Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Energy transformation cost for the Japanese mid-century strategy

Energy transformation cost for the Japanese mid-century strategy ARTICLE https://doi.org/10.1038/s41467-019-12730-4 OPEN Energy transformation cost for the Japanese mid-century strategy 1,2,3 1 4 2,3,5 Shinichiro Fujimori *, Ken Oshiro , Hiroto Shiraki & Tomoko Hasegawa The costs of climate change mitigation policy are one of the main concerns in decarbonizing the economy. The macroeconomic and sectoral implications of policy interventions are typically estimated by economic models, which tend be higher than the additional energy system costs projected by energy system models. Here, we show the extent to which policy costs can be lower than those from conventional economic models by integrating an energy system and an economic model, applying Japan’s mid-century climate mitigation target. The GDP losses estimated with the integrated model were significantly lower than those in the conventional economic model by more than 50% in 2050. The representation of industry and service sector energy consumption is the main factor causing these differences. Our findings suggest that this type of integrated approach would contribute new insights by providing improved estimates of GDP losses, which can be critical information for setting national climate policies. 1 2 Department of Environmental Engineering, Kyoto University, C1-3 361, Kyotodaigaku Katsura, Nishikyoku, Kyoto city, Japan. Center for Social and Environmental Systems Research, National Institute for Environmental Studies (NIES), 16–2 Onogawa, Tsukuba, Ibaraki 305–8506, Japan. International Institute for Applied System Analysis (IIASA), Schlossplatz 1, A-2361 Laxenburg, Austria. School of Environmental Science, The University of Shiga Prefecture, Hikone, Japan. Department of Civil and Environmental Engineering, College of Science and Engineering, Ritsumeikan University, Kyoto, Japan. *email: sfujimori@athehost.env.kyoto-u.ac.jp NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 1 1234567890():,; ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 limate change mitigation is one of the greatest societal consistently dealing with energy systems and the stability of challenges facing most countries as reduction of energy- power generation. Crelated CO emissions is key to reducing greenhouse gas Here, we describe the macroeconomic implications of climate (GHG) emissions. In 2015, more than 190 countries reached the mitigation policy using an integrated modelling framework Paris Agreement (PA) and each country submitted their own wherein an energy system model, Asian-Pacific Integrated Model/ Nationally Determined Contribution (NDC). Along with those Enduse (AIM/Enduse), and a power-dispatch model, AIM/Power, targets, countries were also asked to engage in long-term plan- are inter-linked with the multi-sector economic model (AIM/ 2,3 ning, known as a mid-century strategy . Under the long-term CGE). We call this new soft-linking modelling framework an global goal in the PA of keeping the global mean temperature integrated model, which allows us to assess the macroeconomic increase well below 2 °C compared with the pre-industrial level, impacts of climate change mitigation with concrete specification the net CO emissions in this mid-century must be close to of detailed energy technologies, ensuring a stable power supply neutral according to numerous studies carried out using Inte- with consideration of long-term (seasonal and daily) and short- grated Assessment Models (IAMs) . term (less than hourly) power fluctuations. Macroeconomic costs of climate change mitigation is a great The principle of this methodology is based on the concept that concern for climate policy settings . The Intergovernmental Panel energy simulation from the energy system model is more reliable on Climate Change (IPCC) fifth assessment report summarises than that from the economic model, as energy supply and climate mitigation costs, and GDP or consumption losses in 2050 demand are technologically represented in detailed in the energy are around 2–6% to achieve the abovementioned 2 °C goal. system model. Similarly, the technological representation of There are multiple ways to interpret these numbers. It may be too power supply in the power-dispatch model is more reliable than expensive to pay for climate change prevention that delays GDP that in the energy system model. We overcome the disadvantages growth for a couple of years or low enough for avoiding wide- of these models by exchanging information and iterating it spread climate change impacts and irreversible risks associated among models. We begin with the AIM/Enduse run, which with catastrophic events. To address macroeconomic mitigation provides energy system information to AIM/CGE and AIM/ costs, IAMs normally represent GHG emissions reduction costs Power. Then, these two models’ outcomes are further fed into either through an energy system model or an economic model, AIM/Enduse. Finally, we confirm whether the models reach often termed bottom-up and top-down models, respectively. sufficient convergence for our purposes (see Supplementary Although there are other ways to classify the IAMs, in this Information for more detailed discussion about reaching con- paper, we define economic model as the model that includes vergence). See the Methods for indicators exchanged among multi-sectoral CGE model within the IAM framework, and models. Note that for CGE results, we compare the stand-alone energy system model as the model that does not. Note that a CGE model with the integrated model. power-dispatch model is also used in this study although that We applied this framework to Japan as a case study. The 6–8 is not usually classified as IAMs. There are many global , and Japanese government has declared a long-term GHG emissions 9,10 29 national energy system models as well as the economic reduction target of 80% by 2050 . As mitigation costs in Japan 11,12 30–32 models , which are based on multi-sectoral CGE models. estimated in previous studies vary significantly across IAMs , Traditionally, CGE models tend to project higher policy costs application of this framework would be beneficial for Japan’s than those of energy system models (see also Supplementary climate policies to communicate with the stakeholders. We ana- Note 1 and Supplementary Table 1). One possible reason for this lysed scenarios with and without climate mitigation policy, which tendency is that parameters in CGE models are calibrated against are the mitigation and baseline scenarios, respectively. a historical period in which it is difficult to decouple economic As results, we found that the macroeconomic costs are not as growth and CO emissions. Some argue that aggregated energy high as previously reported when energy system information is system representation is disadvantageous to understanding appropriately reflected in the economic model. The critical drastic energy system changes and their macroeconomic impli- determinants of mitigation costs that changed in the newly cations. Thus, incorporating energy system model information developed integrated model were identified as the representation into CGE models may lead lower macroeconomic costs than of industry and service sectors’ energy consumption, which is previously reported. associated with production functions. These findings may change Integrating CGE and energy system model offers a great the general perception of climate change mitigation costs in terms advantage in representing the feedbacks inherent across economic of macroeconomic losses and provide important policy insights. and energy systems. For the policy makers, macroeconomic implications including sectoral impacts provided by CGE models is more meaningful than energy system costs alone. To this end, Results 14–16 several attempts have been made , whereas investigators such Energy system in Japan’s mid-century strategy. An 80% 17–20 as Bohringer et al. incorporated disaggregated information reduction of GHG emissions requires substantial changes in the on power sectors. An extended literature list is shown in Sup- energy system compared to the current system or the baseline plementary Table 2 and there are more examples if we include scenario (Fig. 1a). As a result of Japan’s unique socioeconomic 21,22 non-multi-sector CGE models . circumstances, with a decreasing population and modest eco- At the meantime, drastic energy transformation requires large- nomic growth (Supplementary Fig. 1), the overall energy system scale variable renewable energy penetration. The key issue of the shows little changes in the future under the baseline scenario. The variability in renewable energy is strongly dependent on national- main changes of the baseline 2050 from the base year are the and local-scale grid systems, availability of solar and wind power, higher share of coal relative to other fossil fuels, and the decrease battery technology, and other energy sources that can be used to in the share of nuclear energy, which reflects the current societal balance demand and supply. Recently, some national modelling attitude toward nuclear power that limits new construction 23–25 studies have addressed these issues and integration of a (Fig. 1b). Regarding CO emissions, the baseline level is stable or power-dispatch model with an energy system model has been may even decline over time (Fig. 1d). Meanwhile, the mitigation 26 27,28 attempted . In IAMs, they are represented to some degree , scenario exhibits large-scale renewable energy penetration, slight which are adequate to provide global-scale energy analyses. energy demand reduction, compositional changes characterised However, no studies showed macroeconomic implications of by the use of more carbon-neutral energy sources, and 2 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE 20 1250 ab c d 0 0 0 ef g 0 0 0 Coal|w/o CCS Nuclear Wind Other Heat Solid|Biomass Liquid|Oil Baseline Mitigation Hydro Oil|w/o CCS Biomass|w/o CCS Gas Solid|Coal Solar Geothermal Gas|w/o CCS Electricity Liquid|Biomass Fig. 1 Main energy and emissions related variables. Primary energy source (a, e), power generation (b, f), final energy demand (c, g), CO emissions (d), and carbon price (h) projections. a, b, and c show the baseline scenario, whereas e, f, and g show mitigation scenarios. The other in a and e includes secondary energy net imports electrification (Fig. 1f). The price of carbon in the mitigation namely, without nuclear and without carbon capture and storage scenario increases over time and reaches ~1000$/tCO in 2050, (CCS). These results can be interpreted as a simple uncertainty which is due to the low-carbon technological availability in the analysis, but they have more meaningful policy implications AIM/Enduse model. because the perception of nuclear power in Japan has changed The power system relying heavily on variable renewable energy drastically since the Fukushima Daiichi accident, and there is requires measures to stabilise the power supply system and limited geologically appropriate space for CCS on Japanese demand responses. Curtailment in onshore wind increases, territory. Figure 3c illustrates the relationship of mitigation costs particularly after 2020 when variable renewables start to expand in the CGE stand-alone and integrated models for this sensitivity (Fig. 2a). Furthermore, when coal-fired power is completely analysis. Here, we again see systematically higher costs in the phased out around 2040, offshore wind also exhibits a clear stand-alone model than in the integrated model. Comparison of curtailment increase. The battery requirements for short-term these integrated model’s GDP losses and additional energy system fluctuations also increase sharply after 2020, whereas the capacity costs derived from AIM/Enduse shows a similar trend to that factor of thermal power plants declines (Fig. 2b, c). We also show in Fig. 3d. The overall energy and emissions trends for the daily electricity supply and demand profiles for selected days this sensitivity cases are provided in Supplementary Fig. 2 and in 2050 (Fig. 2d). Supplementary Fig. 3. Mitigation costs. Mitigation costs, as measured by GDP loss rates Mechanism causing the differences in macroeconomic costs. (hereafter GDP is accounted by the total final consumption), The central mechanisms for changing the macroeconomic increase over time as emissions reductions become deeper, as implications are changes in the productivity of primary factors illustrated in Fig. 3a. The CGE stand-alone results reach more (labour and capital) constituting value-added, which is the GDP than 2.5% after 2030, whereas the integrated model is lower, measure in production side. This is because the primary factor around 1.2% in 2050 (Fig. 3a). The equivalent variation also inputs are constrained exogenously for each year in our economic 33,34 shows similar trend as GDP losses (Fig. 3b). The additional model while the capital and labour inputs change dynami- energy system costs in the AIM/Enduse stand-alone are plotted in cally with population development, and GDP growth. The the same figure, and are notably similar to the integrated model straight-forward reason that GDP losses are lower in the inte- results (blue lines in Fig. 3a). The mitigation costs under such grated model than in the stand-alone model is that the parameter deep emissions reductions from CGE studies are usually not as assumptions in CGE models differ between the stand-alone low as our estimates (2–6% of GDP losses in 2050) . Once the model and integrated model. The former relies on the existing energy system model’s results are reflected in the economic literature and the latter on the energy system model outputs. model, the integrated model would be able to estimate similar Consequently, the primary factor productivity is higher in the mitigation costs to those from energy system models. integrated model than in the stand-alone model. Then, the dif- We further implemented sensitivity scenarios with varying ferences in productivity are mainly driven by two things. One is technological availability, which may lead to non-linear energy the productivity decreases associated with emissions reductions in system responses, to investigate the robustness of our findings. energy end-use sectors, such as industry, transport and service For this purpose, we selected two technological variation sectors (e.g. capital replacement by expensive but energy-efficient scenarios wherein more power stability measures are needed; ones). The other is sectoral allocation changes in primary factors. NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 3 Primary energy (EJ/year) Power generation (EJ/year) Final energy (EJ/year) Mt CO /yr US$2005/t CO 2 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ab Solar|PV Pumped hydro Wind|Offshore Battery for long-term Wind|Onshore Battery for short-term 0 0 Year Year c d Summer Intermediate Summer weekday weekend weekday PV_H PV_H PV_L Thermal power total Charge Gas Pumping Biomass Discharge Coal Biomass|w/o CCS Pumped hydro Hydro Coal|w/ CCS PV Nuclear Coal|w/o CCS Wind Loss Gas|w/ CCS Gas|w/o CCS Demand_w/_HP&EV Demand_w/o_HP&EV –50 Hour Year Fig. 2 Power system reactions to large-scale renewable energy penetration. a Wind and solar curtailment rate, which is the unused energy divided by total power generation for each energy source, b the installed capacity of technologies to stabilise fluctuations in the electricity supply, c the capacity factor of thermal power, and d the profiles of electricity demand and supply on selected three typical days (PV_H and PV_L indicate sunny and cloudy days, respectively) Regarding the first factor, Fig. 4a illustrates the capital The second factor, namely the effect of sectoral primary factor input efficiency of major industrial sectors (top 10 industries, allocation changes, is mainly driven by the power generation sector. which account for 95% of GDP in the base year) in the The electricity generation in the mitigation scenario compared to mitigation scenario compared to the baseline scenario for the baseline scenario is about 20% higher in the stand-alone model, stand-alone and integrated models in 2050. Here, we define the but almost the same in the integrated model (Fig. 4). There are capital input efficiency as capital input per output for each certainly differences in technological shares between the stand-alone sector, which is a model outcome. Higher values indicate and integrated models, but, in summary, it seems that the difference that additional capital inputs are needed in the mitigation in total electricity generation between the models is the dominant scenario compared to the baseline scenario. In general, the factor, where the stand-alone model requires additional capital and stand-alone model requires larger capital inputs than the labour inputs, accounting for 0.4 percentage points of GDP, relative integrated model in the mitigation scenario. We can roughly to the integrated model, which relies on the AIM/Enduse outputs compute the value-added losses associated with these capital (Fig. 4d). With respect to the representation of electricity demand, productivity losses by multiplying the value-added of each the total electricity demand is determined by energy consumption sector Fig. 4b, c). These eventually account for 1.3 percentage in the energy end-use sectors, which are represented by a CES points of the total value-added (GDP). Then, the productivity function, as mentioned above. The fuel-wise share is determined differences between the stand-alone and integrated model are using a logit function in both the stand-alone and integrated mainly caused by differences in the functional form and models. A parameter representing the preferences or technological parameters particular to the value-added and energy bundle. choices in the logit function is determined endogenously in the Here, we use a CES function in which the substitution elasticity, integrated model, based on the AIM/Enduse results, whereas they share parameters and future autonomous energy efficiency are are exogenous parameters in the stand-alone model. We describe defined in the stand-alone model. The integrated model the detailed mathematical formation and assumptions in the uses a function of the same form, but the additional investment Supporting Information. and energy inputs are exogenously given by AIM/Enduse, In addition to the two main mechanisms mentioned above, the whereas the CES shift parameters are determined endogenously productivity changes and sectoral shifts in other sectors certainly (sector-wise additional investments are shown in Supplemen- occur, but are relatively minor. In summary, the differences in tary Table 3). GDP changes between the stand-alone and integrated models are 4 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 2040 2020 % % GW GW NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE a b 4 4 3 3 2 2 0 0 Year Year CGE stand-alone CGE stand-alone Integrated model Integrated model Enduse stand-alone cd 4 1.5 1.0 0.5 0 0.0 CGE stand-alone (%/year) Energy system model (%/year) Default Default No CCS No CCS No nuclear No nuclear Fig. 3 Climate change mitigation cost. a, b Time-series mitigation cost AIM/CGE results are represented as GDP loss rates and equivalent variation change rates relative to baseline scenarios. AIM/Enduse results are expressed as additional energy system costs of GDP relative to baseline scenarios. c, d 5-year mitigation costs with varying technological availability; c illustrates the relationship of GDP losses in the CGE stand-alone and integrated models, and d shows GDP losses in the integrated model and additional energy system costs in AIM/Enduse. The energy system model results shown here correspond to Enduse_results1 in Supplementary Fig. 5 explained above, but, generally speaking, many interactions model runs in Fig. 5. The stand-alone CGE model shows simultaneously occur in the CGE model and sometimes the remarkable value-added decreases in the industry (IND) and cause and consequences are not clear. service sectors (SER) in 2030, whereas the integrated model does not. These trends remained consistent for the year 2050, with the CGE stand-alone model showing large changes in the service Decomposition of mitigation costs and sectoral contributions. sector. This result is consistent with those described in the pre- vious section, wherein the industry and service sector’s energy To identify which sectors contribute to GDP losses, the value- added by each sector, as estimated by the economic model, is system information, i.e. the representation of production func- decomposed into three factors of output changes, value-added tions in those sectors, are critical factors for differentiating overall productivity (output per value-added), and residuals. Moreover, GDP losses between the two models. The output decrease in the we compared the outputs of stand-alone CGE and integrated service sector is the largest element to change the GDP in the NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 5 0.0 0.5 1.0 1.5 Integrated model (%/year) GDP loss rates (%/year) Integrated model (%/year) Equivalent variation (%/year) ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ab 1.2 0.6 1.1 0.4 CGE stand-alone 1.0 Integrated model 0.2 0.9 0.0 cd Sectors Sectors Geothermal 01 Other services Biomass 02 Other manufacturing Wind Stand-alone model 03 Transport and communications Solar 04 Construction Hydro 05 Chemical, plastic and rubber products Nuclear 06 Food products Gas 07 Paper, paper products and pulp Oil 08 Iron and steel Coal 09 Non-ferrous products Integrated model 10 Light industry 11 Others Contribution of capital productivity change to value added changes (% of GDP) Fig. 4 Valued-added differences between baseline and mitigation scenarios. a Sectoral capital input efficiency for the top 10 industrial activities in 2050. Capital efficiencies in baseline and mitigation scenarios are computed and then the capital efficiency in the mitigation scenario relative to the baseline scenario is shown. b Sectoral value-added share in the baseline scenario for the top 10 industrial activities. c The capital productivity value-added effects compared to the total value-added for the top 10 industrial activities. Dots means net total changes. Negative and positive values mean capital productivity gain and losses compared to baseline scenarios respectively. d The value-added share of power sectors in terms of total economy-wide value-added CGE stand-alone model. This result may be driven by changes in overall insights are clear, that the industry and service sectors are household expenditures for services, which were around 3.4 and key in determining macroeconomic implications. 0.0% in the CGE stand-alone and integrated models, respectively, in 2050. These differences may be due to changes in total income. We ran further diagnostic scenarios with and without Discussion incorporating energy system information by sectors (see Methods Our newly proposed integrated model approach implicitly for more details) to investigate the extent to which the energy assumes that the energy productivity in the CGE model is system model’s output information for each sector contributes to endogenized by using the energy system model information. This mitigation cost differences compared to the stand-alone CGE. treatment is somewhat different from the conventional approach, Comparing scenarios that include a single sector’s information in which CGE models use the same Autonomous Energy Effi- from AIM/Enduse and the stand-alone model (Row 1–6 ciency Improvement (AEEI) and constant elasticity substitution in Table 1, respectively), the inclusion of the industry and service parameters, with and without mitigation policies. Based on the sector information from AIM/Enduse makes a remarkable results showing that the macroeconomic costs associated with difference in the GDP loss rate (Row 5 and 4 in Table 1, climate change mitigation policies are lower than estimated using respectively). From the opposite side, the scenarios taking out the conventional approaches, we can interpret the energy pro- AIM/Enduse information for each sector (Row 7–11 in Table 1) ductivities in the mitigation scenarios as being higher than in the show that excluding the industry and service sectors consistently conventional approach. This would imply that the AIM/Enduse generates GDP loss differences compared to the integrated model model incorporates higher productivity technological information (Row 12 in Table 1). Conversely, the incorporation of residential, than the conventional CES approach. transport and energy supply sector information given by AIM/ Overall, as long as an energy system model is more reliable than Enduse has a small impact on GDP losses, or even has the the CGE model in terms of energy-related variables, the energy opposite effect in some cases. Finally, we can see cross-sectional representation in the conventional CGE should be replaced by the effects in other scenarios in Supplementary Table 4, which energy system model outputs. The contributions of the industry and indicates the complexity of the results and shows that the service sectors to GDP loss differences are caused by the production influence of each sectoral impact is not additive. However, the function form and its parameters. Basically, for most conventional 6 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 01 Other services 02 Other manufacturing 03 Transport and communications 04 Construction 05 Chemical, plastic and rubber products 06 Food products 07 Paper, paper products and pulp 08 Iron and steel 09 Non-ferrous products 10 Light industry 01 Other services 0.0 02 Other manufacturing 0.5 03 Transport and communications 1.0 04 Construction 1.5 05 Chemical, plastic and rubber products 06 Food products 07 Paper, paper products and pulp 08 Iron and steel 09 Non-ferrous products 10 Light industry Baseline.Integrated model Mitigation.Integrated model Baseline.Stand-alone model Mitigation.Stand-alone model Models Relative ratio of capital efficiency Electricity sector’s value-added Value added share in GDP (–) share in GDP (%) NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE 2030 2050 –1 –2 –1 –2 Sectors Output change Value-added_Output ratio Residual Fig. 5 Decomposition analysis of GDP changes across sectors. Value-added changes relative to baseline scenarios are expressed as percentages of GDP. Legend entries Output change, Value-added_output ratio, and Residual refer to output changes, value-added productivity changes, and residuals, respectively. The top and bottom panels show CGE stand-alone and integrated model results, respectively. Sectors are BIO Bioenergy industry, SER service sector, CCS CCS industries, TRS Transportation, IND manufacturing and construction, PWR power, OEN other energy supply, AGR agriculture, and FFE fossil fuel extraction Table 1 Diagnostic scenarios and their GDP loss rates in 2050. Energy supply Industry Service Transport Residential GDP loss rate (%) 2030 2050 1 off off off off off 1.1 2.4 2 off off off off on 0.9 2.3 3 off off off on off 1.1 2.4 4 off off on off off 0.6 1.7 5 off on off off off 0.4 0.8 6 on off off off off 0.9 2.2 7 off on on on on 0.1 0.2 8 on off on on on 0.5 2.2 9 on on off on on 0.4 1.2 10 on on on off on −0.1 0.6 11 on on on on off 0.1 0.8 12 on on on on on 0.0 0.8 Column names are sectors, and on and off refer to whether AIM/Enduse information is incorporated. The red and blue rows indicate the stand-alone and integrated models, respectively. Yellow and green rows indicate scenarios that include and exclude information from a single sector given by AIM/Enduse, respectively CGE models, the substitution elasticity of energy and value-added There can be a discussion on the parameter choices in the in these sectors use values referenced from the literature .This conventional CGE models and a question whether our results are representation has two possible disadvantages. First, historical price- robust to the key parameter assumptions. To this end, we con- induced energy and capital substitutability data are based on past ducted a sensitivity analysis, varying the elasticity substitution events and limited to developed countries. Future technological between energy and value-added from 0.2 to 0.8, taking the range availability, which is represented by the energy system model in this from the literature . The results showed that the cost differences study, may change drastically. Second, the elasticity parameter is associated with variation in the substitution elasticity parameter sometimes assumed to be uniform, but it should differ among are much smaller than the differences between the integrated and sectors, and probably regions (this study uses the global model’s stand-alone models (see Supplementary Fig. 4). This implies that uniform value for the stand-alone model). even if the wide range of values for the substitution elasticity NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 7 BIO SER CCS TRS IND PWR OEN AGR FFE BIO SER CCS TRS IND PWR OEN AGR FFE Change in % of GDP Integrated model CGE stand-alone ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 parameter (as seen historically in the literature) is considered, For now, this study’sapproachand theimplications future technological changes represented by the energy system thereof are applicable only to Japan, within the context of our model cannot be expressed. modelling framework. Application to other fields by different To represent the production functions, an alternative approach modelling teams is needed to demonstrate that our findings can to CES-type methods already exists in the econometric method . be generalised. In contrast to this approach, our method relies on realistic For future researches, as reported in the results section, some representation of technological availability. Therefore, we can variables show discrepancies between the two models in the base identify explicit technological changes that are consistent with the year. Although we think that this discrepancy does not affect our general equilibrium framework. Note that this process implicitly main conclusion, a more consistent understanding of this type of assumes that currently non-existent technologies are excluded, modelling framework is needed. This understanding may be whereas the conventional approach using possible substitution accomplished by calibrating both models, but such calibration could implicitly assume an infinite possibility to decrease energy will require substantial additional efforts to fully harmonise the consumption in response to energy price signals. base year data. Although this calibration is not expected to change GDP loss differences associated with the household sector’s our conclusions, it is a worthwhile endeavour for future research. representation in the conventional and integrated models were Another future potential research based on this modelling is that small, but we need to consider the disadvantages of measuring the hard-linkage among the models and in particular, electricity mitigation cost as GDP loss. Household expenditure is a major market is now highly demanded to investigate in terms of component of GDP in the expenditure accounting system, and intermittent supply of solar and wind power generation. increases in household expenditure directly boost GDP. Hence, purchasing relatively expensive energy devices such as electric Methods vehicles and heat pumps will not directly decrease GDP, but Overview of the method. Here, we developed an integrated modelling framework that incorporates energy system, power-dispatch, and CGE models, as illustrated in rather may offset the negative impacts of climate change miti- Supplementary Fig. 5. Each model’s output is exchanged with the others. We gation costs. Notably, this GDP increase is attributed to the executed five model iterations and assessed the second iteration because the dis- additional expenditure, which may not contribute to an increase crepancy improvements were sufficiently small at the second iteration. The cal- in actual welfare. This finding may show one of the limitations of culation begins with an AIM/Enduse run and then uses AIM/CGE and AIM/ Power. AIM/Enduse is run again, considering the AIM/CGE and AIM/Power accounting for climate mitigation costs using this type of model. outputs. The electricity demand and supply system under stringent emissions An energy system model simply represents the reduction reduction targets would be highly dependent on fluctuations in the electricity potential of energy-consuming devices, but numerous other supply and demand patterns, which requires operation on an hourly basis. possibilities exist to change the energy service itself. Artificial Therefore, we used AIM/Power in this model. We conducted scenario-based simulations through 2050. The individual models were solved from 2010 to 2050, intelligence may maintain energy devices more efficiently, or then the results from each were input to the other models. If models interact each transport demand could be reduced. Material consumption can other for each year, the convergence could be much faster since current approach also change through sharing of goods and services. From that can remain the gaps among the models each year, which can be amplified parti- perspective, the mitigation potential and associated cost may be cularly latter period. However, fortunately we have already had good convergences underestimated. Meanwhile, these societal changes could have with less iterations. The energy system and related CO emissions are the scope of this study, as Japanese GHG emissions are associated with these factors. In this indirect effects in the opposite direction in terms of energy study, we excluded the effect of climate change damage on the economy to avoid consumption, as information technology would require addi- complexity (e.g. isolating mitigation effects from the mixture of climate change tional electricity. The monetary savings realised by decreasing mitigation and damage impact, and additional assumptions on other countries’ energy usage could be spent on other things, and if it were spent emissions situations). The baseline socioeconomic assumptions are based on Shared Socioeconomic Pathways 2 described in Fujimori et al. . on energy-intensive activities (e.g. tourism using air travel), energy consumption and emissions could increase. A computable general equilibrium model. The CGE model used in this study is a The energy system model’s representation of technological recursive dynamic general equilibrium model that covers all regions of the world diffusion is based on linear programming with some constraints. 39–43 and is widely used in climate mitigation and impact studies . The main inputs Thus, this model may be interpreted as the extreme case where a for the model are socioeconomic assumptions of the drivers of GHG emissions single technology is selected at some point under certain price such as population, total factor productivity (TFP), which should reproduce the GDP assumptions in baseline scenarios, energy technology, and consumer pre- conditions, such as only electric vehicles being sold in a private ferences on diet. The production and consumption of all goods and GHG emis- car market. Meanwhile, the CES or logit formulations that are sions are the main outputs based on price equilibrium. The base year is the typically used in economic models allow multiple possibilities, year 2005. implicitly assuming heterogeneity in goods and consumers, One characteristic of our industrial classification is that energy sectors, whose real behaviour should be represented by a utility function including power sectors, are disaggregated in detail, because energy systems and their technological descriptions are crucial for the purposes of this study. that accounts for non-monetary value . This notion is important Moreover, to appropriately assess bioenergy and land-use competition, agricultural when interpreting household results derived from integrated sectors are highly disaggregated . Details of the model structure and its model results, where some may select economically irrational 45 mathematical formulas were provided by Fujimori et al. and wiki page . technologies and non-monetary factors are present. However, Production sectors are assumed to maximise profits under multi-nested constant elasticity substitution (CES) functions at each input price. Energy according to our results, industrial activities have more influence transformation sectors (Supplementary Table 5) input energy and are value- over mitigation cost and our conclusions would hold true if we added based on a fixed coefficient, whereas all energy end-use sectors included such heterogeneity. (Supplementary Table 6) have elasticities between energy and the value-added We achieved relatively fast convergence compared with existing (CES aggregation of capital and labor) amount. These sectors are treated in this manner to account for energy conversion efficiency in the energy transformation studies. There are two possible reasons for the rapid convergence. sectors. Power generation from several energy sources is combined using a logit First, on AIM/CGE side, the energy consumption is forced to be function ,althoughaCESfunctionisoften used in otherCGE models.We AIM/Enduse by endogenising parameters that are exogenous in the chose this method to represent energy balance because the CES function does 47 44 conventional CGE formula. Second, the major information pro- not guarantee a physical balance . As discussed by Fujimori, Hasegawa ,an energy or physical balance violation in the CES would not be critical if the power vided by AIM/CGE to AIM/Enduse that changes the AIM/Enduse generation shares of each technology in total power generation were similar to response is the energy service changes (output of sectors and total the calibrated information. The hydrogen production sectors have similar household consumption), but the difference from the previous structure as power generation. In this study, climate mitigation changes the iteration is less than 1%, which would not change AIM/Enduse power generation mix when compared to that of the base year, and therefore is a results in terms of carbon price or power generation. key treatment. The variable renewable energy cost assumption is shown in SI 8 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE section 2. Household expenditures on each commodity are described with a In addition, the demand-supply balance of electricity within an hour is modelled linear expenditure system (LES) function. The savings ratio is endogenously using the fluctuations and flexible range of each generator. Although generators determined to balance savings and investment, and capital formation for each and flexible resources are modelled in detail, electricity demands are provided item is determined using a fixed coefficient. The Armington assumption, which exogenously. Thus, the power-dispatch stand-alone model does not determine the assumes imperfect substitutability between domestically produced and traded total electricity consumption and installed capacity by technology, which are given goods , is used for trade, and the current account is assumed to be balanced. parameters. Note that there are buffers to deal with seasonal fluctuations, such as To construct energy supply cost curves, we implemented multiple sources of fossil fuel CCS thermal plants, in the mitigation scenarios, and thus, even if we information. Solar and wind supply curves are from a study considering urban consider battery storage for seasonal fluctuations, it would remain unused due to distance . Biomass potential and supply curve data is from a land-use allocation the cost competitiveness. The no CCS scenario also uses gas thermal plants to model . adjust for seasonal differences. From the energy system model to the economic model. The following infor- An energy system model. The energy system model used in this study is a mation is given to AIM/CGE from AIM/Enduse outputs. First, Change ratio of recursive dynamic partial equilibrium model based on detailed descriptions of final energy consumption by sector and energy type; second, power generation energy technologies in the end-use and supply sectors. In this study, we used the multi-region version of AIM/Enduse [Japan] , which divides Japan into 10 regions share by energy source; third, battery capacity for stabilising fluctuations of the power supply and its capacity factor, which is taken from AIM/Power (this capacity (see Supplementary Fig. 6) based on the power grid system. The model covers energy-related GHG emissions from both energy end-use and energy supply sec- factor means the total hours that the battery used divided by a year); forth, CCS installation; fifth, investment in energy end-use sectors; sixth, carbon prices; tors. The end-use sectors are composed of industry, buildings and transportation sectors, and they are disaggregated into several subsectors with respect to types of seventh transmission losses. Final energy consumption is classified into four sectors (industry, transport, products, buildings, and transportation mode based on the IEA energy balances. The CO emissions constraint is assumed for every simulation year of the AIM/ service and residential) and fed into the CGE model. We exogenously represent these sectors, while autonomous energy efficiency improvement (AEEI) parameters are Enduse model under the mitigation scenario. Within this study, the carbon price trajectory is almost exponential as a consequence. Therefore, even if we adopt an endogenised. This treatment maintains the same number of equations and variables as in the conventional CGE approach. To integrate household energy consumption inter-temporal optimisation scheme, it would not markedly affect the results. However, this might not be the case for other carbon constraints. Mitigation and energy device purchase activities in the household, we divided the household expenditure into four categories, such as car-use activities and other energy options are selected based on linear programming to minimize total energy system consumption activities, as illustrated in the Supplementary Fig. 7 (see more detailed costs that include investments for mitigation options and energy costs subject to information in Supplementary Note 2 and Supplementary Table 11). Because the exogenous parameters such as cost and efficiency of technology, primary energy absolute value of energy consumption is not fully harmonised between these two prices, energy service demands, and emission constraints. Detailed information on the model structure and parameter settings are provided in Kainuma et al. and a models, we compare the change ratios of energy consumption with 2010 levels, which is the base year of the AIM/Enduse model, for final energy consumption list of technologies is given in Supplementary Table 7. As the models used in this study were recursive dynamic, we did not consider discounting the energy system determination. If the corresponding energy consumption was zero or very low in 2010 (less than 1 ktoe), the change ratio can lead to unrealistic projections; therefore, we costs. Nevertheless, the AIM/Enduse model annualises the capital costs of energy technologies using a discount rate in the range 5–33% (Oshiro et al.) . The sec- use absolute values. The investment in energy end-use sectors is input as an incremental capital cost compared to the baseline case, where investment costs in the toral discount rate is 5% for power and industry, 10% for transportation, and 33% for other sectors. These individual discount rates are only applied to simulate baseline is modelled by CES substitution. Moreover, the capital input coefficients are fixed at baseline levels so that additional energy investment is represented by AIM/ technology selection in the energy system model. Consequently, the energy investment data fed into the economic model are not discounted by these rates. Enduse information rather than CES substitution elasticity in the mitigation scenarios. The power sector is modelled in detail, considering the balances of electricity supply and demand in 3-h steps to assess the impacts of variable renewable From the economic model to the energy system model. Because the sectoral energies (VREs). This sector also includes measures to integrate VREs into the disaggregation of AIM/Enduse basically complies with the IEA energy balance, grid, such as electricity storage, demand response (DR) using battery-powered there are inconsistencies in the AIM/CGE, which is based on an input-output table. electric vehicles and heat-pump devices, and interconnections. The total capacity Thus, in terms of data exchange from AIM/CGE to AIM/Enduse, the subsectors was calculated based on the capacity of newly installed power plants, which was are aggregated so that the granularity of the sectors is in agreement. Nevertheless, determined endogenously, as well as that of existing plants. In the AIM/Enduse given the large share of industrial GHG emissions in Japan’s long-term low-carbon model, the residual capacities of the existing power plants in operation in 2010 scenarios, iron, chemical, paper, non-metallic minerals, and non-ferrous metals are were calculated based on individual powerplant information, such as year exempted from the sector aggregation. AIM/Enduse uses the following information constructed, capacity of each plant, and expected lifetime. generated by AIM/CGE: first, GDP changes; second, household consumption In the industry, building and transportation sectors, wide mitigation options are changes; third, industry and service sector outputs; fourth energy price changes included, such as energy-efficient devices and fuel switching. The industrial sector Economic information from AIM/CGE is input into AIM/Enduse as changes in also includes innovative technologies such as carbon capture and storage (CCS). energy service demand for each sector. Transport demand is associated with GDP However, the AIM/Enduse stand-alone model does not account for some projection in AIM/Enduse and we proportionally change the transport demand based mitigation options that contribute to reduction in energy service demands. The key on changes in GDP. The energy service demand in the industrial sectors, such as steel power generation technoeconomic information is shown in Supplementary and cement production, and outputs of other industrial sectors, is altered by the Table 8. The cost information is based on METI data (2015) , as they are outputs from AIM/CGE. Energy service demand in the household and industrial consistent with the assumptions in Japan’s NDC. Note that the estimated sectors could have low or high elasticities to relevant economic activity variables, such mitigation cost may become much lower under more optimistic assumptions as household consumption and outputs of service sectors, but remains an uncertain regarding future cost reductions, especially for renewable energies. Moreover, factor. According to the Swedish econometric analysis ,elasticity between monetary powerplant information in 2010 and fuel assumptions are shown in Supplementary and physical units of energy services can be assumed to be ~1.0. This elasticity Table 9 and Supplementary Table 10. accounts for the percent change in physical energy services caused by a 1% change in monetary outputs. Furthermore, the GDP losses indicated in this study are relatively small, less than 3%, in the CGE stand-alone model, as shown in Fig. 4a. Thus, we A power-dispatch model. The power-dispatch model used in this study is a tentatively applied an elasticity value of 1.0. Meanwhile, we varied the elasticity from recursive dynamic partial equilibrium model focused on generation planning for 0.5 to 2.0 and observed that the policy costs change slightly, but the qualitative the power sector. In other words, unlike the AIM/Enduse model covering all conclusion still holds (Supplementary Fig. 8). energy-related sectors, the AIM/Power model only covers the power generation sector. This model can simulate hourly or annual electricity generation, generation capacity, plant locations, and multiple flexible resources, and includes interregional From the energy system model to the power-dispatch model. AIM/Power’s role transmission, dispatchable power, storage, and demand responses. These variables is to present the feasibility of power-dispatch given an electricity demand and were selected based on linear programming while minimising the total system installed power capacity. Thus, AIM/Enduse provides the following items to AIM/ costs, including capital costs, operation and maintenance costs, and fuel costs Power: first, electricity demand; second, power generation installed capacity; third, under several constraints, including satisfying electricity demand and CO emis- demand response technological availability, such as heat-pump water heaters and sion reduction targets. In this study, we used a version of the model that classifies electric vehicles Japan into 10 regions (see Supplementary Fig. 6). Detailed information about this model can be found in Shiraki et al. . Note that as AIM/Enduse provides power generation installed capacity for AIM/Power, AIM/Power does not make invest- From the power-dispatch model to the economic model. AIM/Power provides ment decisions, except for making additional investments in storage and power more realism in terms of technologies to stabilise short-term fluctuations in the plants aimed at hourly and within hourly power demand-supply management power system than the other two models used in this study. Moreover, the power AIM/Power can explicitly simulate the hourly demand-supply balance of system would respond to large-scale renewable energy installations by adjusting the electricity, with consideration of daily variations in photovoltaic output caused by capacity factor for conventional power generation systems (e.g. coal-fired power) in weather conditions as well as seasonal and weekday/weekend variations in demand. addition to curtailing the output from variable renewables. These measures for NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 balancing short-term fluctuations reduce the electricity output per installed capa- 4. Clarke, L. et al. Assessing Transformation Pathways. Climate Change 2014: city, and thus affect investment decisions. It is necessary to consider this feedback Mitigation of Climate Change. Contribution of Working Group III to the Fifth from AIM/Power to AIM/Enduse. In summary, the following AIM/Power infor- Assessment Report of the Intergovernmental Panel on Climate Change. mation is given to AIM/Enduse: first, battery capacity needed to stabilise short- (Cambridge University Press, Cambridge, United Kingdom and New York, term electricity fluctuations; second, Curtailment ratio; third, capacity factors. NY, USA, 2014). Note that generation capacity, although not directly related to balancing short- 5. McCollum, D. L. et al. Energy investment needs for fulfilling the Paris term fluctuations, is decided by AIM/Enduse. Thus, the battery capacity is Agreement and achieving the Sustainable Development Goals. Nat. Energy 3, determined by AIM/Enduse when considering long-term electricity fluctuations, 589–599 (2018). whereas that for short-term fluctuations is provided by AIM/Power. 6. Calvin, K. et al. The SSP4: a world of deepening inequality. Glob. Environ. Change 42, 284–296 (2017). 7. Fricko, O. et al. The marker quantification of the Shared Socioeconomic Convergence of iterations. We confirm fast convergence between the models. Pathway 2: a middle-of-the-road scenario for the 21st century. Glob. Environ. Detailed discussion related to the convergence is made by Supplementary Note 3 Change 42(Supplement C), 251–267 (2017). where Supplementary Figs. 9–20, Supplementary Table 12 and Supplementary Table 13 shows actual convergence situation. As stated, the discrepancy between 8. van Vuuren, D. P. et al. Energy, land-use and greenhouse gas emissions trajectories under a green growth paradigm. Glob. Environ. Change 42, the model almost reaches convergences in the second step. 237–250 (2017). 9. Strachan, N. & Kannan, R. Hybrid modelling of long-term carbon reduction Scenario assumptions. There are two basic assumptions for future scenarios, scenarios for the UK. Energy Econ. 30, 2947–2963 (2008). namely, baseline and mitigation, which are carried out with and without carbon 10. Turton, H. ECLIPSE: An integrated energy-economy model for climate policy pricing to reduce GHG emissions by 80% in 2050. Basic assumptions on techno- and scenario analysis. Energy 33, 1754–1769 (2008). logical conditions, such as nuclear scenarios and CCS capacities, are taken from 56 11. Jacoby, H. D., Reilly, J. M., McFarland, J. R. & Paltsev, S. Technology and previous studies . In the results section, we describe how the mitigation cost technical change in the MIT EPPA model. Energy Econ. 28, 610–631 (2006). differs from that in the stand-alone AIM/CGE run to identify each sector’s con- 12. Fujimori, S. et al. SSP3: AIM implementation of Shared Socioeconomic tribution to the changing mitigation costs. Pathways. Glob. Environ. Change 42, 268–283 (2017). 13. Markandya, A., Halsnaes, K. Costing methodologies. Guidance Papers on the Analytical method for the diagnostic scenario runs. To investigate the extent to Cross Cutting Issues of the Third Assessment Report of the IPCC.15–31 (Global which AIM/Enduse output information for each sector contributes to mitigation Industrial and Social Progress Research Institute (GISPRI), 2000). cost adjustment compared to the conventional CGE approach, we ran diagnostic 14. Waisman, H., Guivarch, C., Grazi, F. & Hourcade, J. C. The Imaclim-R model: scenarios with and without incorporation of AIM/Enduse data by sector, as noted infrastructures, technical inertia and the costs of low carbon futures under in the supplementary information. Ultimately, we conducted 32 scenarios with imperfect foresight. Climatic Change 114, 101–120 (2012). various combinations of AIM/Enduse information for energy supply, industry, 15. Lanzi, E., Chateau, J. & Dellink, R. Alternative approaches for levelling carbon service, transport, and residential sectors taken into account or excluded. The prices in a world with fragmented carbon markets. Energy Econ. 34, indicator shown below is adopted. i, t, and s are sets of variables (e.g. energy S240–S250 (2012). demand), years and scenarios, respectively. X and Y are AIM/CGE and AIM/ i,t,s i,t,s 16. Vandyck, T., Keramidas, K., Saveyn, B., Kitous, A. & Vrontisi, Z. A global Enduse outputs, respectively. stocktake of the Paris pledges: implications for energy systems and economy. vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Glob. Environ. Change 41(Supplement C), 46–63 (2016). u X  Y t t;i;s t;i;s 17. Tuladhar, S. D., Yuan, M., Bernstein, P., Montgomery, W. D. & Smith, A. A ErrInd ¼ hi i;s t X þY top–down bottom–up modeling approach to climate change policy analysis. t;i;s t;i;s t 2 Energy Econ. 31, S223–S234 (2009). 18. Arndt, C. et al. A sequential approach to integrated energy modeling in South Africa. Appl. Energy 161, 591–599 (2016). Reporting summary. Further information on research design is available in 19. Böhringer, C. & Rutherford, T. F. Combining bottom-up and top-down. the Nature Research Reporting Summary linked to this article. Energy Econ. 30, 574–596 (2008). 20. Böhringer, C. & Rutherford, T. F. Integrated assessment of energy policies: decomposing top-down and bottom-up. J. Economic Dyn. Control 33, Data availability 1648–1661 (2009). Scenario data are accessible online via HARVARD Dataverse (https://doi.org/10.7910/ 21. Hartwig, J., Kockat, J., Schade, W. & Braungardt, S. The macroeconomic DVN/QE6ERU). The data which are derived from the original scenario database shown effects of ambitious energy efficiency policy in Germany—Combining bottom- as figures but not in the above database is available upon requests. The source data up energy modelling with a non-equilibrium macroeconomic model. Energy underlying Figs. 1–5 and Supplementary Figs 1–20 are provided as a Source Data file. 124, 510–520 (2017). 22. Kriegler, E. et al. Fossil-fueled development (SSP5): an energy and resource Code availability intensive scenario for the 21st century. Glob. Environ. Change-Hum. Policy The source code for generating figures used in the main text and supplementary Dimens. 42, 297–315 (2017). information is available in HARVARD Dataverse (https://doi.org/10.7910/DVN/ 23. Jacobson, M. Z., Delucchi, M. A., Cameron, M. A. & Frew, B. A. Low-cost QE6ERU), which name is the Sourcecode.zip. The current code base of the AIM/Enduse, solution to the grid reliability problem with 100% penetration of intermittent AIM/CGE and AIM/Power developed over more than two decades at Kyoto University, wind, water, and solar for all purposes. Proc. Natl. Acad. Sci. USA 112, Shiga Prefecture University and National Institute for Environmental Studies. AIM/ 15060–15065 (2015). Enduse is available at AIM website (http://www-iam.nies.go.jp/aim/data_tools/index. 24. Zerrahn, A. & Schill, W.-P. Long-run power storage requirements for high html#enduse) and others are not available in a publicly shareable version. The code will shares of renewables: review and a new model. Renew. Sustain. Energy Rev. 79, continue to be developed and hosted by Kyoto University, Department of Environmental 1518–1534 (2017). Engineering (http://www.athehost.env.kyoto-u.ac.jp/). Requests for code should be 25. Santos-Alamillos, F. J., Archer, C. L., Noel, L., Budischak, C. & Facciolo, W. addressed to Shinichiro Fujimori. Assessing the economic feasibility of the gradual decarbonization of a large electric power system. J. Clean. Prod. 147, 130–141 (2017). 26. Pye, S., Li, F. G. N., Price, J. & Fais, B. Achieving net-zero emissions through Received: 7 November 2018; Accepted: 27 September 2019; the reframing of UK national targets in the post-Paris Agreement era. Nat. Energy 2, 17024 (2017). 27. Luderer, G. et al. Assessment of wind and solar power in global low-carbon energy scenarios: an introduction. Energy Econ. 64, 542–551 (2017). 28. Pietzcker, R. C. et al. System integration of wind and solar power in integrated assessment models: a cross-model evaluation of new approaches. Energy Econ. References 64, 583–599 (2017). 1. UNFCCC. FCCC/CP/2015/L.9/Rev.1: Adoption of the Paris Agreement. 29. Japan MoE. Overview of the Plan for Global Warming Countermeasures (UNFCCC, Paris, 2015). -Cabinet decision on May 13, 2016. 2016 https://www.env.go.jp/press/files/en/ 2. The White House. United States mid-century strategy for deep decarbonization. 676.pdf (2016). (United Nations Framework Convention on Climate Change, Washington, 30. Aldy, J. et al. Economic tools to promote transparency and comparability in DC, 2016). the Paris Agreement. Nat. Clim. Change 6, 1000 (2016). 3. Iyer, G. et al. Measuring progress from nationally determined contributions to 31. Oshiro, K., et al. Mid-century emission pathways in Japan associated with the mid-century strategies. Nat. Clim. Change 7, 871–874 (2017). global 2 °C goal: national and global models’ assessments based on carbon 10 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE budgets. Climatic Change 1–15 https://link.springer.com/article/10.1007/ 55. Shiraki, H., Ashina, S., Kameyama, Y., Hashimoto, S. & Fujita, T. Analysis of s10584-019-02490-x. (2019). optimal locations for power stations and their impact on industrial symbiosis 32. Sugiyama, M. et al. Japan’s long-term climate mitigation policy: multi-model planning under transition toward low-carbon power sector in Japan. J. Clean. assessment and sectoral challenges. Energy 167, 1120–1131 (2019). Prod. 114,81–94 (2016). 33. van der Mensbrugghe, D. The environmental impact and sustainability 56. Oshiro, K., Kainuma, M. & Masui, T. Implications of Japan’s 2030 target for applied general equilibrium (ENVISAGE) model. The World Bank, January long-term low emission pathways. Energy Policy 110, 581–587 (2017). 2008 http://siteresources.worldbank.org/INTPROSPECTS/Resources/334934- 1193838209522/Envisage7b.pdf (2008). Acknowledgements 34. Babiker, M. H., et al. The MIT Emissions Prediction and Policy Analysis S.F., K.O., and T.H. are supported by the Environment Research and Technology (EPPA) Model; Revisions, Sensitivities, and Comparisons of Results. Development Fund (2-1702, 2-1908) of the Environmental Restoration and Conservation (Cambridge, MA, 2001). Agency of Japan and JSPS KAKENHI Grant (No. 19H02273). S.F. is supported by 35. van der Werf, E. Production functions for climate policy modeling: an COMMIT (Climate pOlicy assessment and Mitigation Modeling to Integrate national empirical analysis. Energy Econ. 30, 2964–2979 (2008). and global Transition pathways) financed by Directorate General Climate Action (DG 36. Jorgenson, D. W. Econometric methods for modeling producer behavior. CLIMA) and EuropeAid under grant agreement No. 21020701/2017/770447/SER/ Handb. Econ. 3, 1841–1915 (1986). CLIMA.C.1 EuropeAid/138417/DH/SER/MulitOC. H.S. is supported by the Murata 37. McCollum, D. L. et al. Improving the behavioral realism of global integrated Science Foundation. assessment models: an application to consumers’ vehicle choices. Transportation Res. Part D Transp. Environ. 55, 322–342 (2017). 38. Fujimori, S. et al. SSP3: AIM implementation of Shared Socioeconomic Author contributions Pathways. Glob. Environ. Change-Hum. Policy Dimens. 42,268–283 (2017). S.F. designed the research; S.F carried out analysis of the modelling results; S.F. and H.S. 39. Hasegawa, T. et al. Climate change impact and adaptation assessment on food created figures; S.F. wrote the draft of the paper; S.F., K.O. and H.S. provided economic, consumption utilizing a new scenario framework. Environ. Sci. Technol. 48, energy system, and power model data; and S.F., K.O., H.S., and T.H. contributed to the 438–445 (2014). discussion and interpretation of the results. 40. Hasegawa, T. et al. Consequence of climate mitigation on the risk of hunger. Environ. Sci. Technol. 49, 7245–7253 (2015). 41. Mittal S., Dai H., Fujimori S., Masui T. Bridging greenhouse gas emissions and Competing interests renewable energy deployment target: comparative assessment of China and The authors declare no competing interests. India. Appl. Energy 166, 301–313 (2016). 42. Hasegawa, T., Fujimori, S., Takahashi, K., Yokohata, T. & Masui, T. Economic Additional information implications of climate change impacts on human health through Supplementary information is available for this paper at https://doi.org/10.1038/s41467- undernourishment. Climatic Change 136,1–14 (2016). 019-12730-4. 43. Fujimori, S., Kainuma, M., Masui, T., Hasegawa, T. & Dai, H. The effectiveness of energy service demand reduction: a scenario analysis of global Correspondence and requests for materials should be addressed to S.F. climate change mitigation. Energy Policy 75(Supplement C), 379–391 (2014). 44. Fujimori, S., Hasegawa, T., Masui, T. & Takahashi, K. Land use representation Peer review information Nature Communications thanks Per Ivar Helgesen, Ronald in a global CGE model for long-term simulation: CET vs. logit functions. Food Sands and the other, anonymous, reviewer(s) for their contribution to the peer review of Sec 6, 685–699 (2014). this work. Peer reviewer reports are available. 45. IAMC. The common Integrated Assessment Model (IAM) documentation. https://www.iamcdocumentation.eu/index.php/IAMC_wiki (2018). Reprints and permission information is available at http://www.nature.com/reprints 46. Clarke, J. F. & Edmonds, J. A. Modelling energy technologies in a competitive market. Energy Econ. 15, 123–129 (1993). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 47. Schumacher, K. & Sands, R. D. Innovative energy technologies and climate published maps and institutional affiliations. policy in Germany. Energy Policy 34, 3929–3941 (2006). 48. Armington, S. P. A theory of demand for products distinguished by place of production. Staff Pap. 16, 159–178 (1969). Open Access This article is licensed under a Creative Commons 49. Herran, D. S., Dai, H., Fujimori, S. & Masui, T. Global assessment of onshore wind power resources considering the distance to urban areas. Energy Policy Attribution 4.0 International License, which permits use, sharing, 91,75–86 (2016). adaptation, distribution and reproduction in any medium or format, as long as you give 50. Hasegawa, T., Fujimori, S., Ito, A., Takahashi, K. & Masui, T. Global land-use appropriate credit to the original author(s) and the source, provide a link to the Creative allocation model linked to an integrated assessment model. Sci. Total Environ. Commons license, and indicate if changes were made. The images or other third party 580, 787–796 (2017). material in this article are included in the article’s Creative Commons license, unless 51. Oshiro, K. & Masui, T. Diffusion of low emission vehicles and their impact on indicated otherwise in a credit line to the material. If material is not included in the CO2 emission reduction in Japan. Energy Policy 81, 215–225 (2015). article’s Creative Commons license and your intended use is not permitted by statutory 52. Kainuma, M. M. Y. & Morita, T. Climate policy assessment: Asia-Pacific regulation or exceeds the permitted use, you will need to obtain permission directly from integrated modeling. (Springer, Japan, 2003). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 53. Oshiro, K., Kainuma, M. & Masui, T. Assessing decarbonization pathways and licenses/by/4.0/. their implications for energy security policies in Japan. Clim. Policy 16(sup1), S63–S77 (2016). 54. METI. Report on analysis of generation costs, etc. for subcommittee on long- © The Author(s) 2019 term energy supply and demand outlook. (Ministry of Economy, Trade and Industry, 2015). NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 11 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nature Communications Springer Journals

Energy transformation cost for the Japanese mid-century strategy

Loading next page...
 
/lp/springer-journals/energy-transformation-cost-for-the-japanese-mid-century-strategy-0K6vf4yt58

References (63)

Publisher
Springer Journals
Copyright
Copyright © 2019 by The Author(s)
Subject
Science, Humanities and Social Sciences, multidisciplinary; Science, Humanities and Social Sciences, multidisciplinary; Science, multidisciplinary
eISSN
2041-1723
DOI
10.1038/s41467-019-12730-4
Publisher site
See Article on Publisher Site

Abstract

ARTICLE https://doi.org/10.1038/s41467-019-12730-4 OPEN Energy transformation cost for the Japanese mid-century strategy 1,2,3 1 4 2,3,5 Shinichiro Fujimori *, Ken Oshiro , Hiroto Shiraki & Tomoko Hasegawa The costs of climate change mitigation policy are one of the main concerns in decarbonizing the economy. The macroeconomic and sectoral implications of policy interventions are typically estimated by economic models, which tend be higher than the additional energy system costs projected by energy system models. Here, we show the extent to which policy costs can be lower than those from conventional economic models by integrating an energy system and an economic model, applying Japan’s mid-century climate mitigation target. The GDP losses estimated with the integrated model were significantly lower than those in the conventional economic model by more than 50% in 2050. The representation of industry and service sector energy consumption is the main factor causing these differences. Our findings suggest that this type of integrated approach would contribute new insights by providing improved estimates of GDP losses, which can be critical information for setting national climate policies. 1 2 Department of Environmental Engineering, Kyoto University, C1-3 361, Kyotodaigaku Katsura, Nishikyoku, Kyoto city, Japan. Center for Social and Environmental Systems Research, National Institute for Environmental Studies (NIES), 16–2 Onogawa, Tsukuba, Ibaraki 305–8506, Japan. International Institute for Applied System Analysis (IIASA), Schlossplatz 1, A-2361 Laxenburg, Austria. School of Environmental Science, The University of Shiga Prefecture, Hikone, Japan. Department of Civil and Environmental Engineering, College of Science and Engineering, Ritsumeikan University, Kyoto, Japan. *email: sfujimori@athehost.env.kyoto-u.ac.jp NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 1 1234567890():,; ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 limate change mitigation is one of the greatest societal consistently dealing with energy systems and the stability of challenges facing most countries as reduction of energy- power generation. Crelated CO emissions is key to reducing greenhouse gas Here, we describe the macroeconomic implications of climate (GHG) emissions. In 2015, more than 190 countries reached the mitigation policy using an integrated modelling framework Paris Agreement (PA) and each country submitted their own wherein an energy system model, Asian-Pacific Integrated Model/ Nationally Determined Contribution (NDC). Along with those Enduse (AIM/Enduse), and a power-dispatch model, AIM/Power, targets, countries were also asked to engage in long-term plan- are inter-linked with the multi-sector economic model (AIM/ 2,3 ning, known as a mid-century strategy . Under the long-term CGE). We call this new soft-linking modelling framework an global goal in the PA of keeping the global mean temperature integrated model, which allows us to assess the macroeconomic increase well below 2 °C compared with the pre-industrial level, impacts of climate change mitigation with concrete specification the net CO emissions in this mid-century must be close to of detailed energy technologies, ensuring a stable power supply neutral according to numerous studies carried out using Inte- with consideration of long-term (seasonal and daily) and short- grated Assessment Models (IAMs) . term (less than hourly) power fluctuations. Macroeconomic costs of climate change mitigation is a great The principle of this methodology is based on the concept that concern for climate policy settings . The Intergovernmental Panel energy simulation from the energy system model is more reliable on Climate Change (IPCC) fifth assessment report summarises than that from the economic model, as energy supply and climate mitigation costs, and GDP or consumption losses in 2050 demand are technologically represented in detailed in the energy are around 2–6% to achieve the abovementioned 2 °C goal. system model. Similarly, the technological representation of There are multiple ways to interpret these numbers. It may be too power supply in the power-dispatch model is more reliable than expensive to pay for climate change prevention that delays GDP that in the energy system model. We overcome the disadvantages growth for a couple of years or low enough for avoiding wide- of these models by exchanging information and iterating it spread climate change impacts and irreversible risks associated among models. We begin with the AIM/Enduse run, which with catastrophic events. To address macroeconomic mitigation provides energy system information to AIM/CGE and AIM/ costs, IAMs normally represent GHG emissions reduction costs Power. Then, these two models’ outcomes are further fed into either through an energy system model or an economic model, AIM/Enduse. Finally, we confirm whether the models reach often termed bottom-up and top-down models, respectively. sufficient convergence for our purposes (see Supplementary Although there are other ways to classify the IAMs, in this Information for more detailed discussion about reaching con- paper, we define economic model as the model that includes vergence). See the Methods for indicators exchanged among multi-sectoral CGE model within the IAM framework, and models. Note that for CGE results, we compare the stand-alone energy system model as the model that does not. Note that a CGE model with the integrated model. power-dispatch model is also used in this study although that We applied this framework to Japan as a case study. The 6–8 is not usually classified as IAMs. There are many global , and Japanese government has declared a long-term GHG emissions 9,10 29 national energy system models as well as the economic reduction target of 80% by 2050 . As mitigation costs in Japan 11,12 30–32 models , which are based on multi-sectoral CGE models. estimated in previous studies vary significantly across IAMs , Traditionally, CGE models tend to project higher policy costs application of this framework would be beneficial for Japan’s than those of energy system models (see also Supplementary climate policies to communicate with the stakeholders. We ana- Note 1 and Supplementary Table 1). One possible reason for this lysed scenarios with and without climate mitigation policy, which tendency is that parameters in CGE models are calibrated against are the mitigation and baseline scenarios, respectively. a historical period in which it is difficult to decouple economic As results, we found that the macroeconomic costs are not as growth and CO emissions. Some argue that aggregated energy high as previously reported when energy system information is system representation is disadvantageous to understanding appropriately reflected in the economic model. The critical drastic energy system changes and their macroeconomic impli- determinants of mitigation costs that changed in the newly cations. Thus, incorporating energy system model information developed integrated model were identified as the representation into CGE models may lead lower macroeconomic costs than of industry and service sectors’ energy consumption, which is previously reported. associated with production functions. These findings may change Integrating CGE and energy system model offers a great the general perception of climate change mitigation costs in terms advantage in representing the feedbacks inherent across economic of macroeconomic losses and provide important policy insights. and energy systems. For the policy makers, macroeconomic implications including sectoral impacts provided by CGE models is more meaningful than energy system costs alone. To this end, Results 14–16 several attempts have been made , whereas investigators such Energy system in Japan’s mid-century strategy. An 80% 17–20 as Bohringer et al. incorporated disaggregated information reduction of GHG emissions requires substantial changes in the on power sectors. An extended literature list is shown in Sup- energy system compared to the current system or the baseline plementary Table 2 and there are more examples if we include scenario (Fig. 1a). As a result of Japan’s unique socioeconomic 21,22 non-multi-sector CGE models . circumstances, with a decreasing population and modest eco- At the meantime, drastic energy transformation requires large- nomic growth (Supplementary Fig. 1), the overall energy system scale variable renewable energy penetration. The key issue of the shows little changes in the future under the baseline scenario. The variability in renewable energy is strongly dependent on national- main changes of the baseline 2050 from the base year are the and local-scale grid systems, availability of solar and wind power, higher share of coal relative to other fossil fuels, and the decrease battery technology, and other energy sources that can be used to in the share of nuclear energy, which reflects the current societal balance demand and supply. Recently, some national modelling attitude toward nuclear power that limits new construction 23–25 studies have addressed these issues and integration of a (Fig. 1b). Regarding CO emissions, the baseline level is stable or power-dispatch model with an energy system model has been may even decline over time (Fig. 1d). Meanwhile, the mitigation 26 27,28 attempted . In IAMs, they are represented to some degree , scenario exhibits large-scale renewable energy penetration, slight which are adequate to provide global-scale energy analyses. energy demand reduction, compositional changes characterised However, no studies showed macroeconomic implications of by the use of more carbon-neutral energy sources, and 2 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE 20 1250 ab c d 0 0 0 ef g 0 0 0 Coal|w/o CCS Nuclear Wind Other Heat Solid|Biomass Liquid|Oil Baseline Mitigation Hydro Oil|w/o CCS Biomass|w/o CCS Gas Solid|Coal Solar Geothermal Gas|w/o CCS Electricity Liquid|Biomass Fig. 1 Main energy and emissions related variables. Primary energy source (a, e), power generation (b, f), final energy demand (c, g), CO emissions (d), and carbon price (h) projections. a, b, and c show the baseline scenario, whereas e, f, and g show mitigation scenarios. The other in a and e includes secondary energy net imports electrification (Fig. 1f). The price of carbon in the mitigation namely, without nuclear and without carbon capture and storage scenario increases over time and reaches ~1000$/tCO in 2050, (CCS). These results can be interpreted as a simple uncertainty which is due to the low-carbon technological availability in the analysis, but they have more meaningful policy implications AIM/Enduse model. because the perception of nuclear power in Japan has changed The power system relying heavily on variable renewable energy drastically since the Fukushima Daiichi accident, and there is requires measures to stabilise the power supply system and limited geologically appropriate space for CCS on Japanese demand responses. Curtailment in onshore wind increases, territory. Figure 3c illustrates the relationship of mitigation costs particularly after 2020 when variable renewables start to expand in the CGE stand-alone and integrated models for this sensitivity (Fig. 2a). Furthermore, when coal-fired power is completely analysis. Here, we again see systematically higher costs in the phased out around 2040, offshore wind also exhibits a clear stand-alone model than in the integrated model. Comparison of curtailment increase. The battery requirements for short-term these integrated model’s GDP losses and additional energy system fluctuations also increase sharply after 2020, whereas the capacity costs derived from AIM/Enduse shows a similar trend to that factor of thermal power plants declines (Fig. 2b, c). We also show in Fig. 3d. The overall energy and emissions trends for the daily electricity supply and demand profiles for selected days this sensitivity cases are provided in Supplementary Fig. 2 and in 2050 (Fig. 2d). Supplementary Fig. 3. Mitigation costs. Mitigation costs, as measured by GDP loss rates Mechanism causing the differences in macroeconomic costs. (hereafter GDP is accounted by the total final consumption), The central mechanisms for changing the macroeconomic increase over time as emissions reductions become deeper, as implications are changes in the productivity of primary factors illustrated in Fig. 3a. The CGE stand-alone results reach more (labour and capital) constituting value-added, which is the GDP than 2.5% after 2030, whereas the integrated model is lower, measure in production side. This is because the primary factor around 1.2% in 2050 (Fig. 3a). The equivalent variation also inputs are constrained exogenously for each year in our economic 33,34 shows similar trend as GDP losses (Fig. 3b). The additional model while the capital and labour inputs change dynami- energy system costs in the AIM/Enduse stand-alone are plotted in cally with population development, and GDP growth. The the same figure, and are notably similar to the integrated model straight-forward reason that GDP losses are lower in the inte- results (blue lines in Fig. 3a). The mitigation costs under such grated model than in the stand-alone model is that the parameter deep emissions reductions from CGE studies are usually not as assumptions in CGE models differ between the stand-alone low as our estimates (2–6% of GDP losses in 2050) . Once the model and integrated model. The former relies on the existing energy system model’s results are reflected in the economic literature and the latter on the energy system model outputs. model, the integrated model would be able to estimate similar Consequently, the primary factor productivity is higher in the mitigation costs to those from energy system models. integrated model than in the stand-alone model. Then, the dif- We further implemented sensitivity scenarios with varying ferences in productivity are mainly driven by two things. One is technological availability, which may lead to non-linear energy the productivity decreases associated with emissions reductions in system responses, to investigate the robustness of our findings. energy end-use sectors, such as industry, transport and service For this purpose, we selected two technological variation sectors (e.g. capital replacement by expensive but energy-efficient scenarios wherein more power stability measures are needed; ones). The other is sectoral allocation changes in primary factors. NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 3 Primary energy (EJ/year) Power generation (EJ/year) Final energy (EJ/year) Mt CO /yr US$2005/t CO 2 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ab Solar|PV Pumped hydro Wind|Offshore Battery for long-term Wind|Onshore Battery for short-term 0 0 Year Year c d Summer Intermediate Summer weekday weekend weekday PV_H PV_H PV_L Thermal power total Charge Gas Pumping Biomass Discharge Coal Biomass|w/o CCS Pumped hydro Hydro Coal|w/ CCS PV Nuclear Coal|w/o CCS Wind Loss Gas|w/ CCS Gas|w/o CCS Demand_w/_HP&EV Demand_w/o_HP&EV –50 Hour Year Fig. 2 Power system reactions to large-scale renewable energy penetration. a Wind and solar curtailment rate, which is the unused energy divided by total power generation for each energy source, b the installed capacity of technologies to stabilise fluctuations in the electricity supply, c the capacity factor of thermal power, and d the profiles of electricity demand and supply on selected three typical days (PV_H and PV_L indicate sunny and cloudy days, respectively) Regarding the first factor, Fig. 4a illustrates the capital The second factor, namely the effect of sectoral primary factor input efficiency of major industrial sectors (top 10 industries, allocation changes, is mainly driven by the power generation sector. which account for 95% of GDP in the base year) in the The electricity generation in the mitigation scenario compared to mitigation scenario compared to the baseline scenario for the baseline scenario is about 20% higher in the stand-alone model, stand-alone and integrated models in 2050. Here, we define the but almost the same in the integrated model (Fig. 4). There are capital input efficiency as capital input per output for each certainly differences in technological shares between the stand-alone sector, which is a model outcome. Higher values indicate and integrated models, but, in summary, it seems that the difference that additional capital inputs are needed in the mitigation in total electricity generation between the models is the dominant scenario compared to the baseline scenario. In general, the factor, where the stand-alone model requires additional capital and stand-alone model requires larger capital inputs than the labour inputs, accounting for 0.4 percentage points of GDP, relative integrated model in the mitigation scenario. We can roughly to the integrated model, which relies on the AIM/Enduse outputs compute the value-added losses associated with these capital (Fig. 4d). With respect to the representation of electricity demand, productivity losses by multiplying the value-added of each the total electricity demand is determined by energy consumption sector Fig. 4b, c). These eventually account for 1.3 percentage in the energy end-use sectors, which are represented by a CES points of the total value-added (GDP). Then, the productivity function, as mentioned above. The fuel-wise share is determined differences between the stand-alone and integrated model are using a logit function in both the stand-alone and integrated mainly caused by differences in the functional form and models. A parameter representing the preferences or technological parameters particular to the value-added and energy bundle. choices in the logit function is determined endogenously in the Here, we use a CES function in which the substitution elasticity, integrated model, based on the AIM/Enduse results, whereas they share parameters and future autonomous energy efficiency are are exogenous parameters in the stand-alone model. We describe defined in the stand-alone model. The integrated model the detailed mathematical formation and assumptions in the uses a function of the same form, but the additional investment Supporting Information. and energy inputs are exogenously given by AIM/Enduse, In addition to the two main mechanisms mentioned above, the whereas the CES shift parameters are determined endogenously productivity changes and sectoral shifts in other sectors certainly (sector-wise additional investments are shown in Supplemen- occur, but are relatively minor. In summary, the differences in tary Table 3). GDP changes between the stand-alone and integrated models are 4 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 2040 2020 % % GW GW NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE a b 4 4 3 3 2 2 0 0 Year Year CGE stand-alone CGE stand-alone Integrated model Integrated model Enduse stand-alone cd 4 1.5 1.0 0.5 0 0.0 CGE stand-alone (%/year) Energy system model (%/year) Default Default No CCS No CCS No nuclear No nuclear Fig. 3 Climate change mitigation cost. a, b Time-series mitigation cost AIM/CGE results are represented as GDP loss rates and equivalent variation change rates relative to baseline scenarios. AIM/Enduse results are expressed as additional energy system costs of GDP relative to baseline scenarios. c, d 5-year mitigation costs with varying technological availability; c illustrates the relationship of GDP losses in the CGE stand-alone and integrated models, and d shows GDP losses in the integrated model and additional energy system costs in AIM/Enduse. The energy system model results shown here correspond to Enduse_results1 in Supplementary Fig. 5 explained above, but, generally speaking, many interactions model runs in Fig. 5. The stand-alone CGE model shows simultaneously occur in the CGE model and sometimes the remarkable value-added decreases in the industry (IND) and cause and consequences are not clear. service sectors (SER) in 2030, whereas the integrated model does not. These trends remained consistent for the year 2050, with the CGE stand-alone model showing large changes in the service Decomposition of mitigation costs and sectoral contributions. sector. This result is consistent with those described in the pre- vious section, wherein the industry and service sector’s energy To identify which sectors contribute to GDP losses, the value- added by each sector, as estimated by the economic model, is system information, i.e. the representation of production func- decomposed into three factors of output changes, value-added tions in those sectors, are critical factors for differentiating overall productivity (output per value-added), and residuals. Moreover, GDP losses between the two models. The output decrease in the we compared the outputs of stand-alone CGE and integrated service sector is the largest element to change the GDP in the NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 5 0.0 0.5 1.0 1.5 Integrated model (%/year) GDP loss rates (%/year) Integrated model (%/year) Equivalent variation (%/year) ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ab 1.2 0.6 1.1 0.4 CGE stand-alone 1.0 Integrated model 0.2 0.9 0.0 cd Sectors Sectors Geothermal 01 Other services Biomass 02 Other manufacturing Wind Stand-alone model 03 Transport and communications Solar 04 Construction Hydro 05 Chemical, plastic and rubber products Nuclear 06 Food products Gas 07 Paper, paper products and pulp Oil 08 Iron and steel Coal 09 Non-ferrous products Integrated model 10 Light industry 11 Others Contribution of capital productivity change to value added changes (% of GDP) Fig. 4 Valued-added differences between baseline and mitigation scenarios. a Sectoral capital input efficiency for the top 10 industrial activities in 2050. Capital efficiencies in baseline and mitigation scenarios are computed and then the capital efficiency in the mitigation scenario relative to the baseline scenario is shown. b Sectoral value-added share in the baseline scenario for the top 10 industrial activities. c The capital productivity value-added effects compared to the total value-added for the top 10 industrial activities. Dots means net total changes. Negative and positive values mean capital productivity gain and losses compared to baseline scenarios respectively. d The value-added share of power sectors in terms of total economy-wide value-added CGE stand-alone model. This result may be driven by changes in overall insights are clear, that the industry and service sectors are household expenditures for services, which were around 3.4 and key in determining macroeconomic implications. 0.0% in the CGE stand-alone and integrated models, respectively, in 2050. These differences may be due to changes in total income. We ran further diagnostic scenarios with and without Discussion incorporating energy system information by sectors (see Methods Our newly proposed integrated model approach implicitly for more details) to investigate the extent to which the energy assumes that the energy productivity in the CGE model is system model’s output information for each sector contributes to endogenized by using the energy system model information. This mitigation cost differences compared to the stand-alone CGE. treatment is somewhat different from the conventional approach, Comparing scenarios that include a single sector’s information in which CGE models use the same Autonomous Energy Effi- from AIM/Enduse and the stand-alone model (Row 1–6 ciency Improvement (AEEI) and constant elasticity substitution in Table 1, respectively), the inclusion of the industry and service parameters, with and without mitigation policies. Based on the sector information from AIM/Enduse makes a remarkable results showing that the macroeconomic costs associated with difference in the GDP loss rate (Row 5 and 4 in Table 1, climate change mitigation policies are lower than estimated using respectively). From the opposite side, the scenarios taking out the conventional approaches, we can interpret the energy pro- AIM/Enduse information for each sector (Row 7–11 in Table 1) ductivities in the mitigation scenarios as being higher than in the show that excluding the industry and service sectors consistently conventional approach. This would imply that the AIM/Enduse generates GDP loss differences compared to the integrated model model incorporates higher productivity technological information (Row 12 in Table 1). Conversely, the incorporation of residential, than the conventional CES approach. transport and energy supply sector information given by AIM/ Overall, as long as an energy system model is more reliable than Enduse has a small impact on GDP losses, or even has the the CGE model in terms of energy-related variables, the energy opposite effect in some cases. Finally, we can see cross-sectional representation in the conventional CGE should be replaced by the effects in other scenarios in Supplementary Table 4, which energy system model outputs. The contributions of the industry and indicates the complexity of the results and shows that the service sectors to GDP loss differences are caused by the production influence of each sectoral impact is not additive. However, the function form and its parameters. Basically, for most conventional 6 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 01 Other services 02 Other manufacturing 03 Transport and communications 04 Construction 05 Chemical, plastic and rubber products 06 Food products 07 Paper, paper products and pulp 08 Iron and steel 09 Non-ferrous products 10 Light industry 01 Other services 0.0 02 Other manufacturing 0.5 03 Transport and communications 1.0 04 Construction 1.5 05 Chemical, plastic and rubber products 06 Food products 07 Paper, paper products and pulp 08 Iron and steel 09 Non-ferrous products 10 Light industry Baseline.Integrated model Mitigation.Integrated model Baseline.Stand-alone model Mitigation.Stand-alone model Models Relative ratio of capital efficiency Electricity sector’s value-added Value added share in GDP (–) share in GDP (%) NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE 2030 2050 –1 –2 –1 –2 Sectors Output change Value-added_Output ratio Residual Fig. 5 Decomposition analysis of GDP changes across sectors. Value-added changes relative to baseline scenarios are expressed as percentages of GDP. Legend entries Output change, Value-added_output ratio, and Residual refer to output changes, value-added productivity changes, and residuals, respectively. The top and bottom panels show CGE stand-alone and integrated model results, respectively. Sectors are BIO Bioenergy industry, SER service sector, CCS CCS industries, TRS Transportation, IND manufacturing and construction, PWR power, OEN other energy supply, AGR agriculture, and FFE fossil fuel extraction Table 1 Diagnostic scenarios and their GDP loss rates in 2050. Energy supply Industry Service Transport Residential GDP loss rate (%) 2030 2050 1 off off off off off 1.1 2.4 2 off off off off on 0.9 2.3 3 off off off on off 1.1 2.4 4 off off on off off 0.6 1.7 5 off on off off off 0.4 0.8 6 on off off off off 0.9 2.2 7 off on on on on 0.1 0.2 8 on off on on on 0.5 2.2 9 on on off on on 0.4 1.2 10 on on on off on −0.1 0.6 11 on on on on off 0.1 0.8 12 on on on on on 0.0 0.8 Column names are sectors, and on and off refer to whether AIM/Enduse information is incorporated. The red and blue rows indicate the stand-alone and integrated models, respectively. Yellow and green rows indicate scenarios that include and exclude information from a single sector given by AIM/Enduse, respectively CGE models, the substitution elasticity of energy and value-added There can be a discussion on the parameter choices in the in these sectors use values referenced from the literature .This conventional CGE models and a question whether our results are representation has two possible disadvantages. First, historical price- robust to the key parameter assumptions. To this end, we con- induced energy and capital substitutability data are based on past ducted a sensitivity analysis, varying the elasticity substitution events and limited to developed countries. Future technological between energy and value-added from 0.2 to 0.8, taking the range availability, which is represented by the energy system model in this from the literature . The results showed that the cost differences study, may change drastically. Second, the elasticity parameter is associated with variation in the substitution elasticity parameter sometimes assumed to be uniform, but it should differ among are much smaller than the differences between the integrated and sectors, and probably regions (this study uses the global model’s stand-alone models (see Supplementary Fig. 4). This implies that uniform value for the stand-alone model). even if the wide range of values for the substitution elasticity NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 7 BIO SER CCS TRS IND PWR OEN AGR FFE BIO SER CCS TRS IND PWR OEN AGR FFE Change in % of GDP Integrated model CGE stand-alone ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 parameter (as seen historically in the literature) is considered, For now, this study’sapproachand theimplications future technological changes represented by the energy system thereof are applicable only to Japan, within the context of our model cannot be expressed. modelling framework. Application to other fields by different To represent the production functions, an alternative approach modelling teams is needed to demonstrate that our findings can to CES-type methods already exists in the econometric method . be generalised. In contrast to this approach, our method relies on realistic For future researches, as reported in the results section, some representation of technological availability. Therefore, we can variables show discrepancies between the two models in the base identify explicit technological changes that are consistent with the year. Although we think that this discrepancy does not affect our general equilibrium framework. Note that this process implicitly main conclusion, a more consistent understanding of this type of assumes that currently non-existent technologies are excluded, modelling framework is needed. This understanding may be whereas the conventional approach using possible substitution accomplished by calibrating both models, but such calibration could implicitly assume an infinite possibility to decrease energy will require substantial additional efforts to fully harmonise the consumption in response to energy price signals. base year data. Although this calibration is not expected to change GDP loss differences associated with the household sector’s our conclusions, it is a worthwhile endeavour for future research. representation in the conventional and integrated models were Another future potential research based on this modelling is that small, but we need to consider the disadvantages of measuring the hard-linkage among the models and in particular, electricity mitigation cost as GDP loss. Household expenditure is a major market is now highly demanded to investigate in terms of component of GDP in the expenditure accounting system, and intermittent supply of solar and wind power generation. increases in household expenditure directly boost GDP. Hence, purchasing relatively expensive energy devices such as electric Methods vehicles and heat pumps will not directly decrease GDP, but Overview of the method. Here, we developed an integrated modelling framework that incorporates energy system, power-dispatch, and CGE models, as illustrated in rather may offset the negative impacts of climate change miti- Supplementary Fig. 5. Each model’s output is exchanged with the others. We gation costs. Notably, this GDP increase is attributed to the executed five model iterations and assessed the second iteration because the dis- additional expenditure, which may not contribute to an increase crepancy improvements were sufficiently small at the second iteration. The cal- in actual welfare. This finding may show one of the limitations of culation begins with an AIM/Enduse run and then uses AIM/CGE and AIM/ Power. AIM/Enduse is run again, considering the AIM/CGE and AIM/Power accounting for climate mitigation costs using this type of model. outputs. The electricity demand and supply system under stringent emissions An energy system model simply represents the reduction reduction targets would be highly dependent on fluctuations in the electricity potential of energy-consuming devices, but numerous other supply and demand patterns, which requires operation on an hourly basis. possibilities exist to change the energy service itself. Artificial Therefore, we used AIM/Power in this model. We conducted scenario-based simulations through 2050. The individual models were solved from 2010 to 2050, intelligence may maintain energy devices more efficiently, or then the results from each were input to the other models. If models interact each transport demand could be reduced. Material consumption can other for each year, the convergence could be much faster since current approach also change through sharing of goods and services. From that can remain the gaps among the models each year, which can be amplified parti- perspective, the mitigation potential and associated cost may be cularly latter period. However, fortunately we have already had good convergences underestimated. Meanwhile, these societal changes could have with less iterations. The energy system and related CO emissions are the scope of this study, as Japanese GHG emissions are associated with these factors. In this indirect effects in the opposite direction in terms of energy study, we excluded the effect of climate change damage on the economy to avoid consumption, as information technology would require addi- complexity (e.g. isolating mitigation effects from the mixture of climate change tional electricity. The monetary savings realised by decreasing mitigation and damage impact, and additional assumptions on other countries’ energy usage could be spent on other things, and if it were spent emissions situations). The baseline socioeconomic assumptions are based on Shared Socioeconomic Pathways 2 described in Fujimori et al. . on energy-intensive activities (e.g. tourism using air travel), energy consumption and emissions could increase. A computable general equilibrium model. The CGE model used in this study is a The energy system model’s representation of technological recursive dynamic general equilibrium model that covers all regions of the world diffusion is based on linear programming with some constraints. 39–43 and is widely used in climate mitigation and impact studies . The main inputs Thus, this model may be interpreted as the extreme case where a for the model are socioeconomic assumptions of the drivers of GHG emissions single technology is selected at some point under certain price such as population, total factor productivity (TFP), which should reproduce the GDP assumptions in baseline scenarios, energy technology, and consumer pre- conditions, such as only electric vehicles being sold in a private ferences on diet. The production and consumption of all goods and GHG emis- car market. Meanwhile, the CES or logit formulations that are sions are the main outputs based on price equilibrium. The base year is the typically used in economic models allow multiple possibilities, year 2005. implicitly assuming heterogeneity in goods and consumers, One characteristic of our industrial classification is that energy sectors, whose real behaviour should be represented by a utility function including power sectors, are disaggregated in detail, because energy systems and their technological descriptions are crucial for the purposes of this study. that accounts for non-monetary value . This notion is important Moreover, to appropriately assess bioenergy and land-use competition, agricultural when interpreting household results derived from integrated sectors are highly disaggregated . Details of the model structure and its model results, where some may select economically irrational 45 mathematical formulas were provided by Fujimori et al. and wiki page . technologies and non-monetary factors are present. However, Production sectors are assumed to maximise profits under multi-nested constant elasticity substitution (CES) functions at each input price. Energy according to our results, industrial activities have more influence transformation sectors (Supplementary Table 5) input energy and are value- over mitigation cost and our conclusions would hold true if we added based on a fixed coefficient, whereas all energy end-use sectors included such heterogeneity. (Supplementary Table 6) have elasticities between energy and the value-added We achieved relatively fast convergence compared with existing (CES aggregation of capital and labor) amount. These sectors are treated in this manner to account for energy conversion efficiency in the energy transformation studies. There are two possible reasons for the rapid convergence. sectors. Power generation from several energy sources is combined using a logit First, on AIM/CGE side, the energy consumption is forced to be function ,althoughaCESfunctionisoften used in otherCGE models.We AIM/Enduse by endogenising parameters that are exogenous in the chose this method to represent energy balance because the CES function does 47 44 conventional CGE formula. Second, the major information pro- not guarantee a physical balance . As discussed by Fujimori, Hasegawa ,an energy or physical balance violation in the CES would not be critical if the power vided by AIM/CGE to AIM/Enduse that changes the AIM/Enduse generation shares of each technology in total power generation were similar to response is the energy service changes (output of sectors and total the calibrated information. The hydrogen production sectors have similar household consumption), but the difference from the previous structure as power generation. In this study, climate mitigation changes the iteration is less than 1%, which would not change AIM/Enduse power generation mix when compared to that of the base year, and therefore is a results in terms of carbon price or power generation. key treatment. The variable renewable energy cost assumption is shown in SI 8 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE section 2. Household expenditures on each commodity are described with a In addition, the demand-supply balance of electricity within an hour is modelled linear expenditure system (LES) function. The savings ratio is endogenously using the fluctuations and flexible range of each generator. Although generators determined to balance savings and investment, and capital formation for each and flexible resources are modelled in detail, electricity demands are provided item is determined using a fixed coefficient. The Armington assumption, which exogenously. Thus, the power-dispatch stand-alone model does not determine the assumes imperfect substitutability between domestically produced and traded total electricity consumption and installed capacity by technology, which are given goods , is used for trade, and the current account is assumed to be balanced. parameters. Note that there are buffers to deal with seasonal fluctuations, such as To construct energy supply cost curves, we implemented multiple sources of fossil fuel CCS thermal plants, in the mitigation scenarios, and thus, even if we information. Solar and wind supply curves are from a study considering urban consider battery storage for seasonal fluctuations, it would remain unused due to distance . Biomass potential and supply curve data is from a land-use allocation the cost competitiveness. The no CCS scenario also uses gas thermal plants to model . adjust for seasonal differences. From the energy system model to the economic model. The following infor- An energy system model. The energy system model used in this study is a mation is given to AIM/CGE from AIM/Enduse outputs. First, Change ratio of recursive dynamic partial equilibrium model based on detailed descriptions of final energy consumption by sector and energy type; second, power generation energy technologies in the end-use and supply sectors. In this study, we used the multi-region version of AIM/Enduse [Japan] , which divides Japan into 10 regions share by energy source; third, battery capacity for stabilising fluctuations of the power supply and its capacity factor, which is taken from AIM/Power (this capacity (see Supplementary Fig. 6) based on the power grid system. The model covers energy-related GHG emissions from both energy end-use and energy supply sec- factor means the total hours that the battery used divided by a year); forth, CCS installation; fifth, investment in energy end-use sectors; sixth, carbon prices; tors. The end-use sectors are composed of industry, buildings and transportation sectors, and they are disaggregated into several subsectors with respect to types of seventh transmission losses. Final energy consumption is classified into four sectors (industry, transport, products, buildings, and transportation mode based on the IEA energy balances. The CO emissions constraint is assumed for every simulation year of the AIM/ service and residential) and fed into the CGE model. We exogenously represent these sectors, while autonomous energy efficiency improvement (AEEI) parameters are Enduse model under the mitigation scenario. Within this study, the carbon price trajectory is almost exponential as a consequence. Therefore, even if we adopt an endogenised. This treatment maintains the same number of equations and variables as in the conventional CGE approach. To integrate household energy consumption inter-temporal optimisation scheme, it would not markedly affect the results. However, this might not be the case for other carbon constraints. Mitigation and energy device purchase activities in the household, we divided the household expenditure into four categories, such as car-use activities and other energy options are selected based on linear programming to minimize total energy system consumption activities, as illustrated in the Supplementary Fig. 7 (see more detailed costs that include investments for mitigation options and energy costs subject to information in Supplementary Note 2 and Supplementary Table 11). Because the exogenous parameters such as cost and efficiency of technology, primary energy absolute value of energy consumption is not fully harmonised between these two prices, energy service demands, and emission constraints. Detailed information on the model structure and parameter settings are provided in Kainuma et al. and a models, we compare the change ratios of energy consumption with 2010 levels, which is the base year of the AIM/Enduse model, for final energy consumption list of technologies is given in Supplementary Table 7. As the models used in this study were recursive dynamic, we did not consider discounting the energy system determination. If the corresponding energy consumption was zero or very low in 2010 (less than 1 ktoe), the change ratio can lead to unrealistic projections; therefore, we costs. Nevertheless, the AIM/Enduse model annualises the capital costs of energy technologies using a discount rate in the range 5–33% (Oshiro et al.) . The sec- use absolute values. The investment in energy end-use sectors is input as an incremental capital cost compared to the baseline case, where investment costs in the toral discount rate is 5% for power and industry, 10% for transportation, and 33% for other sectors. These individual discount rates are only applied to simulate baseline is modelled by CES substitution. Moreover, the capital input coefficients are fixed at baseline levels so that additional energy investment is represented by AIM/ technology selection in the energy system model. Consequently, the energy investment data fed into the economic model are not discounted by these rates. Enduse information rather than CES substitution elasticity in the mitigation scenarios. The power sector is modelled in detail, considering the balances of electricity supply and demand in 3-h steps to assess the impacts of variable renewable From the economic model to the energy system model. Because the sectoral energies (VREs). This sector also includes measures to integrate VREs into the disaggregation of AIM/Enduse basically complies with the IEA energy balance, grid, such as electricity storage, demand response (DR) using battery-powered there are inconsistencies in the AIM/CGE, which is based on an input-output table. electric vehicles and heat-pump devices, and interconnections. The total capacity Thus, in terms of data exchange from AIM/CGE to AIM/Enduse, the subsectors was calculated based on the capacity of newly installed power plants, which was are aggregated so that the granularity of the sectors is in agreement. Nevertheless, determined endogenously, as well as that of existing plants. In the AIM/Enduse given the large share of industrial GHG emissions in Japan’s long-term low-carbon model, the residual capacities of the existing power plants in operation in 2010 scenarios, iron, chemical, paper, non-metallic minerals, and non-ferrous metals are were calculated based on individual powerplant information, such as year exempted from the sector aggregation. AIM/Enduse uses the following information constructed, capacity of each plant, and expected lifetime. generated by AIM/CGE: first, GDP changes; second, household consumption In the industry, building and transportation sectors, wide mitigation options are changes; third, industry and service sector outputs; fourth energy price changes included, such as energy-efficient devices and fuel switching. The industrial sector Economic information from AIM/CGE is input into AIM/Enduse as changes in also includes innovative technologies such as carbon capture and storage (CCS). energy service demand for each sector. Transport demand is associated with GDP However, the AIM/Enduse stand-alone model does not account for some projection in AIM/Enduse and we proportionally change the transport demand based mitigation options that contribute to reduction in energy service demands. The key on changes in GDP. The energy service demand in the industrial sectors, such as steel power generation technoeconomic information is shown in Supplementary and cement production, and outputs of other industrial sectors, is altered by the Table 8. The cost information is based on METI data (2015) , as they are outputs from AIM/CGE. Energy service demand in the household and industrial consistent with the assumptions in Japan’s NDC. Note that the estimated sectors could have low or high elasticities to relevant economic activity variables, such mitigation cost may become much lower under more optimistic assumptions as household consumption and outputs of service sectors, but remains an uncertain regarding future cost reductions, especially for renewable energies. Moreover, factor. According to the Swedish econometric analysis ,elasticity between monetary powerplant information in 2010 and fuel assumptions are shown in Supplementary and physical units of energy services can be assumed to be ~1.0. This elasticity Table 9 and Supplementary Table 10. accounts for the percent change in physical energy services caused by a 1% change in monetary outputs. Furthermore, the GDP losses indicated in this study are relatively small, less than 3%, in the CGE stand-alone model, as shown in Fig. 4a. Thus, we A power-dispatch model. The power-dispatch model used in this study is a tentatively applied an elasticity value of 1.0. Meanwhile, we varied the elasticity from recursive dynamic partial equilibrium model focused on generation planning for 0.5 to 2.0 and observed that the policy costs change slightly, but the qualitative the power sector. In other words, unlike the AIM/Enduse model covering all conclusion still holds (Supplementary Fig. 8). energy-related sectors, the AIM/Power model only covers the power generation sector. This model can simulate hourly or annual electricity generation, generation capacity, plant locations, and multiple flexible resources, and includes interregional From the energy system model to the power-dispatch model. AIM/Power’s role transmission, dispatchable power, storage, and demand responses. These variables is to present the feasibility of power-dispatch given an electricity demand and were selected based on linear programming while minimising the total system installed power capacity. Thus, AIM/Enduse provides the following items to AIM/ costs, including capital costs, operation and maintenance costs, and fuel costs Power: first, electricity demand; second, power generation installed capacity; third, under several constraints, including satisfying electricity demand and CO emis- demand response technological availability, such as heat-pump water heaters and sion reduction targets. In this study, we used a version of the model that classifies electric vehicles Japan into 10 regions (see Supplementary Fig. 6). Detailed information about this model can be found in Shiraki et al. . Note that as AIM/Enduse provides power generation installed capacity for AIM/Power, AIM/Power does not make invest- From the power-dispatch model to the economic model. AIM/Power provides ment decisions, except for making additional investments in storage and power more realism in terms of technologies to stabilise short-term fluctuations in the plants aimed at hourly and within hourly power demand-supply management power system than the other two models used in this study. Moreover, the power AIM/Power can explicitly simulate the hourly demand-supply balance of system would respond to large-scale renewable energy installations by adjusting the electricity, with consideration of daily variations in photovoltaic output caused by capacity factor for conventional power generation systems (e.g. coal-fired power) in weather conditions as well as seasonal and weekday/weekend variations in demand. addition to curtailing the output from variable renewables. These measures for NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 balancing short-term fluctuations reduce the electricity output per installed capa- 4. Clarke, L. et al. Assessing Transformation Pathways. Climate Change 2014: city, and thus affect investment decisions. It is necessary to consider this feedback Mitigation of Climate Change. Contribution of Working Group III to the Fifth from AIM/Power to AIM/Enduse. In summary, the following AIM/Power infor- Assessment Report of the Intergovernmental Panel on Climate Change. mation is given to AIM/Enduse: first, battery capacity needed to stabilise short- (Cambridge University Press, Cambridge, United Kingdom and New York, term electricity fluctuations; second, Curtailment ratio; third, capacity factors. NY, USA, 2014). Note that generation capacity, although not directly related to balancing short- 5. McCollum, D. L. et al. Energy investment needs for fulfilling the Paris term fluctuations, is decided by AIM/Enduse. Thus, the battery capacity is Agreement and achieving the Sustainable Development Goals. Nat. Energy 3, determined by AIM/Enduse when considering long-term electricity fluctuations, 589–599 (2018). whereas that for short-term fluctuations is provided by AIM/Power. 6. Calvin, K. et al. The SSP4: a world of deepening inequality. Glob. Environ. Change 42, 284–296 (2017). 7. Fricko, O. et al. The marker quantification of the Shared Socioeconomic Convergence of iterations. We confirm fast convergence between the models. Pathway 2: a middle-of-the-road scenario for the 21st century. Glob. Environ. Detailed discussion related to the convergence is made by Supplementary Note 3 Change 42(Supplement C), 251–267 (2017). where Supplementary Figs. 9–20, Supplementary Table 12 and Supplementary Table 13 shows actual convergence situation. As stated, the discrepancy between 8. van Vuuren, D. P. et al. Energy, land-use and greenhouse gas emissions trajectories under a green growth paradigm. Glob. Environ. Change 42, the model almost reaches convergences in the second step. 237–250 (2017). 9. Strachan, N. & Kannan, R. Hybrid modelling of long-term carbon reduction Scenario assumptions. There are two basic assumptions for future scenarios, scenarios for the UK. Energy Econ. 30, 2947–2963 (2008). namely, baseline and mitigation, which are carried out with and without carbon 10. Turton, H. ECLIPSE: An integrated energy-economy model for climate policy pricing to reduce GHG emissions by 80% in 2050. Basic assumptions on techno- and scenario analysis. Energy 33, 1754–1769 (2008). logical conditions, such as nuclear scenarios and CCS capacities, are taken from 56 11. Jacoby, H. D., Reilly, J. M., McFarland, J. R. & Paltsev, S. Technology and previous studies . In the results section, we describe how the mitigation cost technical change in the MIT EPPA model. Energy Econ. 28, 610–631 (2006). differs from that in the stand-alone AIM/CGE run to identify each sector’s con- 12. Fujimori, S. et al. SSP3: AIM implementation of Shared Socioeconomic tribution to the changing mitigation costs. Pathways. Glob. Environ. Change 42, 268–283 (2017). 13. Markandya, A., Halsnaes, K. Costing methodologies. Guidance Papers on the Analytical method for the diagnostic scenario runs. To investigate the extent to Cross Cutting Issues of the Third Assessment Report of the IPCC.15–31 (Global which AIM/Enduse output information for each sector contributes to mitigation Industrial and Social Progress Research Institute (GISPRI), 2000). cost adjustment compared to the conventional CGE approach, we ran diagnostic 14. Waisman, H., Guivarch, C., Grazi, F. & Hourcade, J. C. The Imaclim-R model: scenarios with and without incorporation of AIM/Enduse data by sector, as noted infrastructures, technical inertia and the costs of low carbon futures under in the supplementary information. Ultimately, we conducted 32 scenarios with imperfect foresight. Climatic Change 114, 101–120 (2012). various combinations of AIM/Enduse information for energy supply, industry, 15. Lanzi, E., Chateau, J. & Dellink, R. Alternative approaches for levelling carbon service, transport, and residential sectors taken into account or excluded. The prices in a world with fragmented carbon markets. Energy Econ. 34, indicator shown below is adopted. i, t, and s are sets of variables (e.g. energy S240–S250 (2012). demand), years and scenarios, respectively. X and Y are AIM/CGE and AIM/ i,t,s i,t,s 16. Vandyck, T., Keramidas, K., Saveyn, B., Kitous, A. & Vrontisi, Z. A global Enduse outputs, respectively. stocktake of the Paris pledges: implications for energy systems and economy. vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Glob. Environ. Change 41(Supplement C), 46–63 (2016). u X  Y t t;i;s t;i;s 17. Tuladhar, S. D., Yuan, M., Bernstein, P., Montgomery, W. D. & Smith, A. A ErrInd ¼ hi i;s t X þY top–down bottom–up modeling approach to climate change policy analysis. t;i;s t;i;s t 2 Energy Econ. 31, S223–S234 (2009). 18. Arndt, C. et al. A sequential approach to integrated energy modeling in South Africa. Appl. Energy 161, 591–599 (2016). Reporting summary. Further information on research design is available in 19. Böhringer, C. & Rutherford, T. F. Combining bottom-up and top-down. the Nature Research Reporting Summary linked to this article. Energy Econ. 30, 574–596 (2008). 20. Böhringer, C. & Rutherford, T. F. Integrated assessment of energy policies: decomposing top-down and bottom-up. J. Economic Dyn. Control 33, Data availability 1648–1661 (2009). Scenario data are accessible online via HARVARD Dataverse (https://doi.org/10.7910/ 21. Hartwig, J., Kockat, J., Schade, W. & Braungardt, S. The macroeconomic DVN/QE6ERU). The data which are derived from the original scenario database shown effects of ambitious energy efficiency policy in Germany—Combining bottom- as figures but not in the above database is available upon requests. The source data up energy modelling with a non-equilibrium macroeconomic model. Energy underlying Figs. 1–5 and Supplementary Figs 1–20 are provided as a Source Data file. 124, 510–520 (2017). 22. Kriegler, E. et al. Fossil-fueled development (SSP5): an energy and resource Code availability intensive scenario for the 21st century. Glob. Environ. Change-Hum. Policy The source code for generating figures used in the main text and supplementary Dimens. 42, 297–315 (2017). information is available in HARVARD Dataverse (https://doi.org/10.7910/DVN/ 23. Jacobson, M. Z., Delucchi, M. A., Cameron, M. A. & Frew, B. A. Low-cost QE6ERU), which name is the Sourcecode.zip. The current code base of the AIM/Enduse, solution to the grid reliability problem with 100% penetration of intermittent AIM/CGE and AIM/Power developed over more than two decades at Kyoto University, wind, water, and solar for all purposes. Proc. Natl. Acad. Sci. USA 112, Shiga Prefecture University and National Institute for Environmental Studies. AIM/ 15060–15065 (2015). Enduse is available at AIM website (http://www-iam.nies.go.jp/aim/data_tools/index. 24. Zerrahn, A. & Schill, W.-P. Long-run power storage requirements for high html#enduse) and others are not available in a publicly shareable version. The code will shares of renewables: review and a new model. Renew. Sustain. Energy Rev. 79, continue to be developed and hosted by Kyoto University, Department of Environmental 1518–1534 (2017). Engineering (http://www.athehost.env.kyoto-u.ac.jp/). Requests for code should be 25. Santos-Alamillos, F. J., Archer, C. L., Noel, L., Budischak, C. & Facciolo, W. addressed to Shinichiro Fujimori. Assessing the economic feasibility of the gradual decarbonization of a large electric power system. J. Clean. Prod. 147, 130–141 (2017). 26. Pye, S., Li, F. G. N., Price, J. & Fais, B. Achieving net-zero emissions through Received: 7 November 2018; Accepted: 27 September 2019; the reframing of UK national targets in the post-Paris Agreement era. Nat. Energy 2, 17024 (2017). 27. Luderer, G. et al. Assessment of wind and solar power in global low-carbon energy scenarios: an introduction. Energy Econ. 64, 542–551 (2017). 28. Pietzcker, R. C. et al. System integration of wind and solar power in integrated assessment models: a cross-model evaluation of new approaches. Energy Econ. References 64, 583–599 (2017). 1. UNFCCC. FCCC/CP/2015/L.9/Rev.1: Adoption of the Paris Agreement. 29. Japan MoE. Overview of the Plan for Global Warming Countermeasures (UNFCCC, Paris, 2015). -Cabinet decision on May 13, 2016. 2016 https://www.env.go.jp/press/files/en/ 2. The White House. United States mid-century strategy for deep decarbonization. 676.pdf (2016). (United Nations Framework Convention on Climate Change, Washington, 30. Aldy, J. et al. Economic tools to promote transparency and comparability in DC, 2016). the Paris Agreement. Nat. Clim. Change 6, 1000 (2016). 3. Iyer, G. et al. Measuring progress from nationally determined contributions to 31. Oshiro, K., et al. Mid-century emission pathways in Japan associated with the mid-century strategies. Nat. Clim. Change 7, 871–874 (2017). global 2 °C goal: national and global models’ assessments based on carbon 10 NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12730-4 ARTICLE budgets. Climatic Change 1–15 https://link.springer.com/article/10.1007/ 55. Shiraki, H., Ashina, S., Kameyama, Y., Hashimoto, S. & Fujita, T. Analysis of s10584-019-02490-x. (2019). optimal locations for power stations and their impact on industrial symbiosis 32. Sugiyama, M. et al. Japan’s long-term climate mitigation policy: multi-model planning under transition toward low-carbon power sector in Japan. J. Clean. assessment and sectoral challenges. Energy 167, 1120–1131 (2019). Prod. 114,81–94 (2016). 33. van der Mensbrugghe, D. The environmental impact and sustainability 56. Oshiro, K., Kainuma, M. & Masui, T. Implications of Japan’s 2030 target for applied general equilibrium (ENVISAGE) model. The World Bank, January long-term low emission pathways. Energy Policy 110, 581–587 (2017). 2008 http://siteresources.worldbank.org/INTPROSPECTS/Resources/334934- 1193838209522/Envisage7b.pdf (2008). Acknowledgements 34. Babiker, M. H., et al. The MIT Emissions Prediction and Policy Analysis S.F., K.O., and T.H. are supported by the Environment Research and Technology (EPPA) Model; Revisions, Sensitivities, and Comparisons of Results. Development Fund (2-1702, 2-1908) of the Environmental Restoration and Conservation (Cambridge, MA, 2001). Agency of Japan and JSPS KAKENHI Grant (No. 19H02273). S.F. is supported by 35. van der Werf, E. Production functions for climate policy modeling: an COMMIT (Climate pOlicy assessment and Mitigation Modeling to Integrate national empirical analysis. Energy Econ. 30, 2964–2979 (2008). and global Transition pathways) financed by Directorate General Climate Action (DG 36. Jorgenson, D. W. Econometric methods for modeling producer behavior. CLIMA) and EuropeAid under grant agreement No. 21020701/2017/770447/SER/ Handb. Econ. 3, 1841–1915 (1986). CLIMA.C.1 EuropeAid/138417/DH/SER/MulitOC. H.S. is supported by the Murata 37. McCollum, D. L. et al. Improving the behavioral realism of global integrated Science Foundation. assessment models: an application to consumers’ vehicle choices. Transportation Res. Part D Transp. Environ. 55, 322–342 (2017). 38. Fujimori, S. et al. SSP3: AIM implementation of Shared Socioeconomic Author contributions Pathways. Glob. Environ. Change-Hum. Policy Dimens. 42,268–283 (2017). S.F. designed the research; S.F carried out analysis of the modelling results; S.F. and H.S. 39. Hasegawa, T. et al. Climate change impact and adaptation assessment on food created figures; S.F. wrote the draft of the paper; S.F., K.O. and H.S. provided economic, consumption utilizing a new scenario framework. Environ. Sci. Technol. 48, energy system, and power model data; and S.F., K.O., H.S., and T.H. contributed to the 438–445 (2014). discussion and interpretation of the results. 40. Hasegawa, T. et al. Consequence of climate mitigation on the risk of hunger. Environ. Sci. Technol. 49, 7245–7253 (2015). 41. Mittal S., Dai H., Fujimori S., Masui T. Bridging greenhouse gas emissions and Competing interests renewable energy deployment target: comparative assessment of China and The authors declare no competing interests. India. Appl. Energy 166, 301–313 (2016). 42. Hasegawa, T., Fujimori, S., Takahashi, K., Yokohata, T. & Masui, T. Economic Additional information implications of climate change impacts on human health through Supplementary information is available for this paper at https://doi.org/10.1038/s41467- undernourishment. Climatic Change 136,1–14 (2016). 019-12730-4. 43. Fujimori, S., Kainuma, M., Masui, T., Hasegawa, T. & Dai, H. The effectiveness of energy service demand reduction: a scenario analysis of global Correspondence and requests for materials should be addressed to S.F. climate change mitigation. Energy Policy 75(Supplement C), 379–391 (2014). 44. Fujimori, S., Hasegawa, T., Masui, T. & Takahashi, K. Land use representation Peer review information Nature Communications thanks Per Ivar Helgesen, Ronald in a global CGE model for long-term simulation: CET vs. logit functions. Food Sands and the other, anonymous, reviewer(s) for their contribution to the peer review of Sec 6, 685–699 (2014). this work. Peer reviewer reports are available. 45. IAMC. The common Integrated Assessment Model (IAM) documentation. https://www.iamcdocumentation.eu/index.php/IAMC_wiki (2018). Reprints and permission information is available at http://www.nature.com/reprints 46. Clarke, J. F. & Edmonds, J. A. Modelling energy technologies in a competitive market. Energy Econ. 15, 123–129 (1993). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in 47. Schumacher, K. & Sands, R. D. Innovative energy technologies and climate published maps and institutional affiliations. policy in Germany. Energy Policy 34, 3929–3941 (2006). 48. Armington, S. P. A theory of demand for products distinguished by place of production. Staff Pap. 16, 159–178 (1969). Open Access This article is licensed under a Creative Commons 49. Herran, D. S., Dai, H., Fujimori, S. & Masui, T. Global assessment of onshore wind power resources considering the distance to urban areas. Energy Policy Attribution 4.0 International License, which permits use, sharing, 91,75–86 (2016). adaptation, distribution and reproduction in any medium or format, as long as you give 50. Hasegawa, T., Fujimori, S., Ito, A., Takahashi, K. & Masui, T. Global land-use appropriate credit to the original author(s) and the source, provide a link to the Creative allocation model linked to an integrated assessment model. Sci. Total Environ. Commons license, and indicate if changes were made. The images or other third party 580, 787–796 (2017). material in this article are included in the article’s Creative Commons license, unless 51. Oshiro, K. & Masui, T. Diffusion of low emission vehicles and their impact on indicated otherwise in a credit line to the material. If material is not included in the CO2 emission reduction in Japan. Energy Policy 81, 215–225 (2015). article’s Creative Commons license and your intended use is not permitted by statutory 52. Kainuma, M. M. Y. & Morita, T. Climate policy assessment: Asia-Pacific regulation or exceeds the permitted use, you will need to obtain permission directly from integrated modeling. (Springer, Japan, 2003). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 53. Oshiro, K., Kainuma, M. & Masui, T. Assessing decarbonization pathways and licenses/by/4.0/. their implications for energy security policies in Japan. Clim. Policy 16(sup1), S63–S77 (2016). 54. METI. Report on analysis of generation costs, etc. for subcommittee on long- © The Author(s) 2019 term energy supply and demand outlook. (Ministry of Economy, Trade and Industry, 2015). NATURE COMMUNICATIONS | (2019) 10:4737 | https://doi.org/10.1038/s41467-019-12730-4 | www.nature.com/naturecommunications 11

Journal

Nature CommunicationsSpringer Journals

Published: Oct 18, 2019

There are no references for this article.