Supplementary Material to

“Transport electrification: a key element for energy system transformation and climate stabilization”

in Climatic Change David McCollum1#*, Volker Krey1#, Peter Kolp1, Yu Nagai1, Keywan Riahi1

1International Institute for Applied Systems Analysis, Laxenburg 2361, Austria.

#D.M. and V.K. contributed equally to this work.*e-mail: mccollum@iiasa.ac.at

Table of Contents1. Brief description of the MESSAGE-MACRO integrated assessment modeling framework............................2

2. Brief description of the MAGICC reduced-complexity global climate model................................................5

3. Brief description of the stylized transport module in MESSAGE-MACRO.....................................................6

4. Assumptions for electric vehicle stock calculations in the main text............................................................8

5. Feedstock CO2 emissions sensitivity analysis................................................................................................8

6. Measuring diversity through the Shannon-Wiener index...........................................................................10

7. Further details on mitigation costs and scenario feasibility........................................................................10

References...........................................................................................................................................................13

1

1. Brief description of the MESSAGE-MACRO integrated assessment modeling framework

The MESSAGE (Model for Energy Supply Strategy Alternatives and their General Environmental Impact) integrated assessment model (IAM) is a global systems engineering optimization model used for medium- to long-term energy system planning, energy policy analysis, and scenario development (Messner and Strubegger 1995; Riahi et al. 2012; van Vliet et al. 2012). Developed at the International Institute for Applied Systems Analysis (IIASA) for more than two decades, MESSAGE is an evolving framework that, like other global IAMs in its class (e.g., MERGE, ReMIND, IMAGE, WITCH, GCAM, etc.), has gained wide recognition over time through its repeated utilization in developing global energy and emissions scenarios (e.g., Nakicenovic and Swart (2000)).

The MESSAGE model divides the world up into eleven (11) regions (Supplementary Figure 1, Supplementary Table 1) in an attempt to represent the global energy system in a simplified way, yet with many of its complex interdependencies, from resource extraction, imports and exports, conversion, transport, and distribution, to the provision of energy end-use services such as light, space conditioning, industrial production processes, and transportation. Trade flows (imports and exports) between regions are monitored, capital investments and retirements are made, fuels are consumed, and emissions are generated. In addition to the energy system, the model includes also the other main greenhouse-gas emitting sectors, agriculture and forestry. MESSAGE tracks a full basket of greenhouse gases and other radiatively active gases – CO2 , CH4 , N2O , NOx , volatile organic compounds (VOCs), CO, SO2, PM, BC, OC, NH3, CF4, C2F6, HFC125, HFC134a, HFC143a, HFC227ea, HFC245ca, and SF6 – from both the energy and non-energy sectors (e.g., deforestation, livestock, municipal solid waste, manure management, rice cultivation, wastewater, and crop residue burning). In other words, all Kyoto gases plus several others are accounted for.

NAM PAO

WEU

EEU

FSU

MEA

AFR

LAM

SAS

PAS

CPA

1 NAM North America 2 LAM Latin America & The Caribbean 3 WEU Western Europe 4 EEU Central & Eastern Europe

5 FSU Former Soviet Union 6 MEA Middle East & North Africa 7 AFR Sub-Saharan Africa 8 CPA Centrally Planned Asia & China

9 SAS South Asia 10 PAS Other Pacific Asia 11 PAO Pacific OECD

OECD

REFS

ALM

ASIA

Supplementary Figure 1. Map of 11 regions in MESSAGE model

2

Supplementary Table 1. Listing of 11 MESSAGE regions by country

11 MESSAGE regions

Definition (list of countries)

NAMNorth America(Canada, Guam, Puerto Rico, United States of America, Virgin Islands)

WEU

Western Europe(Andorra, Austria, Azores, Belgium, Canary Islands, Channel Islands, Cyprus, Denmark, Faeroe Islands, Finland, France, Germany, Gibraltar, Greece, Greenland, Iceland, Ireland, Isle of Man, Italy, Liechtenstein, Luxembourg, Madeira, Malta, Monaco, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, United Kingdom)

PAOPacific OECD(Australia, Japan, New Zealand)

EEU

Central and Eastern Europe(Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, The former Yugoslav Rep. of Macedonia, Hungary, Poland, Romania, Slovak Republic, Slovenia, Estonia, Latvia, Lithuania)

FSUFormer Soviet Union(Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Republic of Moldova, Russian Federation, Tajikistan, Turkmenistan, Ukraine, Uzbekistan)

CPACentrally Planned Asia and China(Cambodia, China (incl. Hong Kong), Korea (DPR), Laos (PDR), Mongolia, Viet Nam)

SASSouth Asia(Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, Sri Lanka)

PAS

Other Pacific Asia(American Samoa, Brunei Darussalam, Fiji, French Polynesia, Gilbert-Kiribati, Indonesia, Malaysia, Myanmar, New Caledonia, Papua, New Guinea, Philippines, Republic of Korea, Singapore, Solomon Islands, Taiwan (China), Thailand, Tonga, Vanuatu, Western Samoa)

MEA

Middle East and North Africa(Algeria, Bahrain, Egypt (Arab Republic), Iraq, Iran (Islamic Republic), Israel, Jordan, Kuwait, Lebanon, Libya/SPLAJ, Morocco, Oman, Qatar, Saudi Arabia, Sudan, Syria (Arab Republic), Tunisia, United Arab Emirates, Yemen)

LAC Latin America and the Caribbean(Antigua and Barbuda, Argentina, Bahamas, Barbados, Belize, Bermuda, Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, Dominica, Dominican Republic, Ecuador, El Salvador, French Guyana, Grenada, Guadeloupe, Guatemala, Guyana, Haiti, Honduras, Jamaica, Martinique, Mexico, Netherlands Antilles, Nicaragua, Panama,

3

Paraguay, Peru, Saint Kitts and Nevis, Santa Lucia, Saint Vincent and the Grenadines, Suriname, Trinidad and Tobago, Uruguay, Venezuela)

AFR

Sub-Saharan Africa(Angola, Benin, Botswana, British Indian Ocean Territory, Burkina Faso, Burundi, Cameroon, Cape Verde, Central African Republic, Chad, Comoros, Cote d'Ivoire, Congo, Democratic Republic of Congo, Djibouti, Equatorial Guinea, Eritrea, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Namibia, Niger, Nigeria, Reunion, Rwanda, Sao Tome and Principe, Senegal, Seychelles, Sierra Leone, Somalia, South Africa, Saint Helena, Swaziland, Tanzania, Togo, Uganda, Zambia, Zimbabwe)

A typical model application is constructed by specifying performance characteristics of a set of technologies and defining a Reference Energy System (RES) that includes all the possible energy chains that MESSAGE can make use of. In the course of a model run, MESSAGE determines how much of the available technologies and resources are actually used to satisfy a particular end-use demand, subject to various constraints (both technological and policy), while minimizing total discounted energy system costs over the entire model time horizon (1990-2110). It does this based on a linear programming, optimization solution algorithm. The representation of the energy system includes vintaging of the long-lived energy infrastructure, which allows for consideration of the timing of technology diffusion and substitution, the inertia of the system for replacing existing facilities with new generation systems, clustering effects (technological interdependence) and – in certain versions of the model – the phenomena of increasing returns (i.e., the more a technology is applied the more it improves and widens its market potentials). Combined, these factors can lead to “lock-in” effects (Arthur 1989; Arthur 1994) and path dependency (change occurs in a persistent direction based on an accumulation of past decisions). As a result, technological change can go in multiple directions, but once change is initiated in a particular direction, it becomes increasingly difficult to alter its course.

Important inputs for MESSAGE are technology costs and technology performance parameters (e.g., efficiencies and investment, variable, and O&M costs). For the scenarios included in this paper, technical, economic and environmental parameters for over 100 energy technologies are specified explicitly in the model. Costs of technologies are assumed to decrease over time as experience (measured as a function of cumulative output) is gained. For assumptions concerning the main energy conversion technologies see the following references: Riahi et al. (2007), Nakicenovic and Swart (2000), Riahi et al. (2012), and van Vliet et al. (2012). For information on carbon capture and storage technologies specifically, see Riahi et al. (2004).

MESSAGE is able to choose between both conventional and non-conventional technologies and fuels (e.g., advanced fossil, nuclear fission, biomass, and renewables), and in this respect the portfolio of technologies/fuels available to the model obviously has an important effect on the model result. In the version of the model used in this study, we consider a portfolio of technologies whose components are either in the early demonstration or commercialization phase (e.g., coal, natural

4

gas, oil, nuclear, biomass, solar, wind, hydro, geothermal, carbon capture and storage, hydrogen, biofuels, and electrified transport, to name just a subset). Notably, this portfolio includes bio-CCS, a technology that can potentially lead to negative emissions (i.e., permanent underground storage of CO2 which was originally pulled out of the atmosphere by photosynthesis). Exceedingly futuristic technological options, such as nuclear fusion and geo-engineering, are, however, not considered.

Other important input parameters for our modeling include fossil fuel resource estimates and potentials for renewable energy. For fossil fuel availability, the model distinguishes between conventional and unconventional resources for eight different categories of (oil, gas, coal) occurrences (Riahi et al. 2012; Rogner 1997). For renewable potentials we rely on spatially explicit analysis of biomass availability and adopt the assumptions discussed in Riahi et al. (2012).

Price-induced changes in energy demand (i.e., elastic demands) are also modeled in MESSAGE via an iterative link to MACRO, a top-down, macro-economic model of the global economy (Messner and Schrattenholzer 2000). Through an iterative solution process, MESSAGE and MACRO exchange information on energy prices, energy demands, and energy system costs until the demand responses are such (for each of the six end-use demand categories in the model: electric and thermal heat demands in the industrial, residential, commercial, and transportation sectors) that the two models have reached equilibrium. This process is parameterized off of a baseline scenario (which assumes some autonomous rate of energy efficiency improvement, AEEI) and is conducted for all eleven MESSAGE regions simultaneously. Therefore, the demand responses motivated by MACRO are meant to represent the additional (compared to the baseline) energy efficiency improvements and conservation that would occur in each region as a result of higher prices for energy services. The macro-economic response captures both technological and behavioral measures (at a high level of aggregation), while considering the substitutability of capital, labor, and energy as inputs to the production function at the macro level.

Further and more detailed information on the MESSAGE modeling framework is available, including documentation of model set-up and mathematical formulation (Messner and Strubegger 1995; Riahi et al. 2012) and the model’s representation of technological change and learning (Rao et al. 2006; Riahi et al. 2004; Roehrl and Riahi 2000).

2. Brief description of the MAGICC reduced-complexity global climate model

MAGICC (Model for the Assessment of Greenhouse-gas Induced Climate Change), version 5.3, has been used in this study to estimate the climate system impacts of the varying greenhouse gas emission trajectories of the scenarios in the ensemble. MAGICC is a reduced-complexity coupled global climate-carbon cycle model, in the form of a user-friendly software package that runs on a personal computer (Wigley 2008). In its standard form, MAGICC calculates internally consistent projections for atmospheric concentrations, radiative forcing, global annual-mean surface air temperature, ice melt, and sea level rise, given emissions trajectories of a range of gases (CO2, CH4, N2O, CO, NOx, VOCs, SO2, and various halocarbons, including HCFCs, HFCs, PFCs, and SF6), all of which are outputs from MESSAGE. The time horizon of the model extends as far back as 1750 and can make projections as far forward as 2400. The climate model in MAGICC is an upwelling-diffusion,

5

energy-balance model, which produces output for global- and hemispheric-mean temperature and for oceanic thermal expansion. Climate feedbacks on the global carbon cycle are accounted for through the interactive coupling of the climate model and a range of gas-cycle models. MAGICC has been used in all IPCC Assessment reports, dating back to 1990, and its strength lies in its ability to replicate the more complex global climate models that run on supercomputers. For our analysis, we use a version of the software that is consistent with the IPCC Fourth Assessment Report, Working Group 1, except that the model has been slightly modified to permit the explicit treatment of black and organic carbon (BC and OC) and their impacts on the global climate.1 The 550 ppm and the 450 ppm CO2-eq climate targets adopted by the EMF27 modeling protocol are implemented based on the increase in radiative forcing (RF) – from all greenhouse gases and forcing agents, excluding contributions from albedo change, nitrate aerosols, and mineral dust – compared to the pre-industrial era (1750). CO2-eq concentrations are then calculated from radiative forcing using the standard approximation formula: C0 exp(RF/α), where C0 = 278 ppm and α=5.35.

3. Brief description of the stylized transport module in MESSAGE-MACRO

The version of MESSAGE-MACRO employed in this study includes a quite stylized representation of the transport sector that essentially captures only fuel switching and price-elastic demands as mechanisms to respond to climate and energy policies. The following brief description elaborates the main characteristics of this transport module.

The model chooses between different final energy forms to provide useful energy for transportation. This decision is based primarily on the energy service costs by fuel, taking into account fuel prices at the final energy level and the respective final-to-useful energy conversion efficiencies. (In addition, “inconvenience” or “disutility” costs are applied to non-liquid fuels, in order to capture market adoption hurdles that MESSAGE-MACRO is not equipped to handle in its current form.) These conversion efficiencies vary by energy carrier. Useful energy demands (for the aggregate transportation sector of each region) are first specified in terms of ICE-equivalent, which therefore by definition have a conversion efficiency of final to useful energy of 1. Relative to that, the conversion efficiency of alternative fuels is higher, for example electricity in 2010 has a factor of ~3x higher final-to-useful efficiency than the regular oil-product based ICE. The assumed efficiency improvements of the ICE vehicles in the transportation sector, as well as mode-switching and other behavioral changes, are implicitly embedded in the baseline demand specifications. These come from the MESSAGE scenario generator2 (see Riahi et al. (2007) for more information).

1 We gratefully acknowledge Dr. Steve Smith of the Pacific Northwest National Laboratory (USA) for sharing a modified version of MAGICC (v5.3), which explicitly takes user-specified trajectories of BC and OC as inputs.2 Energy service demands are provided exogenously to MESSAGE; they are then adjusted endogenously based on energy prices thanks to the linkage with MACRO. There are seven demands in the stylized end-use version of the model, one of which is transport. These demands are generated using an R-based model called the scenario generator. This model uses country-level historical data of GDP per capita (PPP) and final energy use, as well as projections of GDP|PPP and population, to extrapolate the seven energy service demands into the future. The sources for the historical and projected datasets come from, for example, the World Bank, UN, OECD, and IEA. Using the historical datasets, the scenario generator conducts regressions that describe the historical relationship between the independent variable (GDP|PPP per capita) and several dependent variables, including total final energy intensity (MJ/2005USD) and the shares of final energy in several energy sectors (%).The historical data are also used in quantile regressions to develop global trend lines that represent each percentile of the cumulative distribution function (CDF) of each regressed variable. Given the regional

6

Additional demand reduction in response to price increases (e.g., in policy scenarios) then occurs via two mechanisms: (i) the fuel switching option (due to the fuel-specific relative efficiencies), and (ii) the linkage with the macro-economic model MACRO (see Section 1). Supplementary Figure 2 graphically illustrates the main components of the stylized transport sector representation in MESSAGE-MACRO.

Supplementary Figure 2. Schematic diagram of the stylized transport sector representation in MESSAGE-MACRO

To reflect limitations of switching to alternative fuels, for example as a result of limited infrastructure availability (e.g., rail network) or some energy carriers being largely unsuitable for certain transport modes (e.g., electrification of aviation), share constraints are imposed on certain energy carriers (e.g., electricity) and energy carrier groups (e.g., liquid fuels) of the transport sector. In addition, the diffusion speed of alternative fuels is limited to mimic known bottlenecks in the

regressions and global trend lines, final energy intensity and sectoral shares can be extrapolated forward in time based on projected GDP per capita. Several user-defined inputs allow the user to tailor the extrapolations to individual socio-economic scenarios. The total final energy in each region is then calculated by multiplying the extrapolated final energy intensity by the projected GDP|PPP in each time period. Next, the extrapolated shares are multiplied by the total final energy to identify final energy demand for each of the seven energy service demand categories. Finally, final energy is converted to useful energy in each region by using the average final-to-useful energy efficiencies reported by the IEA for each country.

7

supply chain, particularly those not explicitly represented in MESSAGE (e.g., non-energy related infrastructure). Both the share and diffusion constraints are typically parameterized based on transport sector studies that analyze such developments and their feasibility in much greater detail – the current paper being a prime example of this.

The cost-markup of the different fuel options in MESSAGE-MACRO can essentially be interpreted as incremental annualized non-fuel life-cycle costs of vehicles compared a reference technology (e.g., ICE vehicle). However, by making a few reasonable assumptions, this cost-markup can also be interpreted as a combination of incremental investment and operation and maintenance (O&M) costs. Reporting of this metric, particularly for BEVs, may help in the interpretation of the results and when comparing with other studies that employ an explicit representation of vehicle technologies.

While investment costs tend to be higher for BEVs than for conventional ICEs, a number of studies indicate that O&M costs (excluding fuel costs) for BEVs are actually lower by some 35-40% (Diez 2012; OTT 2002). Assuming O&M costs for ICE cars (depending on vehicle size/type and country) in the range of 500-1000 US$2005 per year (OTT 2002; RACV 2013), a vehicle lifetime of 15 years, a discount rate of 5%/yr and annual mileage of 10,000 to 20,000 km/vehicle/yr (as also assumed in Section 4 below), the resulting investment cost increment for BEVs compared to ICE vehicles in the 2020 period amounts to about 1950 to 4450 US$2005. Beyond 2020 we do not assume any further decrease in the cost increment. However, the variation of the upper bound on the electrification rate (i.e., the constraint modified in our “ET” sensitivity analysis) is consistent with the assumption that for the same cost-markup the vehicle range of BEVs varies. For example, the high electrification scenario ET75% implies that range limitations at the above cost increment (relative to ICE vehicles) are essentially not an issue, and therefore the light duty vehicle fleet can be almost completely electrified.

Finally, the demand for international shipping is modeled in a very simple way with a number of different energy carrier options (light and heavy fuel oil, biofuels, natural gas, and hydrogen). Demand is coupled to global GDP development with an income elasticity.

4. Assumptions for electric vehicle stock calculations in the main text

The calculation of the number of battery-electric vehicles (BEV) in the different transport electrification sensitivity cases, as discussed in the main text, depends entirely on vehicle efficiency assumptions and the assumed distance that a typical BEV is driven each year. For this reason, we give ranges in the paper. The lower-end estimates assume the following: 20,000 km/vehicle/yr (12,420 miles/vehicle/yr) and 0.185 kWh/km (0.300 kWh/mile). The upper-end estimates assume: 10,000 km/vehicle/yr (6,200 miles/vehicle/yr) and 0.125 kWh/km (0.200 kWh/mile). We further assume that light-duty vehicles are responsible for 70% of the incremental transport electricity demand after 2010. Note that if some of these vehicles were plug-in hybrid-electrics (PHEV), the calculated number of vehicles would be far higher.

5. Feedstock CO 2 emissions sensitivity analysis

8

Due to its versatile use in different sectors, bioenergy is a valuable, but scarce resource under scenarios of stringent climate mitigation. The current version of MESSAGE relies on bioenergy potentials based on van Vuuren et al. (2009), which takes into account a number of sustainability criteria to derive the respective potentials. In the standard EMF27 runs, global bioenergy supply is limited to about 145 EJ/yr in 2050 and 220 EJ/yr in 2100. In the limited bioenergy runs, the global bioenergy potential is limited to 100 EJ/yr excluding traditional biomass use (cf. Table 1 of the main text).

One of the principal reasons biofuels become so attractive as an industrial input – once biofuel demands in transport are relaxed – has much to do with how MESSAGE accounts for emissions from fossil-based feedstocks (coal, oil, natural gas, synthetic liquid fuels) used to produce industrial goods, such as poly-ethylene and other petro-chemicals. In reality, a part of the carbon contained in these feedstocks becomes “sequestered” in these products (e.g., in building materials of different types, in landfills, or simply in storage) for years, if not decades, before ultimately disintegrating into the lithosphere, hydrosphere, or atmosphere. In MESSAGE, however, the standard assumption is that 100% of the feedstock carbon is emitted to the atmosphere as CO2; hence, there is a strong incentive within the model to replace fossil-based feedstocks with biomass-based alternatives when biofuels are freed up from transport in the second half of the century, as is the case at higher levels of transport electrification. In such instances, the marginal value of biofuels in the industrial sector begins to approach that in transport. But how would this dynamic change if the assumption on feedstock emissions were to better reflect reality? We carried out a small experiment in MESSAGE to answer this question, varying (downwards) the assumed share of carbon contained in fossil feedstocks that ultimately ends up in the atmosphere.3 Our scenarios show that prior to 2050 biofuel demands in industry slightly decrease, a result that is perhaps expected given the decreased marginal value of biofuels as a low-carbon industrial feedstock in these sensitivity cases. After 2050, the dynamics become more complicated, owing to two competing effects: (i) fossil fuel feedstocks contribute to GHG emissions to a lesser extent and therefore serve as an alternative to biomass-based feedstocks, and (ii) biomass-based feedstocks are able to sequester carbon and thus convey the same benefits as bioenergy with CCS. With respect to allocation of biomass across sectors, these sensitivity cases lead us to conclude that the standard feedstock emission assumptions in MESSAGE are not overly prescriptive. That said, a robust finding of the analysis is that these assumptions have a considerable impact on the costs of mitigation and, ultimately, on the feasibility of ambitious climate targets (see Supplementary Table 2); thus, the topic deserves additional attention in future research.

3 Six additional scenarios were run for this sensitivity analysis. We took the 450 FullTech, 450 FullTech ET5% (low transport electrification), and 450 FullTech ET75% (high transport electrification) scenarios and simply assumed that instead of 100% either 50% or 0% of the feedstock carbon is emitted to the atmosphere as CO2.

9

Supplementary Table 2. Cumulative CO2 emissions from industrial feedstocks, consumption losses, and CO2 prices in the various feedstock emissions sensitivity cases discussed in the paper. The percentage values in the leftmost column represent the assumed share of carbon emissions from feedstocks that are released to the atmosphere in the particular scenario.

Feedstock sensitivity case

Feedstock CO2

emissions(Gton, 2010-2100)

Consumption losses

(as % ofbaseline GDP)

CO2 price(US$/tCO2)

450 FullTech

100% 269 0.7% 23

50% 154 0.6% 20

0% 2 0.5% 16

450 FullTech ET5%

100% 259 0.9% 30

50% 151 0.7% 26

0% 3 0.6% 21

450 FullTech ET75%

100% 256 0.5% 17

50% 148 0.5% 14

0% 6 0.4% 12

6. Measuring diversity through the Shannon-Wiener index

The Shannon-Wiener diversity index (SWDI), referred to in the main text, has been used in previous studies to measure different aspects of energy system diversity (Jansen et al. 2004; Kim et al. 2009; Kruyt et al. 2009; Riahi et al. 2012; Stirling 1994). The exact value of the SWDI has little intuitive meaning; rather, the indicator’s true explanatory power rests on its ability to shed light on relative changes in diversity over time and across countries/regions/sectors. In this sense, the higher the diversity indicator, the greater the diversity.

SWDI=−∑j

( p j ∙ ln p j )

where:- pj: share of component j in the total mix

7. Further details on mitigation costs and scenario feasibility

Similar to the availability of supply-side technologies and stringency of energy efficiency and conservation efforts, transport sector electrification has important implications for the feasibility of the ambitious 450 ppm CO2-eq stabilization target. As shown in Table 1 of the main text, some of the 450 ppm MESSAGE scenarios from the standard EMF27 set turned out to be quite close to the feasibility threshold of the model. By varying the assumptions on transportation sector electrification, two of these scenarios became feasible (in the case of higher electrification), and one became infeasible (lower electrification). Most notably, the 450 NoCCS case turned feasible after

10

slightly increasing the maximum electrification rate to 35% (instead of 25% in the regular setup), while the 450 ppm scenario with limited renewables and bioenergy supply (450 Conv) became feasible at the highest maximum electrification rate that we explored, 75%. In contrast, the 450 scenario with limited bioenergy (450 LimBio) turned infeasible at our lowest maximum electrification rate of 5%. For all other scenarios, feasibility did not change.

The scale of electrification of the transport sector, however, has a major impact on the total costs of climate mitigation, in particular of meeting a stringent stabilization target like 450 ppm CO2-eq. Supplementary Figure 3 shows for the 450 FullTech family of scenarios two different mitigation cost measures – CO2 prices in 2020 and consumption losses over the century (2010-2100) – and how these measures vary depending on the cumulative share of final energy demand in transport that is met by electricity between 2010 and 2100.4 (As a point of reference, this share is approximately 17% for the standard 450 FullTech scenario.) Carbon prices give an indication of the marginal cost of abatement, whereas consumption losses are more of an average, bulk measure. In both cases, mitigation costs dramatically decrease at higher electrification levels; or put the other way, if the potential for electrification in transport is restricted, then the system-wide costs of climate mitigation (considering all energy producing and consuming sectors) are likely to be significantly higher. These dynamics are due to the ripple effects that transport electrification has on fuel consumption in the industry and buildings sectors. More specifically, by obviating the need for biofuels in transport, electrification frees up valuable biomass resources that can then be used in other parts of the energy system (see Section 3.3 of the main text).5

To put the sensitivity of mitigation costs to transport electrification into context, it is useful to compare to a couple of the reduced technology portfolio cases in the standard set of EMF27 scenarios (Supplementary Figure 3). Whereas the carbon price (consumption loss) in the 450 FullTech scenario is $23/tCO2 (0.7% loss), the constrained renewables 450 LimSW and constrained bioenergy 450 LimBio scenarios see mitigation costs rising to $31/tCO2 (0.9% loss) and $52/tCO2 (1.2% loss), respectively. Thus, restrictions to the penetration of electric vehicles across the various transport sub-sectors appear to have as great of an impact on mitigation costs as restrictions to wind and solar electricity generation on the supply side.

4 This cumulative share is calculated as follows: annual final electricity demands in transport are cumulated from 2010 to 2100 and subsequently divided by the cumulative amount of final energy consumed in the entire transport sector over the same time period.5 An important caveat to this discussion is that the costs of electric vehicles and their requisite recharging infrastructure are not completely accounted for in MESSAGE. While these costs will indeed have an impact on the uptake of advanced vehicles by consumers in the future, what is important within the decision-making framework of the model is the incremental cost of electric-drive technologies relative to their conventional counterparts. We therefore implicitly assume that electric technologies and infrastructure become increasingly cost-competitive in different applications, irrespective of any carbon pricing.

11

0.2%

0.4%

0.6%

0.8%

1.0%

1.2%

1.4%

1.6%

10

15

20

25

30

35

40

45

50

55

60

65

0% 5% 10% 15% 20% 25% 30% 35% 40% 45%

Cons

umpt

ion L

osses

(% of

basel

ine G

DP)

Carb

on Pr

ice (U

S$20

05/tC

O2)

Electrification Share in Transport

left axis

right axis

450 FullTech

450 NoCCSET35%

450 LimBio

450 NucOff

450 LimSW450 EERE

450 LowEI

Supplementary Figure 3. Mitigation costs in each of the electrified transport variants of the 450 FullTech scenario (dots connected by lines) and in a selected sub-set of 450 ppm technology variants (other dots). Horizontal axis shows the cumulative share of final energy demand in transport that is met by electricity (2010-2100). Blue dots connected by line show global carbon prices in 2020, which grow with the discount rate (5% p.a.) throughout the century (left axis). Orange dots connected by line show global consumption losses as a share of globally-aggregated GDP (in both cases, annual values from 2010-2100 are discounted back to 2010 at the discount rate and then cumulated).

12

References

Arthur WB (1989) Competing Technologies, Increasing Returns, and Lock-in by Historical Events. Econ J 99:116-131.

Arthur WB (1994) Increasing Returns and Path Dependence in the Economy. Ann Arbor, University of Michigan Press, MI.

Diez W (2012) Implications of electromobility for the service business of car dealers (in German: Auswirkungen der Elektromobilität auf das Servicegeschäft vertragsgebundener Autohäuser). Working Paper. Institute for Automotive Research (IFA), Geislingen.

Jansen JC, Arkel WGv, Boots MG (2004) Designing indicators of long-term energy supply security. Energy Research Centre of the Netherlands (ECN), p. 35.

Kim HJ, Jun E, Chang SH, Kim WJ (2009) An assessment of the effectiveness of fuel cycle technologies for the national energy security enhancement in the electricity sector. Annals of Nuclear Energy 36:604-611.

Kruyt B, van Vuuren DP, de Vries HJM, Groenenberg H (2009) Indicators for energy security. Energy Policy 37:2166-2181.

Messner S, Schrattenholzer L (2000) MESSAGE-MACRO: linking an energy supply model with a macroeconomic module and solving it iteratively. Energy 25:267-282.

Messner S, Strubegger M (1995) User's guide for MESSAGE III, Working Paper WP-95-069. International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria, p. 164.

Nakicenovic N, Swart R (2000) IPCC Special Report on Emissions Scenarios. Cambridge University Press, Cambridge.

OTT (2002) Quality Metrics 2003. Office of Transportation Technologies (OTT), U.S. Department of Energy, Washington, D.C.

RACV (2013) RACV's car owning and operating costs guide (http://www.racv.com.au/wps/wcm/connect/racv/Internet/Primary/my+car/advice+_+information/vehicle+operating+costs). Last accessed on September 20, 2013. Royal Automobile Club of Victoria.

Rao S, Keppo I, Riahi K (2006) Importance of technological change and spillovers in long-term climate policy. The Energy Journal, Special Issue: Endogenous Technological Change and the Economics of Atmospheric Stabilisation 27.

Riahi K, Dentener F, Gielen D, Grubler A, Jewell J, Klimont Z, Krey V, McCollum D, Pachauri S, Rao S, van Ruijven B, van Vuuren DP, Wilson C (2012) Chapter 17 - Energy Pathways for Sustainable Development. Global Energy Assessment - Toward a Sustainable Future, Cambridge University Press, Cambridge, UK and New York, NY, USA and the International Institute for Applied Systems Analysis, Laxenburg, Austria, pp. 1203-1306.

Riahi K, Grübler A, Nakicenovic N (2007) Scenarios of long-term socio-economic and environmental development under climate stabilization. Technological Forecasting and Social Change 74:887-935.

Riahi K, Rubin ES, Schrattenholzer L (2004) Prospects for carbon capture and sequestration technologies assuming their technological learning. Energy 29:1309-1318.

Roehrl RA, Riahi K (2000) Technology Dynamics and Greenhouse Gas Emissions Mitigation: A Cost Assessment. Technological Forecasting and Social Change 63:231-261.

Rogner HH (1997) An assessment of world hydrocarbon resources. Annu Rev Energ Env 22:217-262.Stirling A (1994) Diversity and ignorance in electricity supply investment : Addressing the solution

rather than the problem. Energy Policy 22:195-216.van Vliet O, Krey V, McCollum D, Pachauri S, Nagai Y, Rao S, Riahi K (2012) Synergies in the

Asian energy system: Climate change, energy security, energy access and air pollution. Energy Economics.

van Vuuren DP, van Vliet J, Stehfest E (2009) Future bio-energy potential under various natural constraints. Energy Policy 37:4220-4230.

Wigley TML (2008) MAGICC/SCENGEN 5.3: User Manual (version 2). National Center for Atmospheric Research, Boulder.

13