO) 2010 (IEA, 2010b). This reference case represents a world in which no additional climate policies are implemented. Global power generation is assumed to grow by 3% and nuclear energy by 1.9% per year between 2010 and 2020 to meet rising energy demand, in particular in developing countries ( Figure 1). Between 2010 and 2020, global electricity generation in the reference baseline was assumed to increase by 34% (from 20,700 to 27,700 TWh). Fossil fuels remain the dominant source of power generation in 2010 and 2020 (a share of 65%), with a key role played by coal (about 40%). Between 2010 and 2020, global coal-based power generation increases by more than 40%. This above-average growth is mostly driven by the demand in emerging economies.
The second-most important fuel in 2010 is natural gas (21%), which grows by 32% until 2020, and keeps its share in the global power mix about constant. The share of renewables in global electricity generation rises from 20% in 2010 to 22% in 2020, which corresponds to an increase in generation of 45% over this 10-year span. Finally, the share of nuclear power decreases from 13.5% in 2010 to 12% in 2020, while absolute generation increases by 20% and installed nuclear capacity by 23%. This growth in nuclear energy is mainly driven by emerging economies (particularly in China), with a strong increase from 14 to 58GW installed capacity. India almost doubles its installed capacity between 2010 and 2020 (from 6 to 11 GW).
Moreover, nuclear power generation is concentrated in only a few countries: the US, China, France, and Japan account for more than 60% of global nuclear power production. South Korea, Russia, and Canada together produce another 15%. The share of nuclear power in the national power mix differs substantially and ranges from 72% in France to 36% in South Korea, 31% in Japan, 18% in the US, 15% in Russia and Canada, and 7% in China. In contrast to most other countries, nuclear power production in Germany decreases between 2010 and 2020, even in the reference baseline, because it had decided to phase out nuclear power before the Fukushima accident (but at a somewhat slower rate, see e.g. Lechtenböhmer & Samadi, 2012, for details). This also translates into a small decrease in nuclear power generation for the EU in the reference baseline. In India, nuclear power accounts for a rather small share in the power mix, i.e. 3.5% in 2010 and 4.2% in 2020. As a consequence, the country-specific analyses offered below often disregard India.
In the nuclear phase-out baseline, no new nuclear capacities are built, and existing nuclear capacities are progressively decommissioned over the next four decades. The speed of the phase-out was determined on a country-by-country level, based on the average age of nuclear power plants. Although by 2050 not all nuclear power plants are phased out, the production of electricity from such plants is reduced to about 1% (i.e. 500 TWh) of global power generation, compared to 11% in the reference baseline. In the medium term, nuclear power accounts for about 11% (3350 TWh) of global power generation in the WEO baseline by 2020, but only 8% (2100 TWh) in the nuclear phase-out baseline.
The decrease in nuclear power generation in 2020 by 1250 TWh in the nuclear phase-out baseline corresponds to about 5% of global power generation and is mostly compensated by a stronger deployment of fossil fuels ( Figure 2). The shares of coal and natural gas in global power generation increase, respectively, from 42% to 45% and from 21% to 22%, while the share of renewables increases from 22% to 23%. The higher generation costs of the power mix lead to higher electricity prices and a decrease in global power production by 60 TWh (0.2%).
In the nuclear phase-out baseline, global GHG emissions in 2020 are about 2.2%, i.e. 800 million metric tons (Mt) CO 2 e, higher compared to the reference baseline. For most countries, the nuclear phase-out leaves baseline GHG emissions almost unchanged, because nuclear energy is not an important part of their national power mixes. As expected, countries with a high dependency on nuclear power generation in the reference baseline tend to experience a significant increase in GHG emissions, in particular Japan, Canada, and Russia ( Figure 3). In Japan, where two-thirds of power generated from nuclear plants is replaced by natural gas (and not by coal), the phase-out increases GHG emissions by 7% (80 MtCO 2 e) in 2020. Because of the lower share of nuclear energy in their national power mixes, GHG emissions rise less in China (3%) and the US (2%), although nuclear power production is mainly replaced by coal. In absolute terms, however, the increase in emissions is largest in China (300 MtCO 2 e) and the US (100 MtCO 2 e).
Interestingly, GHG emissions for a small number of countries are lower in the nuclear phase-out baseline than in the reference baseline in 2020. For example, in Finland, France, and Sweden, where nuclear energy plays a key role and strong renewable support policies are deployed, the power mix in the reference baseline largely relies on a mix of nuclear and renewable energy. A nuclear phase-out then leads to an offset of nuclear power by renewable energy in those countries. 5 Moreover, energy prices increase (compared to the reference baseline) and lead to lower energy demand and also to lower CO 2 e emissions compared to the reference baseline. 6 Total emissions in the EU in the nuclear phase-out baseline are about the same as in the reference baseline because the phase-out results in higher CO 2 e emissions of comparable magnitude in other EU Member States.

Climate policy scenario
The climate policy scenario used includes GHG emission targets for Annex I and non-Annex I countries for 2020 that are deemed consistent with meeting the 28C target.

Emissions targets for 2020
The aggregate emissions target for Annex I countries for 2020 was taken from the proposal by the European Commission (EC, 2009a), which assumes emissions reductions of 30% below 1990 levels. Following Peterson et al. (2011), this scenario may be interpreted as an illustrative example for possible post-Kyoto climate targets, which are consistent with the 28C target. Although, in principle, there are numerous ways of splitting the 30% reduction target between Annex I countries, the simplest type of burden-sharing rule was chosen: each Annex I country faces a uniform reduction rate of 30% below 1990 levels. 7 For non-Annex I countries the targets were derived from the NAMAs they submitted under the Copenhagen Accord and Cancun Agreements (i.e. only non-Annex I countries that submitted a NAMA [NAMA-NAI] face emission targets in the climate policy analysis). Several NAMAs have defined the emissions target as a target rate below baseline emissions and not as an absolute emission target derived from emission levels in a historic base year. As most NAMA submissions have not provided quantitative reduction targets, these submissions were translated into quantitative reduction targets. In the case of China and India, which provided CO 2 e emissions intensity targets, the targets were calculated using emissions and real (2005) GDP based on market exchange rates. 8 For non-Annex I countries that submitted specific measures rather than general emissions reduction targets, the associated emissions reductions were calculated. To do so, these reductions were assumed to correspond to a threshold price of EUE10/tCO 2 e in 2020 (that is, NAMA-NAI countries are expected to implement the cheapest reduction measures available in the countries as NAMAs, where the cost of the most expensive measure implemented is E10/tCO 2 e). The emissions reductions that can be realized at this price are between 5% below baseline in Jordan and 20% below baseline in several African countries. The marginal abatement cost (MAC) curves of the reference baseline were used to derive these emissions reductions. 9 Unlike the reduction targets of Annex I countries, those of non-Annex I countries -expressed in absolute levels -vary between the reference and the nuclear phase-out climate policy scenario as they depend on the baseline emissions in 2020.  Table 1 shows the emissions reduction targets as a percentage of baseline emissions for the reference and the nuclear phase-out baseline. At the global level, the climate policy scenario implied GHG emissions reductions of 12% compared to baseline emissions in 2020 in both scenarios. Emissions for Annex I countries were, on average, 28% below emissions in the reference climate policy scenario, and 30% below emissions in the nuclear phase-out climate policy scenario. Thus, on average, the nuclear phase-out increases required emissions reductions of about 7% as baseline emissions were 2% higher compared to the reference climate policy scenario. In comparison, for non-Annex I countries the policy targets translate into GHG emissions that are 3% below baseline emissions in both scenarios.
In both scenarios, Australia, Canada, Japan, and the US faced more ambitious emissions targets than the group of Annex I countries on average. Japan and the US were most affected by the phase-out of nuclear energy. The differences in emissions targets below baseline increased by four percentage points for Japan and two percentage points for the US. This corresponds to an increase in total emission reductions of 13% for Japan and 5% for the US. For Russia, the uniform 30% reduction target implies rather modest reductions compared to baseline emissions. Because baseline emissions in Russia are higher in the nuclear phase-out baseline, required emissions reductions increased from 7% to 9% below baseline compared to the reference climate policy scenario. For the Ukraine the uniform reduction rate means that GHG emissions may exceed baseline emissions by 36% and 37% in 2020.
As NAMAs for non-Annex I countries are calculated below baseline emissions, percentage figures do not change for the two scenarios. Among the non-Annex I countries listed in Table 1, South Africa, South Korea, and Mexico face the most ambitious reduction targets relative to both baseline scenarios. For China and India, the efficiency targets pledged under the Copenhagen Accord and Cancun Agreements translate into emissions reduction targets for 2020 that correspond to the baseline emissions. 10

Certificate trading
All Annex I countries are allowed to trade emissions certificates with one another, i.e. they may exchange Assigned Amount Units (AAUs). Non-Annex I countries may sell offsetting credits (CERs) to any Annex I country. However, trading of CERs is assumed to be governed by three restrictions. First, to avoid double counting, NAMA-NAI countries can only generate and sell CERs for emissions reductions that go beyond their own domestic NAMA targets. Second, non-Annex I countries can realize only 20% of their mitigation potential via CERs. This share is consistent with Castro (2010), who found that only a small amount of a country's mitigation potential is realized under the Clean Development Mechanism (CDM; see also Duscha & Schleich, 2013). Third, Annex I countries face a limit in the use of CERs to fulfil their own reduction targets, as Annex I countries debated during the discussions for the Copenhagen Accord. This CER quota is set to 20% of the emissions reductions below baseline and applies to all Annex I countries.
The Annex I countries allowed to trade in either of the scenarios need to fulfil at least 50% of the required emissions reductions below baseline domestically (domestic compliance quota). Because the domestic compliance quota may prevent perfect arbitrage, the costs of domestic mitigation efforts in countries in which the domestic compliance quota is binding will exceed the market price of AAUs. Although the CER-quota and the domestic compliance quota reflect features of actual climate policy discussions, they prevent achieving the globally cost-efficient outcome via the trading mechanism.

Results of climate policy scenario
For all countries and regions included in the model, sets of MAC curves were generated from the reference and nuclear phase-out baselines, following a approach similar to that taken by Anger (2008), den Elzen et al. (2011), andSchleich (2013), by progressively introducing a range of carbon prices. Higher CO 2 e prices not only increased the deployment of nuclear power to reduce the CO 2 e emissions in the reference scenario, but also spurred other mitigation options such as energy efficiency improvements, fuel switching from coal to gas, and the deployment of renewables.
Based on the two sets of MAC curves, the impact of the nuclear phase-out on certificate prices, domestic mitigation effort, certificate trading, power generation, and compliance costs may be evaluated for the climate policy scenario. Table 2 displays the prices of AAUs and CERs in 2020 for the reference and nuclear phase-out baselines.

Certificate prices
Owing to the fact that trading between Annex I countries is not limited, Annex I countries faced equal marginal abatement costs of E61/tCO 2 e in the reference climate policy scenario, unless their domestic compliance quota is binding. The nuclear phase-out resulted in an increase in the price of AAUs of about 24% compared to the reference climate policy scenario. This increase reflected the (small) increase in required GHG emissions reductions in the nuclear phase-out climate policy scenario compared to the reference climate policy scenario (baseline effect) and the fact that nuclear power plants were no longer available as a mitigation option (mitigation cost effect). Similarly, the price of CERs was about 19% higher in the nuclear phase-out climate policy scenario compared to the reference climate policy scenario. Because the CER quota of 20% is binding in both cases in some Annex I countries, the price of CERs was below the price of AAUs. The vast majority of CERs were generated in China and India. This reflects both rather lenient emissions targets (equal to baseline emissions) and large potentials of low-cost mitigation options in these countries.

Emissions reductions and pattern of compliance
For most countries, the increase in prices for emission certificates between the reference and the nuclear phase-out climate policy scenarios was associated with changes in emissions reductions and in the pattern of compliance -i.e. whether countries met their emissions targets via domestic mitigation or via purchasing certificates from abroad ( Figure 4).
Typically for Annex I countries, the nuclear phase-out means that the emission targets become more ambitious (because of the baseline effect) and also lead to a change in the share of domestic mitigation efforts in total required compliance efforts (domestic compliance share) compared to the reference climate policy scenario. Figure 4 shows that the domestic compliance share ranges from a minimum of 50% in countries with particularly high mitigation costs, such as Australia, Canada, and Japan (i.e. the use of certificates is limited by the domestic compliance quota of 50%), and up to 71% in the EU.
For countries that employ nuclear power, the impact of nuclear phase-out on domestic compliance share was governed by two countervailing effects. First, the mitigation cost effect resulted ceteris paribus in a lower domestic compliance share. Second, higher prices for AAUs rendered additional domestic mitigation options profitable, leading ceteris paribus to a higher domestic compliance share. For countries that do not rely on nuclear power, only the second effect matters. As a consequence, for most countries the nuclear phase-out was associated with a higher domestic compliance share. In Russia, domestic emissions reductions were lower in the nuclear phase-out than in the reference climate policy scenario. Russia not only faces higher baseline emissions in the nuclear phase-out climate policy scenario, but also loses nuclear as a mitigation option. Both effects lower Russia's supply of AAUs (despite higher certificate prices) by about 80 million AAUs.
For China and the Ukraine, who are net sellers of certificates, the trading volume was noticeably higher in the nuclear phase-out than in the reference climate policy scenario. In contrast to Russia, countries such as India and the Ukraine, in which the share of nuclear power in the reference baseline is rather low, benefited from the higher certificate prices without losing a significant share of their mitigation potential. For China, emissions were substantially higher in the nuclear phase-out baseline (by 350 MtCO 2 e), and certificates sales increased by around 30 MtCO 2 e.

Power sector
As described in Section 3, a phase-out of nuclear power substantially affects the fuel mix in the baseline. Meeting ambitious climate policy targets will lead to additional adjustments in the power sector ( Figures 5 and 6). In both policy scenarios, the generation of coal-fired power was lower than in the baseline scenarios. While in the reference climate policy scenario nuclear power generation increased particularly, in the nuclear phase-out climate policy scenario natural gas and wind increased the most. In particular, power generation from natural gas, solar, and wind increased by 9%, 21%, and 34%, respectively, in the nuclear phase-out scenario compared to the reference climate policy scenario. These developments go together with a 4% reduction in global and Annex I electricity demand in the reference climate policy scenario and -because of the fairly stronger deployment of more expensive low-carbon technologies -with a 5% reduction in the nuclear phase-out climate policy scenario.
A comparison across countries reveals that the pattern of adjustment in the power mix will be quite similar in most countries and in line with the overall global pattern. Figure 6 shows the extent to which the effects differed across countries/regions with a high share of nuclear power. For example, the US will rely heavily on the expansion of nuclear power to meet their climate policy targets and showed a much stronger increase in electricity generated by wind (+ 65%), biomass (+ 28%), and natural gas (+ 6%) in the nuclear phase-out than in the reference climate policy scenario. In the US and China, the effects were an order of magnitude larger than in other 'high nuclear' countries. For France, the effects were small, because only a relatively small share of the nuclear capacity was phased out prior to 2020 in the nuclear phase-out baseline.
Consequently, all Annex I countries (with the exception of those in the EU) experienced an increase in CO 2 e emissions in the power sector in the nuclear phase-out climate policy scenario compared to the reference climate policy scenario. Hence, those Annex I countries will need to realize additional mitigation efforts in other domestic sectors or purchase certificates abroad.
In all Annex I countries, however, the power sector hosted a substantial share of domestic mitigation efforts in both policy scenarios, covering between 28% (Ukraine, nuclear phase-out) and 54% (US, Figure 5 Global electricity generation by fuel in 2020 for baseline and climate policy scenarios Costs of foregoing nuclear power 339 reference) of total emissions reductions. The share of the power sector in domestic mitigation efforts was lower in the nuclear phase-out than in the reference climate policy scenario in all countries with the exception of Canada. This difference was particularly large in Japan and Russia, where their power sectors' share of domestic mitigation efforts decreased from 41% to 32% and from 33% to 28%, respectively. The main increases in the contributions of other sectors were found in industry (Japan: + 5 percentage points) and residential and services (Japan and Russia: + 2 percentage points).

Compliance costs
Compliance costs reflect a country's costs of meeting its emissions target and are measured as the sum of the mitigation costs for domestic efforts in the energy system (i.e. domestic mitigation costs) plus the net costs of purchasing and selling certificates (i.e. trade costs). The compliance costs in 2020 for both the reference and the nuclear phase-out climate policy scenario are shown in Figure 7. The compliance costs of each country were disaggregated into domestic mitigation costs and trading costs. Accordingly, the phase-out of nuclear power increased compliance costs in the group of Annex I countries by 28%, although the effects vary significantly across countries.
As expected, the US and the EU, where the uniform 30% reduction targets imply the largest required emissions reductions below baseline of all Annex I countries, also carry the highest compliance costs in both policy scenarios. The US also faced the largest increase in absolute compliance costs due to the nuclear phase-out (E21 billion). In comparison, Japan faced the highest relative increase in compliance costs (+ 58%), followed by the US (+ 28%). By contrast, the Ukraine and Russia (who are net sellers of Figure 6 Changes in electricity generation in 2020 for countries with a high share of nuclear energy (policy scenarios versus reference and nuclear phase-out baselines) AAUs), as well as India and China (who are net sellers of CERs), benefited from the increase in certificate prices. At the same time, however, the nuclear phase-out also led to higher domestic mitigation costs for these countries. Taking both effects into account, India and China were better off in the nuclear phase-out climate policy scenario. Russia and the Ukraine also benefited from the phase-out of nuclear energy, even though it sold fewer certificates in the nuclear phase-out compared to the reference climate policy scenario, but at a higher price.
In general, the nuclear phase-out tended to increase a country's domestic mitigation costs, in combination with an increase in either trade costs (if the country is a net buyer of certificates) or trade revenues (if the country is a net seller). One exception to this pattern was the EU, in which the increase in the price of AAUs led to additional domestic mitigation efforts, and hence reduced the amount of AAUs purchased from abroad ( Figure 8). Figure 9 displays the compliance costs as a share of GDP for the group of Annex I countries and for Annex I countries with positive compliance costs. Compliance costs for the entire Annex I group were around 0.4% of GDP. In other words, in general, although costs to meet the 30% reduction targets will be low, there will be differences for each country. These differences will generally depend on the strictness of the targets and the countries' mitigation potential and mitigation costs. Total compliance costs were by far the highest in the US, followed by the EU. Compliance costs were quite modest if they were measured as a share of GDP; i.e. they were below 1% for the US and 0.5% for the EU. By contrast, although absolute compliance costs for Australia and Canada were significantly lower compared to Costs of foregoing nuclear power 341 those of the US and the EU, they still accounted for a significantly larger amount of GDP (between 1.5 and 2% of GDP). Thus, a 30% reduction target will have a more significant economic impact on these two countries than on the US or the EU.
Similarly, the nuclear phase-out increased the compliance costs as a share of GDP in Annex I countries by only 0.1 percentage points, which is rather small when measured for the group of Annex I countries. For Australia and Canada this share increased by 0.3 and 0.4 percentage points, and less for the US (0.15 percentage points), the EU (0.03 percentage points), and Japan (0.14 percentage points). Thus, for Japan, in which the nuclear phase-out led to the largest increase in total compliance costs of any country (Figure 10), this increase still amounted to a relatively small overall economic burden, even though nuclear is an important technology in its power production. By contrast, because of the resulting increase in prices on the carbon market, the global nuclear phase-out resulted in a more pronounced increase in costs in Australia and Canada, measured as a share of GDP, even though these countries did not rely on nuclear power. Figure 7 illustrates that the effects of a nuclear phase-out will differ across countries depending on the share of nuclear in their power mixes and on the importance of nuclear compared to other domestic Costs of foregoing nuclear power 343 mitigation options. To gain additional insights into the factors underlying the differences in countries' compliance costs in response to a nuclear phase-out, compliance costs changes were decomposed into two effects. The first effect reflects the difference in compliance costs due to the global increase in baseline emissions in the nuclear phase-out compared to the reference baseline (baseline effect). The second effect captures the additional compliance costs from losing nuclear power as a mitigation option (mitigation cost effect).

Decomposition of changes in compliance costs in baseline effect and mitigation cost effect
To quantify the baseline effect, each country's compliance costs were recalculated, assuming the baseline emissions from the nuclear phase-out baseline and employing the mitigation cost curves derived under the reference scenario (adapted baseline policy scenario). That is, countries in which the phase-out of nuclear energy led to higher (lower) baseline emissions must reduce more (less) emissions in the adapted baseline policy scenario than in the reference climate policy scenario. The baseline effect was then calculated as the differences in compliance costs between the adapted baseline policy scenario and the reference climate policy scenario. Note that, because the phase-out of nuclear energy leads to higher global baseline emissions than in the reference baseline, certificate prices were also higher in the adapted baseline policy scenario.
To quantify the mitigation cost effect, each country's compliance costs were recalculated, employing the mitigation cost curves from the nuclear phase-out scenario but assuming the baseline emissions from the reference baseline (adapted mitigation cost policy scenario). Now, countries in which the phase-out of nuclear energy led to higher (lower) baseline emissions must reduce less (more) emissions in the adapted mitigation cost policy scenario than in the nuclear phase-out climate policy scenario. The mitigation cost effect was then calculated as the difference in compliance costs between the adapted mitigation cost policy scenario and the reference climate policy scenario.
Any difference in costs between the nuclear phase-out and the reference climate policy scenario that cannot be explained by the sum of the baseline effect and the mitigation costs effect was captured by a residual. This residual reflects the interaction of baseline and mitigation effects and may be positive or negative depending on whether the effects amplify or weaken each other. The results of this decomposition analysis are shown in Figure 10. Note that, for all five countries, the residual is only around 10%.
For all countries but those in the EU, at least half the increase in compliance costs was attributable to the baseline effect. For Japan, the high share of the baseline effect reflected the large increase in baseline emissions (+ 7%) due to the phase-out of nuclear energy. The mitigation cost effect only explains about 30% of the compliance cost increase (i.e. the loss of nuclear power as a mitigation option in Japan only accounts for 30% of the overall compliance cost increase). For the US, the increase in baseline emissions explained about 50% of the overall compliance cost increase, while losing nuclear as a mitigation option accounted for 40% of the increase in compliance costs. Unlike in Japan and the US, the nuclear phase-out did not directly affect the baseline emissions or mitigation options of Australia and Canada. Instead, the increase in compliance costs reflected an indirect effect, i.e. the rise in certificate prices in the nuclear phase-out climate policy scenario.
By contrast, the mitigation cost effect in the EU explained the lion's share of the increase in compliance costs. Two factors drove this result. First, the EU did not experience an increase in baseline emissions, and second, due to relatively low additional domestic mitigation costs, the EU may alleviate the effects of higher certificate prices by increasing domestic reductions in the nuclear phase-out climate policy scenario. Hence, for the baseline effect, there was an indirect effect (i.e. certificate price increase), but no direct effect. For the EU the indirect effect was softened by higher domestic emissions reductions compared to the reference climate policy scenario. For the mitigation costs effect, there was a direct impact (i.e. losing nuclear as a mitigation option) and also an indirect effect.

Results of alternative policy scenarios
To gain further insights into the interaction of a nuclear phase-out and climate policy, an additional scenario analysis was conducted. Due to space limitations, only the main findings of these additional scenarios are presented here. 11

Restricted trading scenario (KP2)
The first additional scenario (KP2) differs from the climate policy scenario in only one aspect: trading of AAUs is limited to those Annex I countries that have committed to a second Kyoto period, i.e. Australia, Belarus, Croatia, Iceland, Kazakhstan, Liechtenstein, Monaco, New Zealand, 12 Norway, Switzerland, Ukraine, and the Member States of the EU. At the time the analysis was conducted, Canada, Japan, and Russia had already stated that they would not participate in a second Kyoto period. Also, the US will continue to abstain from the Kyoto Protocol. Hence, Canada, Japan, and the US can no longer rely on certificate trading for compliance and must intensify their domestic mitigation efforts, while Russia can no longer enjoy revenues from selling certificates.
Compared to the climate policy scenario (with full Annex I trading), prices of AAUs and CERs were 19% and 8% lower, respectively. Furthermore, compliance costs of Annex I countries in the restricted trading scenario were about 28% higher than in the reference policy scenario, and 29% higher than in the nuclear phase-out policy scenario. On the one hand, these figures reflect the savings in overall compliance costs, which may be realized via emissions trading, and on the other hand, they also illustrate that the nuclear phase-out will be more costly if certificate trading were restricted. Most prominently, in Japan, the nuclear phase-out led to a 200% increase in compliance costs (compared to 120% in the climate policy scenario). However, in the US, additional compliance costs due to the nuclear phase-out did not differ much relative to the climate policy scenario because the US may substitute the purchase of certificates with domestic reductions at rather modest additional domestic compliance costs. In contrast to the climate policy scenario, the nuclear phase-out made Russia worse off in KP2 because it no longer enjoyed revenues from selling AAUs. The lower certificate prices made net sellers (e.g. China, India, Ukraine) worse off compared to the climate policy scenario. At the same time, countries that faced stringent emission targets but may purchase certificates (i.e. Australia, EU) benefited from lower certificate prices. Compared to the climate policy scenario, the additional compliance costs due to a nuclear phase-out were almost 60% lower for the EU and about 50% lower for Australia. Overall, net additional compliance costs of a nuclear phase-out for the group of Annex I countries were 35% higher in the KP2 trade scenario than in the climate policy scenario.

Alternative target scenarios
The second set of additional scenarios involved alternatives to the climate policy scenario with respect to (1) the assumed uniform allocation of the reduction target for the group of Annex I countries and (2) the reduction target of 30% for this. These assumptions were varied in two additional target scenarios, while retaining the targets for non-Annex I countries (as in the climate policy scenario).
In principle, there are an infinite number of possible burden-sharing schemes. The uniform reduction target of 30% was compared to the indicator-based burden-sharing scheme developed by the European Commission (EC, 2009b). 13 Accordingly, the 30% reduction target among Annex I countries in this scenario (EC30%) was allocated across regions based on four equally-weighted indicators: GDP per capita (in 2005), which reflects a country's ability to pay; GHG per GDP (in 2005), which recognizes domestic emissions reduction potential; population trend (from 1990 to 2005), which accounts for 'needs'; and GDP trends (from 1990 to 2005), which reflect domestic 'early action'. 14 To test the sensitivity of the findings with respect to target stringency, the policy scenario was also compared to a scenario involving a more stringent 40% uniform reduction target for the group of Annex I countries (Uniform40% scenario). In the online Supplementary Material, Appendix C, Table  A2 provides an overview of the emissions reduction targets for Annex I countries applied within the alternative target scenarios. Figure 11 presents the findings on the change in compliance costs due to a nuclear phase-out for the alternative target scenarios for selected Annex I countries. Because the effects of these scenarios on non-Annex I countries were negligible and involved only indirect effects via changes in CER prices, the focus here is on the results for Annex I countries. Note that, to keep changes in positive compliance costs separate from changes in negative costs, Figure 11 does not show results for Russia and Ukraine. Figure 11 Change in compliance costs induced by nuclear phase-out in the alternative target scenarios. Note: 'Climate Policy' here refers to the 30% uniform reduction target for Annex I countries assessed in detail in this article.
The alternative burden-sharing scheme (EC30%) left the US, Canada, Australia, and Japan with a lower mitigation target than in the uniform 30% scheme. Thus, mitigation costs were lower in these countries. Figure 11 suggests that the phase out of nuclear leads to quite similar effects on percentage changes in compliance costs in Annex I countries in EC30%, as in the original climate policy scenario. For the group of Annex I countries, the nuclear phase-out increased mitigation costs by about 29% in both scenarios. The only noticeable difference was observed for Japan, which needed to offset a large share of nuclear power in the phase-out scenario. The lower mitigation target helped bring down these additional costs. The relative difference in certificate prices between the reference and phase-out scenarios was rather similar in both burden-sharing scenarios (see Table A3, online Supplementary Material, Appendix C). 15 Russia and Ukraine were net sellers of certificates in EC30% as well as in the climate policy scenario, but were better off under the latter scenario because EC30% means less hot air for both countries. 16 Intensifying the reduction target by moving from a 30% to a 40% target increased the level of compliance costs (by around 130% for the group of Annex I countries) and certificate prices (AAUs, + 80%; CERs, + 70%). The additional phase-out of nuclear, however, led to effects on percentage changes in compliance costs of Annex I countries that were quite similar to those in the climate policy scenario, i.e. an increase of about 26% for the group of Annex I countries. For Japan, the relative difference was markedly smaller for the more ambitious Uniform40% scenario. In absolute terms, however, the difference was quite pronounced. The nuclear phase-out increased compliance costs in Japan by about twice as much in the Uniform40% than in the climate policy scenario. Russia and Ukraine remained net sellers of certificates in Uniform40%. Finally, the relative change in certificate prices in response to a nuclear phase-out was similar to that in the other scenarios (see Table A3, online Supplementary Material, Appendix C).
In general, the results of the additional target scenarios suggest that the main findings for the climate policy scenario are fairly robust to the variations in burden-sharing and target level for Annex I countries considered.

Conclusions
In this article, a global energy systems model was used to analyse the effects of a global phase-out of nuclear power on the costs of meeting climate policy targets in 2020 that are consistent with the 28C target. In the climate policy scenario, Annex I countries face a uniform 30% reduction rate compared to their 1990 GHG emission levels, while non-Annex I countries are assumed to meet their NAMA targets. Simulations of the new baseline suggest that a long-term global phase-out of nuclear power by 2050 will lower the share of nuclear in the global power mix from 11% to 8% in 2020. This reduction will be offset almost entirely by a stronger deployment of fossil fuels and -in countries with ambitious support for renewable energy (e.g. the Member States of the EU) -also by renewables. As a result, global GHG emissions in the baseline will increase by 2% under a nuclear phase-out, and the emissions reductions required to meet the climate policy targets will increase by 3% globally.
Simulations of the climate policy scenario with unrestricted trading reveal that the nuclear phaseout will increase AAU prices by 24% and total compliance costs of Annex I countries by 28%. Japan (+ 58%) and the US (+ 28%) will face the largest relative increase, but China, India, Ukraine, and Russia will benefit because the additional revenues from selling certificates outweigh higher domestic abatement costs. Similar to Edenhofer et al. (2010) or Bauer et al. (2012), it was found that there will be a modest increase in compliance costs in relation to GDP.
To meet the 30% emissions reduction targets for 2020, domestic efforts in Annex I countries will involve the power sector in particular. The share of coal-based power generation will decline and the share of natural gas, nuclear power, and renewables (in particular wind power) will increase in the reference scenario. The nuclear phase-out will increase the share of natural gas, wind, and solar in the power mix of most countries, in particular in those countries that rely strongly on nuclear power (e.g. the US). Somewhat higher electricity prices will lead to a slightly lower demand than in the reference scenario.
Decomposing the overall changes in countries' compliance costs due to a nuclear phase-out into a baseline effect and a mitigation cost effect, it was found that the share of the mitigation cost effect will be about twice as high in the EU as in Australia, Canada, Japan, or the US. While the nuclear phase-out will hardly affect baseline emissions in the EU until 2020, the loss of nuclear power as a mitigation option will weigh rather heavily compared to other regions.
Results from alternative policy scenarios provided additional insights. When trading of AAUs is restricted to those Annex I countries that have committed to a second Kyoto period, compliance costs of Annex I countries in the climate policy scenarios will be about 28% higher than in the reference climate policy scenario, and 29% higher than in the nuclear phase-out climate policy scenario. These figures reflect the savings in overall compliance costs, which may be realized via unlimited emissions trading between Annex I countries. They further illustrate that the nuclear phase-out will be more costly if certificate trading is restricted. Also, the general findings on the relative impact of a global nuclear phase-out on global and regional patterns of compliance costs appear to be fairly robust given alternative ways of sharing the burden of emissions reductions across Annex I countries and more ambitious emission targets for the group of Annex I countries.
The modelling assumptions and findings presented here should, however, be interpreted with caution. Arguably, the assumed global phase-out of nuclear power may overstate actual long-term reactions to the Fukushima accident. Yet, the nuclear phase-out scenario presented serves as an interesting benchmark, as it reflects what may happen should concerns about the future of nuclear energy increase dramatically, and globally. Also, it should be kept in mind that by focusing on the year 2020, in which actual policy targets are available for most countries, the analysis takes on a relatively shortterm perspective. For example, the licences of most nuclear units in the US expire after 2030. In addition, there are the ongoing international climate diplomacy attempts to create binding targets that go beyond 2020. These targets need to be more ambitious than those implemented for 2020 to meet the 28C target with high probability. From this perspective, a global phase-out of nuclear power is expected to bring about stronger economic and environmental implications in the longer run than analysed for 2020. However, in the long run, the energy system will also exhibit higher flexibility and feature learning effects for low-carbon technologies. These factors will lower the adjustment costs of banning a major power generation and mitigation option such as nuclear power. The findings by Bauer et al. (2012) and Edenhofer et al. (2010) suggest that, in the long run, the economic impact of restricting nuclear power will be small compared to the impact of ambitious mitigation policies.