The role of technology for achieving climate policy objectives: overview of the EMF 27 study on global technology and climate policy strategies
- First Online:
- Cite this article as:
- Kriegler, E., Weyant, J.P., Blanford, G.J. et al. Climatic Change (2014) 123: 353. doi:10.1007/s10584-013-0953-7
This article presents the synthesis of results from the Stanford Energy Modeling Forum Study 27, an inter-comparison of 18 energy-economy and integrated assessment models. The study investigated the importance of individual mitigation options such as energy intensity improvements, carbon capture and storage (CCS), nuclear power, solar and wind power and bioenergy for climate mitigation. Limiting the atmospheric greenhouse gas concentration to 450 or 550 ppm CO2 equivalent by 2100 would require a decarbonization of the global energy system in the 21st century. Robust characteristics of the energy transformation are increased energy intensity improvements and the electrification of energy end use coupled with a fast decarbonization of the electricity sector. Non-electric energy end use is hardest to decarbonize, particularly in the transport sector. Technology is a key element of climate mitigation. Versatile technologies such as CCS and bioenergy are found to be most important, due in part to their combined ability to produce negative emissions. The importance of individual low-carbon electricity technologies is more limited due to the many alternatives in the sector. The scale of the energy transformation is larger for the 450 ppm than for the 550 ppm CO2e target. As a result, the achievability and the costs of the 450 ppm target are more sensitive to variations in technology availability.
Anthropogenic climate change is caused by emissions of greenhouse gases (GHGs), aerosols, and other short-lived species from fossil fuel use, industrial processes, agriculture and land use practices. Global GHG emissions have been rising steadily since the industrial revolution and reached 50 GtCO2-eq in 2010 (European Commission and Netherlands Environmental Assessment Agency 2011). Mitigating climate change will require a reversal of this trend by reducing and eventually phasing out GHG emissions. Major questions remain as to how and when this transition to a state of zero or very low GHG emissions should be accomplished, how this transition depends on the choice of climate change target, and what is implied for the underlying transition of global energy and land use patterns.
Low-carbon technologies in the energy system have been identified as a key element for mitigating climate change (e.g. Clarke et al. 2008; Edenhofer et al. 2010), but a clear picture about the role of individual mitigation technologies has yet to emerge (Nakicenovic and Nordhaus 2011). The relative importance of mitigation technologies depends not only on their techno-economic characteristics and how they develop in the future, but also on the competition with other energy technologies, the development of future energy demand and the climate policy architecture. Since all of these factors are interconnected and surrounded by large uncertainty, it is important to investigate technology strategies from a system perspective and under a variety of assumptions. The Stanford Energy Modeling (EMF) Study 27 has employed 18 energy-economy and integrated assessment models (EE&IAMs) from different world regions in a coordinated model comparison exercise to explore the role of various low-carbon technologies in ambitious mitigation scenarios, including different assumptions about technology availability, energy demand and climate policies. The results provide a robust picture of the importance of individual technologies and the determining factors, and constitute a useful resource for climate policy makers.
The EMF27 study builds upon a rich set of model studies on climate mitigation scenarios. Key model comparison studies for instance include the previous EMF climate-change-oriented studies like the EMF19 study on carbon constraints and advanced energy technologies (Weyant 2004), the EMF21 study on non-CO2 Kyoto gas mitigation (Weyant et al. 2006), and the EMF22 study on climate control scenarios including phased participation (Clarke et al. 2009). Other important information comes from recent model intercomparison projects (Edenhofer et al. 2010; Luderer et al. 2012; Calvin et al. 2012) and assessments (Clarke et al. 2008; Krey and Clarke 2011; Riahi et al. 2012) that investigated the role of low-carbon technologies in mitigation scenarios. Compared to these studies, the new contribution of EMF27 is the breadth of technology, energy demand and policy scenarios investigated with a large international consortium, and a detailed exploration of emissions and technology dynamics in individual sectors. Another important new feature of EMF27 is that 14 of the 18 participating EE&IAMs included the availability of Bio-Energy with Carbon Capture and Storage (BECCS) and – in a few cases – forest and soil carbon stock conservation and/or enhancement. Existing studies have shown that BECCS can be a key option for attaining stringent stabilization targets (Azar et al. 2010; Tavoni and Socolow 2013).
While this paper provides a synthesis of key results, the dimensions of technology availability and climate policy regimes are explored in greater depth in two separate overview papers (Krey et al., Blanford et al. this issue). These are augmented by a set of comparative analyses on Non-Kyoto forcing (Rose et al. (b)), fossil fuel use (McCollum et al.), CCS (Koelbl et al.), renewable energy (Luderer et al.), bioenergy (Rose et al. (a)), land use implications (Popp et al.), nuclear energy (Kim et al.) and energy efficiency (Sugiyama et al.). Individual contributions by EMF27 modeling teams analyze additional topics in further detail.
2.1 Participating models
Eighteen global energy-economy and integrated assessment models participated in the EMF27 study, originating from the USA (GCAM, FARM, MERGE, Phoenix), Canada (EC-IAM, TIAM-WORLD , which is now used globally), the European Union (IMACLIM, IMAGE, MESSAGE, POLES, REMIND, WITCH), Japan (AIM-Enduse, BET, DNE21+, GRAPE), India (GCAM-IIM) and the OECD (ENV-Linkages). Further details on these models can be found in the Supplementary Online Material (SOM). The models differ in numerous ways including their sectoral coverage, solution algorithm, representation of GHG emissions, energy demand and supply sectors, baseline assumptions and assumptions about techno-economic parameters. The large ensemble of models permits us to explore ranges of outcomes reflecting both structural as well as parametric uncertainties.
2.2 Scenario design
The technology variations were chosen to reflect generic, but not implausible constraints on the deployment of key mitigation technologies. These include the possibility that CCS will not become commercially available, that nuclear energy is phased out, that the share of intermittent solar and wind power on the electricity grid will be limited, and that modern bioenergy use will be constrained. The study also considered a case with increased energy intensity improvements compared to historical trends included in the baseline.
The policy variations included two different target levels of radiative forcing corresponding to atmospheric concentrations of 450 and 550 ppm CO2 equivalent (CO2e) which represent a range of potential long-term goals of international climate policy. The stringency of the lower target is consistent with concurrent interpretations of the goal to keep global mean warming below 2 °C (Meinshausen et al. 2009). In EMF27, the limit was imposed on the combined radiative forcing from all radiative agents of anthropogenic origin with the exception of nitrate aerosols, mineral dust aerosols, and land use albedo changes (which we call AN3A forcing). The lower 450 limit was set to 2.8 W/m2 allowing for overshoot before 2100, while the higher 550 limit was set to 3.7 W/m2 that was not to be exceeded during the 21st century (Blanford et al. this issue).
The models were asked to implement the targets by assuming full when (timing of emission reductions), where (country or region) and what (sector) flexibility of emissions reductions to ensure cost-effective mitigation efforts. In addition, two policy-driven scenarios were implemented that deviated from these idealized assumptions. The first aimed at extrapolating the Copenhagen pledges for individual world regions until 2100 (called fragmented policy scenario – FP – in the following), while the second took up a proposal by the G8 to reduce global emissions by 50 % until 2050, and modified it by assuming a group of non-participating fossil-fuel-rich countries (Blanford et al. this issue). Additional explanation of the study design, the study protocol and the implementation by modeling teams can be found in the SOM.
2.3 Target feasibility
The question of whether or not a target is feasible is important but subject to different interpretations. Furthermore, it cannot be linked directly to whether or not a model returns a solution. For some, a target is feasible if any set of actions exists that could cause the target to be met, whether or not a given model can find such a path. For others, feasibility is purely a model issue that takes into account assumptions about technology availability and how rapidly technology can deploy. For yet others, it is a matter of assessing whether or not the political process could accept a solution, for example carbon prices might be “unacceptably high.”
We take the approach that it is most useful to report all available model results, since any supposed technical or economic infeasibility can be assessed best in a comparison across model results (see also Tavoni and Tol 2010). In the remaining cases, where models did not return a solution, and where unrelated numerical problems were not identified as the cause, we took it as an indication that the target is technically or economically infeasible for the given model and scenario setup. Such a finding is, of course, contingent on the model. However, a statistics of how many models did not return a solution among those who attempted the scenario should be indicative of the strain that is imposed on the modeled energy-economy-climate system.
3.1 Emissions with and without climate policies
Strong emissions reductions are needed to reach the GHG concentration levels of 450 and 550 ppm CO2e in 2100 (Fig. 1). For 550 ppm CO2e, GHG emissions levels are reduced to 26–38 (Median: 32) GtCO2e in 2050 and 14–24 (Median: 17) GtCO2e in 2100. For 450 ppm CO2e, emissions reductions are even stronger, leading to 20–28 (Median: 24) GtCO2 emissions in 2050 and close to or below zero emissions (−9 to 5 GtCO2e) by the end of the century. Since overshooting the stabilization target prior to 2100 is allowed in the 450 ppm CO2e case but not for 550 ppm CO2e, emissions trajectories actually remain close to each other in the first half of the century but increasingly diverge in the second half.
In reaching the forcing targets, not only emissions of Kyoto gases, but also other forcing constituents such as aerosols and tropospheric ozone play a role. Rose et al. (this issue (b)) describe and assess the state of current non-Kyoto radiative forcing modeling in a subset of EMF27 models. The study finds aerosol emissions mask significant baseline warming. However, there are large differences across models in projected non-Kyoto emissions and forcing, so further evaluation is merited.
Figure 2 shows the range of projected land use and fossil fuel and industry CO2 emissions across various sectors. Despite the large model spread, the different scales of the emissions reductions to reach the 450 and 550 ppm climate targets is clearly visible from 2030 on. There is also a clear distinction between the profiles of direct emissions from electricity generation and the energy end use sectors (see also Figure S3). The electricity sector is decarbonized first, with close to zero (550) or net negative emissions (450) in 2050, and consistently negative emissions in 2100 for those models that include BECCS. These negative emissions compensate for part of the residual emissions from fossil fuel use in the end use sectors. The transport sector shows the largest residual emissions with emissions levels returning to present day levels by 2030–50 for 450 ppm CO2e, and by 2100 for 550 ppm CO2e.
3.2 Economic implications of climate policies
The aggregate macro-economic costs of mitigation are captured to different degrees by different models. Partial equilibrium models report abatement costs for instance as the area under the marginal abatement cost curve (MAC), while general equilibrium models can derive consumption losses including economy wide effects. Such measures (partially) describe the direct costs of mitigation, but neither include climate benefits nor as adverse side effects and co-benefits of climate policy (McCollum et al. 2011; Riahi et al. 2012).
For both climate targets, abatement costs and consumption losses grow faster than the economy in the baseline. Their net present value over the period 2010–2100 (at 5 % discounting) is between 0.4-1.1% and 0.7-2.2% (IMACLIM: 8.0%), respectively, of the baseline economy for 550 ppm CO2e, and 0.8-2.9% and 0.9-3.3% (IMACLIM: 11.7%), respectively, for 450 ppm CO2e. Costs increase between 25 and 200 % between the two targets (Fig. 3). The large variation across models is due to the factors influencing emissions prices, and the extent to which economy-wide effects and market distortions in the baseline are captured. The IMACLIM model with the most extensive treatment of non-climate market distortions shows the highest costs, which could be reduced by exploiting a potential double dividend of climate policy (Bibas et al. this issue). Cost estimates and their increase between the 550 and 450 CO2e target can change significantly when technology availability is constrained (see below; Krey et al. this issue).
All cost estimates for the two climate targets hold for the idealized assumption of universal emissions pricing. Models consistently project higher mitigation costs for the G8 policy scenario for similar levels of abatement (Figure S2). This is due to inefficiencies induced by the fact that a group of fossil resource rich countries does not join the global climate regime (Blanford et al. this issue). The low-cost abatement options in these countries are left untouched and there is a limited amount of emissions leakage. Remarkably, the non-participating countries incur significant costs from the adoption of stringent mitigation policies by the rest of the world due to a loss of fossil fuel export revenues.
3.3 Energy system transformation pathways
Fossil fuel resources are not the limiting factor for GHG emissions. The decarbonization of the energy system leads to a strongly decreasing fossil fuel consumption compared to the baseline development (McCollum et al. this issue). Climate policies could lead to a major reallocation of financial flows in fossil fuel markets between regions and near-term synergies for energy security.
The pace of the transformation of the energy system is accelerated significantly when moving from the 550 to the 450 ppm CO2e target, with both low-carbon supply side options and energy efficiency being significantly upscaled. Energy intensity improvements are accelerated to 1.3–2.9 % (Median: 2.3 %) per year for reaching 450 ppm CO2e compared to the 1970–2010 global rate of 1.3 % per year (Sugiyama et al. this issue). In addition, the direct emissions from the end-use sectors are reduced or compensated by negative emissions to a much larger degree under the 450 ppm CO2e target. CCS and bioenergy, and in particular the combination of both (BECCS), play a crucial role in this (Rose et al. (a) this issue). BECCS notably affects the cost-effective global emissions trajectory by accommodating prolonged use of fossil fuels without CCS. It should be noted that many models lack representation of alternative negative emissions technologies such as afforestation.
CCS is deployed at a substantial scale in almost all EMF27 mitigation scenarios with full technology availability (Figure S4; Koelbl et al. this issue). It can be combined with a variety of feedstocks and energy conversion technologies. While before 2050, coal, gas and biomass are used with CCS at comparable scale, biomass is becoming the most important CCS feedstock in the second half of the century because of the resulting net negative emissions. Rose et al. (this issue (a)) find modern bioenergy projected to grow 1–10 % per annum through 2050, with bioenergy reaching 1–35 % of global primary energy by 2050, and 10–50 % by 2100 exhibiting a wide range across the models. A comparative analysis of the land use implications (Popp et al. this issue) found significant differences between three models with integrated land use components in EMF27 (GCAM, IMAGE, REMIND/MAgPIE). Under climate policy, bioenergy cropland represents 24–36 % of total cropland by 2100, but bioenergy feedstocks, land use emissions and carbon sinks vary notably across models. More research is needed to better understand the role of land use in climate stabilization.
In the climate policy scenarios, renewables contribute significantly to long-term electricity supply, while the contribution of renewables other than biomass to non-electric energy supply is limited. Deployment levels, in particular for wind and solar power, vary considerably across models due to differences in assumptions about costs and resource potentials, and the representation of integration challenges related to fluctuating supply from wind and solar power (Luderer et al. this issue). The majority of models project nuclear energy use for electricity generation to increase in the climate policy scenarios relative to the baseline, but deployment levels vary widely due to different assumptions about costs and uranium availability, and perspectives on nuclear risk factors (Kim et al. this issue). Models show different trade-offs between nuclear and renewable electricity, with some models projecting shares of comparable size (DNE21+, GCAM, EC-IAM, IMAGE, Phoenix, WITCH), and others foreseeing a dominant role of renewable electricity in the long run (AIM-Enduse, BET, MERGE, MESSAGE, POLES, REMIND, TIAM-World; Fig. 3c).
3.4 The impact of limiting technology availability
In many models, transport and industry are the limiting sectors with regards to emissions reductions, with high-cost decarbonization options in these sectors often driving mitigation costs. Consequently, limitations on mitigation options relevant for non-electric energy use, primarily bioenergy and CCS, have a strong impact on mitigation costs (Fig. 5, Krey et al. this issue). Their value is further increased by the fact that their combination (BECCS) can be used to produce negative emissions. By contrast, limited availability and performance of options constrained to the electricity sector such as nuclear, solar and wind power showed a smaller impact on mitigation costs due to a relatively large substitutability among these options. The assumption of approx. 50 % larger energy intensity improvement rates lead to a significant reduction in mitigation costs. This result may not, however, include the full costs of improving energy efficiency, as most models have a very limited accounting of demand side investments and costs.
Reliance on the full portfolio of mitigation technologies increases when moving from the 550 to the 450 ppm CO2e target. Models could identify transformation pathways under the 550 ppm CO2e target for all limited mitigation technology portfolios considered in this study – albeit at different costs. By contrast, only four models could achieve the 450 ppm CO2e target without CCS. Variation of mitigation costs with technology availability is much stronger under a stringent 450 ppm CO2e target compared to the 550 ppm CO2e target (Fig. 5).
Some of the technology variations, most notably if CCS is not available or if bioenergy is limited, lead to significant changes in the emissions profiles compared to the full-technology availability counterparts (Krey et al. this issue; Rose et al. (b) this issue). The reason for this is related to the associated constraint on negative emissions that can no longer be deployed (to the original extent) in the long run to compensate for more emissions in the short term.
The study also investigated a counterfactual scenario with limitations on all mitigation technology clusters, defined as the combination of the individual technology constraints considered in the study (LimTech). Mitigation costs for achieving a 550 ppm CO2e target increased significantly compared to the situation in which the availability of only a subset of technology clusters was limited. No model could find a solution to achieve the 450 ppm CO2e target under these conditions. This highlights the fact that technology is a central component of meeting the climate stabilization challenge.
4 Discussion and conclusions
Emissions: Extrapolating current levels of fragmented policy action or increasing energy intensity improvement rates by 50 % fall short of emissions reductions that would be required for reaching the 450 and 550 ppm CO2e targets. Models exhibit considerable uncertainty about the emissions implications of long-term climate targets and project different allocations of the mitigation effort across sectors with particular uncertainty about the use of land use based mitigation options. Thus, a climate target can be consistent with a range of caps on cumulative future fossil fuel emissions.
Energy transformation and technology value: Mitigation pathways show a massive transformation of the energy system. Robust characteristics of this transformation are increased energy intensity improvements and the electrification of energy end use coupled with a fast decarbonization of the electricity sector. Results indicate that non-electric energy end use is hardest to decarbonize. Technology is a key element for reaching climate targets. Versatile technologies such as CCS and bioenergy have largest value, part of it due to their combined ability to produce negative emissions. The availability of a negative emissions technology seems to be a key element for meeting the climate targets due to the ability to compensate fossil fuel emissions across sectors and time. A multitude of mitigation options in the electricity sector limit dependence on individual mitigation technologies. Rapid energy intensity improvements reduce the mitigation challenge significantly.
The 450 ppm CO2e target requires a stronger decarbonization of non-electric energy use, and a larger deployment of negative emissions compared to the 550 ppm CO2e target. As a result, the achievability and the costs of the 450 ppm CO2e target are more sensitive to variations in technology availability.
Economic implications: Mitigation costs increase by a factor of 1.5 to 2.5 in the large majority of models when moving from 550 ppm to 450 ppm CO2e, but remain below 3.5 % of the baseline economy (NPV for 2010–2100) for all but one model if technology availability is not constrained. The costs are moderated by the availability of BECCS in most models, particularly for the 450 ppm CO2e target. Cost numbers depend on the idealized policy assumption of universal carbon pricing and largely efficient markets. Regional fragmentation leads to higher mitigation costs for achieving the same level of emissions reduction.
Near term implications: Finally, we note that the differences in emissions and energy technology deployments between 450 and 550 ppm mitigation pathways are limited until 2030, even though they are significant in the long term. This is facilitated by the fact that the 450 ppm CO2e target allows for overshooting the target prior to 2100, which can be exploited in particular if negative emissions are available in the long run to compensate for relatively higher near-term emissions. The implications of alternative near-term emissions targets for long-term climate goals are investigated by a concurrent study (Riahi et al., 2013).
Several caveats of the study need to be highlighted. For reaching the long-term climate targets, the study assumes universal emissions pricing starting almost immediately. In the real world, global cooperative action is unlikely to be implemented before 2020, and prospects thereafter are uncertain. In addition, models do not account for the institutional challenges that implementing a price on carbon may bring on the national to regional scale. All of these factors can substantially increase the mitigation challenge, including the costs of mitigation. The study also focused on the direct impacts of mitigation, not including benefits from avoided climate change, co-benefits to other policy objectives such as air pollution, nor adverse side effects, e.g. on food security. These dimensions need to be taken into account when putting the results of this study into a policy context.
We conclude that ambitious climate policy objectives imply a large-scale transformation of the energy and land system. The higher the ambition, the more important the availability of key technologies and an efficient global climate policy regime will be.
Those models used endogenous climate modules that can differ significantly in their response to emissions trajectories, particularly for the climate policy cases. This adds an additional layer of uncertainty to climate outcomes and also affects the amount of residual emissions that models estimate to be consistent with the climate targets.
This is true in the context of this study, since no additional climate policy measures such as technology performance standards or subsidies are assumed.
Most of the EMF27 models assume an interest rate of around 5 % per year. The choice of discount rate affects the average price/net present value cost estimates. Lower discount rates lead to higher average prices/net present value costs, if prices/costs increase over time.
Jae Edmonds and Leon Clarke are grateful for research support provided by the Integrated Assessment Research Program in the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-76RL01830. Results reported for the GCAM model used Evergreen computing resources at the Pacific Northwest National Laboratory’s Joint Global Change Research Institute at the University of Maryland in College Park, which is supported by the Integrated Assessment Research Program in the Office of Science of the U.S. Department of Energy. The views and opinions expressed in this paper are those of the authors alone.
The contribution of Elmar Kriegler, Volker Krey, Gunnar Luderer, Keywan Riahi, Massimo Tavoni and Detlev van Vuuren to this research was supported by funding from the European Commission's Seventh Framework Programme under the LIMITS project (grant agreement no. 282846).