Energy Policy 77 (2015) 118–130

Contents lists available at ScienceDirect

Energy Policy

http://d0301-42

n CorrE-m

mzakermccollu

journal homepage: www.elsevier.com/locate/enpol

Achieving California's 80% greenhouse gas reduction target in 2050:Technology, policy and scenario analysis using CA-TIMES energy eco-nomic systems model

Christopher Yang a,n, Sonia Yeh a, Saleh Zakerinia a, Kalai Ramea a, David McCollum b

a Institute of Transportation Studies, University of California, 1715 Tilia Street, Davis, CA 95616, USAb International Institute for Applied Systems Analysis, Laxenburg 2361, Austria

H I G H L I G H T S

� We model the California Energy System to 2050 under policy and technology scenarios.

� The model optimizes technology and resource investments to meet emissions targets.� Deep emissions cuts (474%) are achieved across all reduction scenarios.� Carbon capture enables negative emission biofuels and allows more petroleum use.� Greenhouse gas mitigation cost is small compared with total economic activity.

a r t i c l e i n f o

Article history:Received 11 June 2014Received in revised form18 November 2014Accepted 4 December 2014

Keywords:Carbon emissionsOptimizationElectricityTransportationFuelsEnergy services

x.doi.org/10.1016/j.enpol.2014.12.00615/& 2014 Elsevier Ltd. All rights reserved.

esponding author.ail addresses: ccyang@ucdavis.edu (C. Yang), sinia@ucdavis.edu (S. Zakerinia), kramea@ucdam@iiasa.ac.at (D. McCollum).

a b s t r a c t

The CA-TIMES optimization model of the California Energy System (v1.5) is used to understand howCalifornia can meet the 2050 targets for greenhouse gas (GHG) emissions (80% below 1990 levels). Thismodel represents energy supply and demand sectors in California and simulates the technology andresource requirements needed to meet projected energy service demands. The model includes as-sumptions on policy constraints, as well as technology and resource costs and availability. Multiplescenarios are developed to analyze the changes and investments in low-carbon electricity generation,alternative fuels and advanced vehicles in transportation, resource utilization, and efficiency improve-ments across many sectors. Results show that major energy transformations are needed but thatachieving the 80% reduction goal for California is possible at reasonable average carbon reduction cost($9 to $124/tonne CO2e at 4% discount rate) relative to a baseline scenario. Availability of low-carbonresources such as nuclear power, carbon capture and sequestration (CCS), biofuels, wind and solargeneration, and demand reduction all serve to lower the mitigation costs, but CCS is a key technology forachieving the lowest mitigation costs.

& 2014 Elsevier Ltd. All rights reserved.

1. Introduction

California is one of the leading nations/states around the worldin developing policy instruments to address the issue of climatechange. In 2005, it set an aspirational long-term goal of reducinggreenhouse gas (GHG) emission 80% below 1990 levels by 2050.The state subsequently implemented legislation setting a bindingtarget that GHG emissions be brought back down to the 1990 level

lyeh@ucdavis.edu (S. Yeh),vis.edu (K. Ramea),

by 2020. This path to achieving the near-term GHG goal includedimplementation of a number of policy mechanisms including thecap-and-trade program, vehicle efficiency and fuel carbon stan-dards, and others (CARB, 2008). There is significant uncertainty asto how to achieve the deep reductions in GHG emissions that areneeded to stabilize atmospheric concentrations of GHGs, whatthey would cost and what policy measures would be needed. TheCA-TIMES model was developed to help answer some of thesequestions about how to achieve the long-term GHG emissionsgoals by 2050 while still meeting the demand for energy(McCollum et al., 2012).

C. Yang et al. / Energy Policy 77 (2015) 118–130 119

1.1. Literature review

A number of scenario and energy modeling analyses have in-vestigated how California can achieve substantial reductions inGHG emissions. Several studies have been performed to under-stand the role of efficiency, advanced technologies and alternativefuels in reducing transportation GHG emissions (CARB, 2009,2012; Leighty, 2012; Yang et al., 2009). A number of scenariostudies have looked across multiple sectors to understand thevarious roles of each in contributing to the GHG targets (CCST,2011; Greenblatt, 2013; Wei et al., 2013; Williams et al., 2012), butthese studies do not explicitly take costs into consideration in anintegrated systems modeling approach. Other more detailedmacroeconomic models of California (e.g. (ICF, 2010; Roland-Holst,2008)) only look to the near-term 2020 targets and have not in-vestigated the very-low GHG, longer-term timeframe of 2050.Other modeling and scenario studies have looked at the dec-arbonization at the global level (GEA, 2012; IEA, 2012), whichprovides a useful look at global trends, energy use and interactionsbetween regions. These studies provide useful points of compar-ison for CA-TIMES modeling and highlight the benefits of using anintegrated cost optimization model to explore technology, re-source and policy scenarios for California's energy future.

1.2. Study goals

The goal of the CA-TIMES model and this particular paper is toprovide an integrated analysis across energy demand and supplysectors for California's policy makers and stakeholders, specificallyfor meeting the 80% GHG reduction goal for 2050. This integrationis important for understanding the role of different sectors,

Fig. 1. Schematic representation of CA-TIMES model. The numbers refer to three categorand Lifecycle (sources 1þ2þ3) emissions explained in Section 3.5.

technologies, resources and policies in the transition to a dec-arbonized energy system, and critical tradeoffs across sectors suchas limits on low-carbon resources, all while attempting to mitigateGHGs in the least-cost manner.

2. Methods

2.1. Modeling overview

CA-TIMES is a bottom up, technologically rich, integrated eco-nomic optimization model of the California Energy System basedupon the MARKAL/TIMES modeling framework (Loulou and Lab-riet, 2007). CA-TIMES covers all sectors of the energy economy ofthe state (Fig. 1), including energy supply (energy resources andfeedstocks (oil, gas, and biomass), fuel production/conversion, fuelimports/exports, electricity production, and fuel delivery), anddemand (i.e., residential, commercial, industrial, transportation,and agricultural end-use sectors). Note that CA-TIMES does notcurrently model or track emissions from non-energy sectors (e.g.landfills, livestock, cement production, refrigerants, etc.).

This version of CA-TIMES (v1.5) is an improved version of amodel described in a previous publication (McCollum et al., 2012).Full documentation of CA-TIMES model and assumptions can befound at Yang et al. (2014).

2.2. Modeling methods

CA-TIMES is primarily run as a cost-minimization model inmost of the scenarios described in this paper. In this approach, theobjective of the model is to meet exogenously specified energy

ies of emissions sources: Included (sources in dashed box 1), Overall (sources 1þ2),

C. Yang et al. / Energy Policy 77 (2015) 118–130120

services demand at the least system cost (i.e. the discounted netpresent value (NPV) of all private economic costs in the model)subject to a series of constraints (used to represent policies as wellas limits on technology and resources). This version of CA-TIMESdoes not include externality or social costs of energy, such aspollution health impacts, ecosystem impacts, or damage asso-ciated with climate change. The decision variables in the model aretechnology investment and operating decisions. This approach isrepresentative of a single, global decision maker with perfectforesight.

“Elasticity scenarios” are also run in which the demand forenergy services is responsive to price changes and total surplus(consumer plus producer surplus) is maximized (Loulou and La-vigne, 1996). Thus, emissions reductions can also be achievedthrough lower demand for energy services and this paper includestwo exploratory “elasticity scenarios”. Elasticity values for servicedemands are summarized in Table SI.1 of the Supportinginformation.

Like many high-level energy models, CA-TIMES models an ag-gregate representation of all consumers and producers. In order toimprove the behavioral realism of the system optimization basedsolely on economic costs, a number of modeling techniques areemployed, including growth constraints (to limit “winner-take-all”behavior), share constraints (to account for consumer hetero-geneity), and hurdle rates (to account for non-cost barriers toadoption of new technologies) (Gillingham et al., 2012; Greene,2011; Yang et al., 2014). See Supporting Information.

2.3. Energy supply sectors

Energy and fuels supply sectors in CA-TIMES cover the extrac-tion of primary energy resources (both in and out of state) as wellas production/conversion and transport of processed fuels andenergy carriers. A primary focus of the model is the representationof transportation fuels production including oil refineries, bio-re-fineries, synfuel and hydrogen production plants and electricpower plants (see Fig. 1). We use exogenous price projections foroil and gas (from AEO's Reference case (AEO, 2013)) as California isassumed to be a price-taker for these resources. More details onthe supply sectors can be found in the companion report.

2.4. Biomass supply and biorefineries

Biomass supply is represented by a series of supply curves to-taling 1767 PJ in 2050 for 12 different feedstock types (Fig. SI.1),broken into five primary categories: woody biomass, herbaceousbiomass, municipal solid waste (MSW), yellow grease (usedcooking oils) and tallow (animal fats) (Parker et al., 2010). Themodel also allows for imports of biofuels from outside the state,including ethanol (from various feestocks), bio-diesel, and bio-based synthetic, drop-in fuels.

Several different types of biorefineries are modeled in CA-TIMES, including cellulosic ethanol production (via biochemical orthermochemical pathways), bio-oil pyrolysis, and Fischer–Tropsch(FT) biomass poly-generation plants. These FT polygen plantsbased upon (Kreutz et al., 2008) can produce synthetic gasoline,diesel, and jet fuel and can take advantage of carbon capture andsequestration (CCS) to produce biofuels with negative carbonintensity.

2.5. Hydrogen infrastructure

Hydrogen production, delivery and refueling are modeled inCA-TIMES based upon the following technologies: Central pro-duction via coal gasification, natural gas steam reforming, biomassgasification and water electrolysis, delivery via gaseous pipelines

or liquid hydrogen trucks, and distributed production via smallscale steam reforming or water electrolysis. The first three centralproduction options can incorporate CCS. The costs and efficienciesfor these plants comes primarily from the DOE H2A database(H2A, 2005). Important caveats to this relatively simple re-presentation of H2 (and biofuel) infrastructure include the lack ofeconomies of scale and spatial detail in fuel production and de-livery, which will tend to underestimate early fuel infrastructurecosts. These topics have been addressed in other work (Yang andOgden, 2013).

2.6. Electricity generation

Electricity supply in the CA-TIMES model is modeled as gen-eration and demand at a single node with average values fortransmission and distribution losses. While the state currentlyimports about 30% of electricity demand, CA-TIMES does not cur-rently model the electricity systems of neighboring regions. Thus,in order to adequately account for costs, technology mix, renew-able fraction and GHG emissions of California's electricity demand,all electricity demand after 2025 is represented as supplied fromdedicated power plants, though not necessarily built within thestate. Power plant costs and efficiencies come primarily from AEOand the SWITCH model (AEO, 2013; Nelson et al., 2012).

CA-TIMES has a representation of spatially disaggregated windand solar resources, and transmission costs based upon a study of59 Competitive Renewable Energy Zones (CREZ) in California andin the Western US, Canada and Mexico (RETI, 2009). They areaggregated into 23 CREZ groups (CGs) and the hourly availabilityprofile for renewable wind and solar resources for each CG isspecified based upon Allan (2011).

CA-TIMES uses 48 sub-annual timeslices, which consist of 6“seasonal” divisions (i.e. 2 month periods) and 8 “hourly” divisions(i.e. 3 h periods in each season). The use of timeslices is especiallyrelevant to the electricity system as this level of temporal resolu-tion can represent much (but certainly not all) of the importantdecisions and variability in supply and demand. Electricity storage(beyond existing pumped hydro) and demand response are notcurrently included in the model.

2.7. Energy demand sectors

The transportation, residential and commercial end-use sectorsare represented in the most detail with specified service demandsand a set of potential technologies that can be chosen by themodel to meet these service demands, depending on costs, energyusage and associated emissions. Technology choices in these sec-tors can reduce emissions via improved efficiency, switching tolower-carbon fuels or both (such as vehicle and building elec-trification). The representation of the industrial and agriculturalsectors is very simplified, represented as exogenous demand sce-narios for fuels and electricity.

2.8. Transportation sector

The transportation sector, including demands and existing andfuture technologies, is specified in the following vehicle cate-gories: light-duty cars and trucks, medium duty trucks, heavy-duty trucks, buses, passenger and freight rail, aviation (intrastate,interstate and international passenger and freight and personalaircraft), marine (international shipping and harbor craft), agri-cultural and off-road vehicles. A wide range of sources were usedto develop the demand (BTS, 2010; CARB, 2004, 2005, 2007, 2011;FAA, 2007; NTD, 2010) and technology data (AEO, 2013; Boeing,2007; EPA, 2014; IEA, 2008; NRC, 2013). More details on the traveldemands and technology performance and costs are found in

C. Yang et al. / Energy Policy 77 (2015) 118–130 121

Appendix B of Yang et al. (2014). A scenario with reduced vehiclemiles traveled (VMT) is also analyzed to understand the role oftravel demand reduction in reducing fuel usage and GHGemissions.

2.9. Residential end-use sector

The residential sector is represented by the demands andtechnologies for thirteen energy services, including space heating,space cooling, water heating, lighting, cooking, refrigeration,clothes washing, clothes drying, dish washing, freezing, TV, poolpumps, and miscellaneous in two building types (single and multi-family housing). Service demands are driven by projections ofpopulation and household growth, housing type, building envel-ope parameters, appliance utilization rates (the expected use perappliance) and saturation rates (the expected number of appli-ances per household) (California Department of Finance, 2013;McCarthy et al., 2008; Rufo and North, 2007). Space heating, spacecooling, water heating, lighting and miscellaneous represent theservice demands with the greatest energy use in 2010, while spacecooling, lighting, and pool pumps represent the service demandswith the greatest expected growth to 2050.

These service demands are satisfied by end-use technologies(i.e. appliances) that vary in efficiency, costs and availability andconsume a range of fuels (electricity, natural gas, LPG, solar andwood biomass) from the CA-TIMES supply sector.

2.10. Commercial end-use sector

Energy usage in the commercial sector is disaggregated into 12building types (restaurants, retail, grocery stores, schools, collegesand universities, hospitals, hotels, small offices, large offices,warehouses, refrigerated warehouses and miscellaneous) and9 different energy services (heating, cooling, ventilation, waterheating, cooking, refrigeration, lighting, office equipment andmiscellaneous). Future energy services for each building type isscaled by growth (�1.3% annual) in the floorspace (CEC, 2011) andare expected to increase between 50 and 100% between 2010 and2050. Time profiles of service demands are constructed fromsurvey data (ITRON, 2006). Technology cost and efficiency para-meters come from NEMS inputs (NEMS, 2013). Detailed listof technology and assumptions are provided in Appendix C ofYang et al. (2014).

2.11. Industrial and agricultural end-use sectors

The industrial sector is a significant challenge to model in CA-TIMES because of the uncertainty in the future composition andscale of the industrial sector to 2050. Industrial activity haschanged dramatically over the last 50 years and we present ag-gregated industrial energy demands for electricity, natural gas,petroleum and other products as an input scenario, rather than theconsequence of endogenous decisions within the model. Theagricultural sector is also treated similarly with specified demandsfor natural gas, electricity and petroleum fuels. As a result, costsfor mitigating emissions from these sectors are not represented,beyond the cost to supply this energy.

We develop two scenarios for energy use in these sectors, onefor scenarios where GHG emissions are not capped and another forthe GHG reduction scenarios. These fuel demands come primarilyfrom scenarios in the literature (IEA, 2010; McCarthy et al., 2008).In the GHG scenario, significant gains in efficiency and elec-trifications are assumed for the industrial sector, reducing totalenergy use 13% relative to the BAU scenario and increasing elec-tricity's share of total industrial energy use from 30% in BAU to 80%in GHG scenarios by 2050. This assumption is quite optimistic, and

reflects a reduction in share of industrial process heating (as pet-roleum extraction and refining decline in the GHG scenario) aswell as a large shift of the remaining industrial heating demandsin manufacturing (i.e. boilers and process heating) to electricity.

2.12. Emissions

Three categories of emissions are differentiated in the CA-TIMES model. The first category, Included emissions accounts forall energy-related GHGs produced from fuel conversion and pro-cessing (e.g., refinery emissions, electric power plants, biofuel orH2 production plant emissions), transport and delivery, and com-bustion activities within the boundaries of California's energysystem, as well emissions from out-of-state electricity productionfor California (i.e. all emissions from activities in the dashed box(1) in Fig. 1). The second category, Overall emissions, includes allIncluded emissions plus emissions associated with the interstateand international aviation and marine trips refueled in California(sources 1 and 2 in Fig. 1). The 3rd category is Lifecycle emissions,which consists of all emissions sources in Overall emissions plusupstream emissions associated with producing and transportingenergy resources from out of state for use in California and in-cluding indirect land use change emissions for biofuels (sources1,2 and 3 in Fig. 1). A carbon cap on Included emissions is con-sistent with California's carbon regulation (CARB, 2009, 2008)with the main difference being that CA-TIMES (and the Includedemissions category) does not track non-energy emissions.

2.13. Scenarios

The value of the CA-TIMES model is to vary the technology,resource and policy-related inputs and analyze the results in thecontext of these assumptions. This section will focus primarily onpolicy and technology choices that influence the least-cost path-ways for reference (BAU) futures (which do not attempt to miti-gate GHG emissions beyond currently implemented policies to2020) and futures that attempt to achieve the state's target of an80% reduction in GHG by 2050 (GHG scenarios). Several GHGscenario variants are analyzed in which policies, technology andresource availability, and technology and resource costs are varied,in order to understand how the transition to a low-carbon econ-omy in California could be different if the potential of certaintechnologies and resources is substantially restricted or enhanced.Scenario results are not forecasts of the future, but rather visionsof what could happen, if the assumptions in the model were rea-lized and real-world decision makers behaved in a manner thatminimizes societal costs.

2.14. Reference (BAU) scenarios

The CA-TIMES Reference scenario (BAU) describes one potentialdevelopment path of California's energy system over the nextseveral decades without constraints on GHG emissions (even in2020). The Intergovernmental Panel on Climate Change (IPCC)develops and analyzes a large number of Reference Case scenarios(O’Neill et al., 2014), but we chose to focus on just two BAU sce-narios to keep the comparisons manageable.

Policy is an important driver in the development of the energysystem, especially in the Reference scenarios. The BAU scenariosinclude representations of the fuel economy standards (CAFE),renewable portfolio standard (RPS), zero emission vehicle (ZEV)mandate, low carbon fuel standard (LCFS) policy, and other taxesand credits/subsidies for biofuels, renewable electricity and elec-tric vehicles.

A separate Reference scenario is run that assumes a lower levelof travel demand (BAU-LoVMT), in order to explore the impact of

Cap

on

Inst

ate

GH

G e

mis

sion

s (M

MT/

yr)

Fig. 2. Carbon cap by year for the GHG policy scenarios, GHG-Line and GHG-Step.Both require 80% reduction from 1990 emissions levels by 2050.

C. Yang et al. / Energy Policy 77 (2015) 118–130122

the state's VMT reduction measures (SB375). The LoVMT as-sumption corresponds to a suite of strong travel demand man-agement (TDM) policies dealing with transit, land use, and auto-pricing. These lead to a reduction in light-duty vehicle (LDV) VMTby 24%, and heavy-duty vehicle (HDV) and medium-duty vehicle(MDV) VMT by 10% in 2050 relative to the Reference case.

And because VMT reduction can lead to large cost savings (frompurchasing fewer vehicles and using less fuels that also leads toless needs for investment in fuel infrastructure in production anddistribution) and lower energy use and emissions, comparing GHGscenarios to both BAU and BAU-LoVMT provides a useful bench-mark to understand how shifting travel demands influences costsand emissions reduction potentials.

2.15. GHG reduction (GHG) scenarios

The GHG scenarios describe a number of different futures inwhich the California's energy system undergoes substantial dec-arbonization as a result of a carbon policy that aims to reduceemission by 80% below 1990 levels in 2050. Again, these scenariosshould not be mistaken as predictions of what will happen as aresult of strong climate policy, but rather as individual visions ofwhat could feasibly happen, under the large set of technologicaland policy assumptions embedded in the model. While theprobability that the future will be identical to any of our scenariosis essentially zero, their value is in understanding the role thattechnologies, resources and policies can play in influencing thedevelopment of the future energy system that can meet the GHGtargets in 2050.

Travel demand in the GHG scenarios is the same as the lowVMT Reference scenario (i.e. BAU-LoVMT). Since policies are amongthe most important drivers for inducing dramatic energy systemtransitions, the GHG scenarios include all policies that are re-presented in the Reference scenarios as well as an additionalcarbon constraint to meet the 2050 GHG target. The primary GHGscenarios do not include new nuclear and CCS technologies (seeSupporting information), given the significant uncertainty abouttheir technical and political viability for large-scale deployment inCalifornia during the modeling period (CCSRP, 2010; Greenblattet al., 2012; Richter et al., 2011). However, sensitivity scenarios inwhich these two key technologies are available to understandtheir potential role in GHG mitigation.

The ARB inventory for California for 1990 (426.59 million me-tric tons of CO2 equivalent, i.e. MMTCO2e1) includes some non-energy emissions, while CA-TIMES only models and tracks emis-sions from the energy usage (390.84 MMTCO2e).2 The cap onemissions in CA-TIMES covers only the Included emissions, whichis all fuel and energy combustion and production emissions in thestate, plus all emissions associated with electricity generation(even for imported electricity), but excluding interstate and in-ternational aviation and marine fuel combustion.

For simplicity and transparency, two types of constraints forthe GHG target are used in the CA-TIMES model. The first type ofconstraint is called the “Step” cap in which a cap is held at the2020 target (1990 levels) between 2020 and 2050 but thendropped to 80% below 1990 emissions in 2050. The second type ofconstraint is a declining carbon cap – specifically, a straight-linetrajectory from 2020 to 2050 is assumed and called “Line”. Fig. 2

1 ARB recently updated their GHG inventory to reflect a higher GWP for me-thane, consistent with IPCC's Fourth Assessment Report (AR4) to 431 MMTCO2e,but CA-TIMES currently uses the older inventory.

2 CA-TIMES ignores approximately 42 MMTCO2e of emissions from cementproduction and other industrial processes, emissions from agriculture and wastetreatment and fugitive emissions from oil, gas production, as well as nearly7 MMTCO2e of negative emissions from natural sinks and sequestration.

shows the numerical values for the GHG emissions caps assumedin both constraints.

While a linear cap or some sort of interim emission reductiongoal might be a more likely approach from the state between 2020and 2050 (CARB, 2014), the “Step” GHG cap scenario provides theadvantage of exploring the “optimal” (i.e. least-cost) emissionspath to meet the 2050 target and the resulting types and timing ofinvestments in low-carbon resources and energy supply and de-mand technologies given our assumptions of discount rate andexogenous assumptions of cost trajectories.

2.16. Scenario variations

Table 1 shows a brief summary of the key sensitivity para-meters for the GHG scenario variations that were developed tounderstand the impact of changing input assumptions on theability of the state to meet the GHG target and the cost and mix ofoptions used to achieve the GHG reductions (also see main report(Yang et al., 2014) for more details). Future work will explore manymore variations on input assumptions, resource availability, andpolicy constraints, to further explore the sensitivity of CA-TIMES tothese model inputs.

3. Results

3.1. GHG emissions

Fig. 3 shows the comparison of Included GHG emissions be-tween the BAU, GHG-Step and GHG-Line scenarios. Emissions in theBAU scenario decline slightly to 2025 and then rise slightly after-wards. The 2020 GHG emissions are 383 MMTCO2e/yr, below the2020 target of 390 MMTCO2e/yr, even though the AB32 cap onemissions is not a constraint in this Reference scenario. This islargely due to the fact that we have incorporated most of the GHGpolicies aiming to achieve the state's 2020 goals in our BAU sce-narios. In 2050, GHG emissions are approximately equivalent to1990 levels (388 vs. 391 MMTCO2e), due to a combination of effi-ciency improvements and carbon intensity reductions in the faceof growing energy service demands. Overall emissions (includinginterstate/international aviation and marine) in the BAU scenarioare 437 MMTCO2e. When adding the 80% GHG emissions reduc-tion constraint (Step or Line), both GHG scenarios can only achieve74.6% GHG reductions below 1990 (not quite enough to meet the80% target-see Supporting information), due to limits on theavailability and adoption of low-carbon resources and technolo-gies within the model (at any price) across all supply and demand

Table 1Sensitivity parameters for GHG scenarios. “GHG-S” represents scenarios with a “Step” constraint on GHG emissions between 2020 and 2050 as described in the previoussection.

Sensitivity case Description

Nuclear power plant availability (GHG-S-Nuclear) New nuclear power plants can be built.Carbon capture and sequestration (CCS) availability(GHG-S-CCS)

CCS is available on many different types of plants.

Nuclear power and CCS availability (GHG-S-NucCCS) Nuclear power plants and CCS technologies are available.Rapid deployment of wind and solar (GHG-S-HiRen) Utility wind and solar power can ramp up 33-100% faster than BAU.Increased biomass supply (GHG-S-HiBio) This scenario assumes that the available biomass supply is 3600 PG (double that of the base case) at the same

costs as in the base GHG case.High and low oil and gas prices (GHG-S-HiOilGas)(GHG-S-LoOilGas)

High price sensitivity case: Uses prices from AEO's (2013) high oil price scenario.Low Price Sensitivity case: Uses prices from AEO’s (2013) high oil/gas resource scenario.

Elastic demand (GHG-S-Elas1)(GHG-S-Elas2)

Sensitivity case (Elas1): Assumes elastic demand in response to price changes. Elas1 allows for reductions inthe investment in end-use appliances as demand is reduced.Sensitivity case (Elas2): Assumes elastic demand in response to price changes. Elas2 constrains the level ofend-use appliances to be equal to the non-elastic scenario even as demand is reduced.

Gre

enho

use

Gas

Em

issi

ons

(MM

TCO

2e/y

r)

BAU GHG-Step GHG-Line

Transportation

Residential

IndustrialCommercial

Fig. 3. Included GHG emissions from 2010 to 2050 by sector for the BAU, GHG-Step and GHG-Line scenarios.

Fig. 4. Included GHG emissions for the four primary CA-TIMES scenarios.

3 Carbon intensity values are calculated on a HHV basis and do not include anenergy efficiency ratio (EER) for hydrogen and electricity fuels consistent with thestate’s Low Carbon Fuel Standard

C. Yang et al. / Energy Policy 77 (2015) 118–130 123

sectors. In the GHG-Step scenario, emissions decline only slightlywith respect to the BAU scenarios to 2030 and then emissions startto diverge significantly after 2030. Transportation continues tomake up the majority of emissions. The GHG-Line scenarioachieves the exact same level of GHG emissions reduction in 2050as GHG-Step, and the distribution of emissions is quite similar, butGHG-Line reduces emission more in the interim years (2030–2045)primarily from transportation, due to the more stringent cap(Fig. 4).

3.2. Transportation sector

Total fuel demand from the transport sector increases only slightlyin the Reference scenario from 2010 to 2050, due to efficiency im-provements that offset increases in travel demand (Fig. 5). The BAUscenario shows a significant increase in the use of biofuels (ethanoland biomass derived diesel, jet and marine residual fuel oil). Biofuelsmake up approximately 25% of BAU fuel usage in 2050 due to therising oil prices (up to $180/bbl) and falling costs of some biofuels. Inthe GHG-Step scenario, a similar quantity of biofuels is produced, butwith a different mix of fuels produced from different lower carbonfeedstocks (see Section 3.5 for further discussion). Given lower overallfuel demand, biofuels make up 37% of fuel demand in 2050. Petro-leum-based fuels usage declines in the GHG scenario in 2050 (41% vs.66% in BAU). This fuel usage includes fuel consumed by interstate/in-ternational aviation and marine travel. If fuels for these out-of-statetrips are excluded, the use of petroleum in the GHG scenario declinessignificantly to �10% of instate fuel demand in 2050.

The average carbon intensity of instate transportation fuelsdeclines by 16% from 2010 to 2050 (to 71 gCO2e/MJ3) in theReference scenario and 53% (to 40 gCO2e/MJ) in the two GHG

* denotes fuel demand for cross-boundary marine and aviation (activities not included in emissions cap)

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

4,500

2010 2015 2020 2025 2030 2035 2040 2045 2050

Fuel

Dem

and

(PJ)

Gasoline

Diesel NG

Jet Ethanol

Biodiesel Marine Petroleum*

Jet* Marine Biofuels*

Biojet*

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

4,500

2010 2015 2020 2025 2030 2035 2040 2045 2050

Gasoline

Diesel

NG

Jet

Ethanol Biodiesel

BioJet Elec H2

Marine Petroleum*

Jet*

Marine Biofuels*

Biojet*

Fig. 5. Overall transportation fuel demand for BAU (left) and GHG-Step (right) scenarios.

Fig. 6. Comparison of carbon per mile and total miles traveled for different vehicletypes and total emissions from LDVs in 2010, 2030 and 2050 for the GHG-Stepscenario. ICE: Internal combustion engine vehicle; HEV: hybrid electric vehicle;PHEV: plug-in hybrid electric vehicle; BEV: battery-electric vehicle; and FCV: fuelcell vehicle.

C. Yang et al. / Energy Policy 77 (2015) 118–130124

scenarios. Accounting for all transportation fuels, including thoseused for out-of-state travel, the 2050 GHG scenario carbon in-tensity is around 36% lower than 2010 (54 gCO2e/MJ). Sinceemissions for these aircraft and ships are not counted in the cap,higher carbon fuels are shifted to these modes while lower carbonfuels are brought to activities inside the cap.

In the light-duty vehicle (LDV) sector, the Reference scenariovehicle mix looks very similar to today's mix: primarily internalcombustion engine (ICE) vehicles (95%) with a small fraction ofalternative fueled vehicles, 5% fuel cell vehicles (FCVs) used tosatisfy the ZEV mandate. In the GHG scenarios, the mix of vehiclesstarts to shift from ICE vehicles to greater use of electric drivevehicles (PHEVs, BEVs and FCVs) in 2030 and completely to thesevehicles by 2050. Fig. 6 shows the emissions intensity (gCO2e/mileof travel) for different vehicle types and the number of milesdriven on each vehicle type for LDVs in 2010, 2030 and 2050. Allthree vehicle types have significant reductions in carbon per milebut BEVs have by far the lowest, due to the decarbonization of theelectric sector and the reliance of FCVs on hydrogen productionfrom natural gas. On-road fuel economy in the LDV sector for theGHG scenarios is over 110 miles per gallon gasoline equivalent(MPGGE) in 2050 vs. 40 MPGGE in the BAU scenario. GHG emis-sions from the light-duty sector decline 89% from 2010 levels.

Other transportation sectors, such as medium-duty, heavy-dutyand aviation exhibit efficiency improvement in the BAU scenariosfrom 20 to 33% between 2010 and 2050. As would be expected, theGHG scenarios exhibit even greater efficiency improvementscompared to the BAU scenarios. Truck fuel economy in the GHGscenario is 40–60% higher than the BAU scenario in 2050 whileaviation efficiency improves about 10%.

Further decarbonization of transportation is not possible withavailable technologies due to assumptions about the applicabilityof electric drive vehicles in some sectors, and limits on thequantity of biofuels and on further reductions to the carbon in-tensity of electricity and hydrogen.

3.3. Residential and commercial end-use sectors

The efficiency improvements associated with technology adoptionover time for the residential and commercial sector in the Referenceand GHG scenarios are shown in Fig. 7. Overall, total energy usage inthese sectors grows by 30% from 2010 to 2050 in the BAU scenariowhile only increasing 3% in the GHG scenario. In 2010, the mix of fuelsused over these sectors is 41% electricity, 49% natural gas and 10%

other. In 2050, the BAU mix is quite similar (43%, 45% and 13%), whilethere is a major shift to electrification in the GHG-Step and GHG-Linescenarios (64% Elec., 25% NG, 10% Other). Overall, there is a 55% re-duction in natural gas usage in the GHG scenarios in 2050 relative toBAU, resulting in significant GHG emissions reductions (�78% below2010 GHG levels). Assumptions about miscellaneous natural gas usageprevent further emissions reductions in these sectors.

3.4. Electric sector

Decarbonizing the electric sector is critical, because in additionto reducing emissions from the generation of electricity, it alsoprovides opportunities to electrify future end-uses that currentlyrely on petroleum and natural gas.

C. Yang et al. / Energy Policy 77 (2015) 118–130 125

Fig. 8 shows the annual electricity generation mix from 2010 to2050 in the Reference and GHG scenarios. Electricity demand in-creased dramatically in the GHG scenarios due to the shift togreater electrification of end-uses in the residential, commercialand industrial sectors (2010:288 TWh; 2050 BAU:407 TWh; 2050GHG:600 TWh). Much of the electricity growth after 2025 comesfrom electrification in the industrial sector. In the Reference case(BAU), the major changes include the phase out of coal and nuclearand growth in wind power, geothermal and biomass to 80, 28 and26 TWh/yr respectively. Overall, renewables make up 33% of totalgeneration, consistent with the state's RPS. Natural gas continuesto be the major contributor to state electricity supply, making upover half of generation.

The large increase in demand for the GHG scenarios (50%greater than BAU) requires significantly more generation fromrenewable sources in 2050. In order to decarbonize the electricitysupply, renewables must shoulder a much larger fraction of totalgeneration (since nuclear and CCS technologies are not available inthese scenarios). Geothermal and tidal generation expand to28 TWh and 22 TWh respectively, while solar and wind power

Fig. 7. Efficiency improvements in residential and commercial sectors for threeprimary scenarios. COM: commercial sector. RES: residential sector.

Fig. 8. Electricity generation by resource type

make up the bulk of the generation in 2050: utility scale solarthermal, solar PV and wind contributing 107 TWh, 110 TWh and221 TWh, respectively. The installed capacity for solar thermal,solar PV and wind generation are 45 GW, 58 GW and 74 GW, re-spectively. Notable is the lack of biomass-based electricity gen-eration in the GHG scenarios since these energy resources are usedto make biofuels. This optimized result indicates that the limitedbiomass resource is best used to offset petroleum emissions intransportation rather than natural gas generation in the electricsector. Renewable contributions to electricity generation are 81%when excluding hydropower (i.e., only RPS eligible generation)and 88% when including hydropower. Natural gas makes up only8% of generation (mostly combined cycle with some contributionfrom combustion turbines) due to the need for balancing supplyand demand. Including demand response and energy storagetechnologies (not currently implemented in the model) could re-duce the need for this remaining natural gas generation and fur-ther reduce emissions.

With the significant increase in renewable generation andphase out of coal, the carbon intensity of electricity drops sig-nificantly in the reference scenario from 350 in 2010 to184 gCO2e/kWh in 2050. However, in the GHG scenario, electricitycarbon intensity drops by over 90% to reach 30 gCO2e/kWh. Theonly remaining emissions from electricity are due to the continueduse of natural gas generation necessary to balance supply anddemand in each timeslice.

3.5. Fuels supply

Biofuels are critical for decarbonizing the transportation sector.Biofuel consumption in 2050 in the BAU scenario is 1080 PJ or8.2 billion gallons of gasoline equivalent (billion GGE). However,not all of this fuel is low carbon (15% is corn-based ethanol) andoverall, ethanol constitutes about 40% of total biofuel production.

Fig. 9 shows the mix of biofuels that are produced in the GHGscenarios. Interestingly, the two GHG scenarios have slightly lowerbiofuel production than the BAU scenario (in 2050, the GHG sce-narios have approximately �7.3 billion GGE compared with8.2 billion GGE for BAU). Ethanol makes up a tiny fraction of bio-fuels production in the GHG scenarios, with most production oc-curring through Fischer–Tropsch synthetic production of gasoline,

in BAU and GHG-Step/GHG-Line scenarios.

Bio

fuel

Dem

and

(GG

E)

Fig. 9. Biofuel production by category for the Reference and GHG-Step scenarios.

C. Yang et al. / Energy Policy 77 (2015) 118–130126

diesel and jet fuels. This scenario utilizes the full amount of low-carbon biomass that is assumed to be available to California, withapproximately 40% of biomass coming from in-state resources, andthe remainder coming from the rest of the Western US. In 2050,the in-state biomass is primarily municipal solid waste (MSW),yellow grease and tallow, while out-of-state biomass comes pri-marily from energy crops, agricultural residues and MSW.

Given all of these changes in fuel usage, it is expected that theprimary energy resource mix for the BAU and GHG scenarios willalso be quite different. Fig. 10 shows how the efficiency improve-ments in the GHG scenarios lead to a significant reduction inoverall primary energy usage, while fossil resources (oil, naturalgas and coal) usage drops to 49% of total primary energy use from93% in 2050. Most of the oil in the GHG scenarios (�83%) is usedfor out-of-state transport. Renewables make up the remaining 51%of primary energy, with biomass contributing 23%. Also importantis that the use of oil and imported liquid fuels is lower in the BAU-LoVMT scenario due to the lower travel demands relative to theBAU scenario.

Prim

ary

Ener

gy R

esou

rces

(PJ)

Oil

NG

Fig. 10. Primary energy resource for 2010, and Reference and

4. Discussion

4.1. Energy system costs

As we note the technologies and resources that are used toreduce GHG emissions, a discussion of the associated costs of thesetransitions is also important. Table 2 shows the sum of un-discounted and discounted (at 4% discount rate) annual energysystem costs (in 2010$) and the cumulative in-state energy systememissions from 2010 to 2050. The lower VMT assumptions in theBAU-LoVMT (and GHG scenarios) lead to a reduction in the vehiclestock, lower fuel use and lower investments in fuel production anddelivery needed to provide VMT and thus lower system costs, re-lative to the BAU scenario. The cost of lower VMT in the BAU-LoVMT and GHG scenarios is not quantified, in part, because it isquite challenging to quantify costs when VMT reduction can takemany different forms such as pricing policies, smart growth andland-use changes, and/or increased transit investments (Kay et al.,2014). Also important is to what extent these VMT reduction ac-tivities will occur even in the absence of a GHG constraint (i.e. theBAU scenario).

The total energy system costs from 2010 to 2050 are on theorder of 3.2 to 3.6 trillion dollars in net present value (NPV) with adiscount rate of 4%, which includes the cost of transportation ve-hicles, residential and commercial appliances, power plants, fuelproduction facilities, fuel transportation and the cost of purchasingor extracting primary energy resources. It does not include anyequipment or end-use conversion devices in the industrial oragricultural sectors, but does include the costs of supplying energy(natural gas, electricity, etc.) to those sectors. Cumulative emis-sions for these four scenarios range from 13.1 GTCO2e to17.4 GTCO2e.

Table 3 shows the differences in costs and emissions betweenthe GHG and BAU scenarios and the average undiscounted anddiscounted cost per tonne of CO2e reduced for the GHG scenarios.The GHG emissions benefits will depend on which Referencescenario is being compared to (comparing to BAU-LoVMT will re-sult in lower GHG reductions than comparing to BAU). At the sametime, the difference in total energy system costs (i.e. total miti-gation costs) will be lower when comparing to the BAU scenariothan comparing with the BAU-LoVMT scenario. These two trendslead to higher mitigation costs ($/tonne CO2e) when comparing to

Biomass

Wind

Solar

GHG scenarios in 2050 (includes out-of-state transport).

Table 2Summary of undiscounted and discounted (4% discount rate) energy system costs(in 2010$) and cumulative emissions for primary CA-TIMES scenarios for 2010 to2050.

Undiscounted TotalCost (2010 to 2050),$B

Discounted TotalCost (2010 to2050), $B

Cumulative Overalla

Emissions (2010 to2050), MMTCO2e

BAU 8663 3472 17,391BAU-LoVMT 7945 3217 16,539GHG-Step 8947 3482 14,278GHG-Line 9162 3550 13,161

a Overall Emissions includes out-of-state transportation emissions from aviationand marine modes.

C. Yang et al. / Energy Policy 77 (2015) 118–130 127

BAU-LoVMT vs. with BAU. We treat LoVMT assumptions as exo-genous and do not assign costs to travel demand reduction giventhat the cost of VMT reduction is highly uncertain. We believe thatcomparing GHG scenarios with BAU underestimates the true cost ofreducing VMT while still accounting for the monetary benefits(reduced vehicle investments and fuel savings) and that compar-ing GHG scenarios with BAU-LoVMT overestimates the cost (orunderestimates the benefits). On the other hand, we do not at-tempt to quantify the indirect benefits of reducing VMT, such asimproved health and quality of life. The “true cost” may besomewhere in between, but it is highly uncertain so at this point,we present both results, recognizing more research is needed forus to better present the true cost estimates of these scenarios.

The differences in discounted energy system cost are between$9 billion to $333 billion over the 40-year modeling period.Compared to projections of discounted state GSP of $68 trillionover the modeling period, these incremental energy system costsare all below 0.5% of state GSP (0.01% to 0.49%). Assuming a lowerGSP growth rate does not change the results significantly. At 2%annual growth in GSP, discounted mitigation costs range from0.02% to 0.62% of cumulative GSP from 2010 to 2050.4

The differences in cumulative emissions between the GHG sce-narios and the BAU scenarios are between 2.2 and 4.2 GTCO2e.Comparing the GHG-Step scenario with the Reference BAU scenario,we see that the discounted average cost of reducing a tonne of CO2eis $3, while the average cost of reducing a tonne of CO2e from theBAU-LoVMT scenario is $117. For the GHG-Line scenario, the dis-counted average cost of reducing a tonne of CO2e relative to theReference scenario is $18 discounted while the cost rises to$99/tonne CO2e relative to the BAU-LoVMT scenario. The GHG-Linescenario exhibits somewhat higher mitigation costs relative to BAUbut lower costs relative to BAU-LoVMT. Marginal abatement costs inthese two GHG scenarios in 2050 are approximately $7000/tonneCO2e, indicating the model resorts to very high cost options to re-duce the last few percent of emissions. The slope of the marginalabatement cost curve is quite vertical near this emissions reductionlimit as the marginal cost is below $2000/tonne CO2e at 72% GHGreduction (see Supporting information for additional discussion).

Given the significant increase in electricity generation, theelectric sector is responsible for the largest increase in discountedsystem costs to 2050. Comparing GHG-Step to BAU, the electricsector is responsible for much more than the total increase insystem cost (since transportation is responsible for a significantcost reduction due to lower travel demand). When comparingGHG-Step to BAU-LoVMT, the electric, transportation and re-sidential sectors are responsible for 44%, 25% and 24% of the costincrease respectively.

4 At a lower 2% annual GSP growth rate, GSP from 2010 to 2050 is $118 trillionundiscounted and $54 trillion discounted.

4.2. Sensitivity scenarios

A number of sensitivity scenarios (see Table 1) were run inorder to understand the impact of technology, resource availabilityand energy prices on the CA-TIMES modeling results. A number ofsensitivity scenarios are able to meet the 80% GHG emissionsreduction target in 2050 (Fig. 11). These scenarios all have greatersupply of low-carbon resources or technologies or mitigationoptions (including nuclear power, CCS, faster renewable growth,greater biomass supply and elastic demand reduction) than theprimary GHG-Step and GHG-Line scenarios. Even those scenariosthat do reach the 80% reduction target for Included emissions in2050, only achieve 68–70% reduction in Overall emissions relativeto 1990 levels, due to the fact that emissions from out-of-statetransport are not capped. Cumulative emissions are lowest inthe GHG-Line scenario due to the earlier schedule of reductionsresulting from the interim emissions targets.

Among all of the sensitivity factors, the introduction of CCStechnology has the greatest impact on the modeling decisions andsystem costs, because it is the largest source of additional low-carbon resources. CCS is used for the production of electricity(from natural gas) and hydrogen (from natural gas and biomass)further reducing their carbon intensities (to below 14 gCO2e/kWhfor electricity and to �51 gCO2e/MJ for hydrogen). The productionof biofuels is coupled with CCS, which provides a large negativeemissions source. Greater than 50% of the carbon in the biomasscan be sequestered in the production of biofuels such that thecarbon sequestered during production of each gallon of biofueloffsets the use of more than one gallon of petroleum fuel. As aresult, the CCS scenarios (GHG-S-NucCCS and GHG-S-CCS) do notrequire the same level of vehicle electrification in the light-dutysector (approximately 90% of vehicles can continue to be primarilycombustion powered, compared to 0% in the primary GHG sce-nario). Petroleum fuel use is approximately doubled in thesetwo scenarios compared to the primary GHG scenarios eventhough the average CI of transportation fuels is 22% lower (42 vs.54 gCO2e/MJ).

In the electric sector, the major changes from the primary GHGscenarios come from the sensitivity scenarios that introduce ad-ditional resources for low-carbon electricity generation (i.e. nu-clear, CCS, greater renewables) and reductions in demand. In thesecases, greater low-carbon electricity supply allows for greaterelectricity demand with even lower carbon intensity. In the GHG-S-Nuclear, GHG-S-HiRen, GHG-NucCCS scenarios, electricity de-mand increases an additional 3-15% while reducing carbon in-tensity (CI) to between 2 and 14 gCO2e/kWh. In the elasticityscenarios (GHG-S-Elas1 and GHG-S-Elas2), the electricity demandis reduced due to demand reduction (10–14%) and the lower de-mand means limited renewables can make up a greater proportionof total generation reducing electricity CI to between 4 and7 gCO2e/kWh.

Total system costs GHG mitigation costs can vary significantlybetween the primary GHG scenarios and sensitivity scenarios(Table 4). Low cost, low-carbon resource options such as nuclearpower, and carbon capture and sequestration lead to the lowestcost of meeting the emissions target requiring an additional $20–40 billion in discounted system cost vs. BAU-LoVMT and negativecosts relative to BAU. The presence of absence of these and otherlow carbon mitigation options can make a large difference in theability and the cost of achieving the GHG targets. Among the twoelastic demand scenarios, GHG-S-Elas1 allows the reduction ininvestments of end use technologies while GHG-S-Elas2 does not,and the difference in costs of mitigation is significant. These twoscenarios are treated as bounding cases as actual consumerbehavior is likely to fall between these two cases: lower VMTdemand should lead to fewer vehicle purchases (as in GHG-S-

2050

GH

G R

educ

tions

bel

ow 1

990

Cum

ulat

ive

Emis

sion

s (G

TCO

2e)

Cumulative Emissions

Emissions Reduction

Fig. 11. California cumulative GHG emissions and 2050 emissions reduction from 1990 levels for Included and Overall (IncludedþOut-of-State Transport) emissions.

Table 3Summary of undiscounted energy system costs (in 2010$) and cumulative emissions for primary CA-TIMES scenarios for 2010 to 2050.

GHG-Step vs. BAU GHG-Step vs. BAU-LoVMT GHG-Line vs. BAU GHG-Line vs. BAU-LoVMT

Difference in Undiscounted Total Costs ($M) 314,105 1,001,674 528,782 1,216,351Difference in Discounted Total Costs ($M) 9270 264,807 77,456 332,994Undiscounted Cost Difference ($/resident/yr) 139 503 250 614Discounted Cost Difference ($/resident/yr) �5 140 32 177Undiscounted Cost Difference (% GSP)a 0.20% 0.62% 0.33% 0.76%Discounted Cost Difference (% GSP) 0.01% 0.39% 0.11% 0.49%Difference in Cumulative Emissions (MMTCO2e) 3113 2261 4229 3378Undiscounted Cost of emissions reduction (2010$/tCO2 reduced) 100.9 443.0 125.0 360.1Discounted Cost of emissions reduction (2010$/tCO2 reduced) 3.0 117.1 18.3 98.6

a This calculation uses a 3.35% annual Gross State Product (GSP) growth rate from Moody's baseline scenario, starting from $1.88 trillion in 2010 for total undiscountedGSP from 2010 to 2050 of $161 trillion and discounted GSP of $68 trillion.

C. Yang et al. / Energy Policy 77 (2015) 118–130128

Elas1) but consumers may still purchase the same number of airconditioning unit even if they reduce their usage (GHG-S-Elas2).

5. Conclusions and policy implications

5.1. Key conclusions

The CA-TIMES model was used to construct a series of referenceand deep GHG emissions reduction scenarios. The integrated model-ing approach is an important strength of CA-TIMES, in that the majorenergy transformations needed to meet GHG emissions targets arechosen while simultaneously considering limited resources, technol-ogy costs, and policy constraints across all energy sectors.

The modeling shows that deep reductions in GHG emissionscan be achieved at low to moderate costs to society, includingmany scenarios that achieve the 80% reduction, but this requiressignificant transformation of the state's energy system. The mostnotable changes on the supply side include major investments inrenewable electricity generation, biofuel production, and hydro-gen production. Renewables produce 60–85% of total generationacross the various GHG scenarios, requiring large investments andfast ramp up and carbon intensity declines by over 90% by 2050.This is even more challenging because of the large increase inelectricity demand that comes from electrification of end-uses.

Transportation reduces emissions through a combination ofhigher efficiency, advanced drivetrain technologies (electric and

fuel cell) and low-carbon transportation fuels. Zero-emission ve-hicles make up between 50 and 90% of LDV fleets in most GHGscenarios and fuel economy climbs as high as 113 MPGGE in manyscenarios. Carbon intensity of transportation fuels falls by 36%with increasing use of low-carbon biofuels (up to 7 billion gallonsof gasoline equivalent (GGE)), hydrogen and electricity. Petroleumusage and resulting emissions are high for cross-boundary marineand aviation travel, since these activities are not included in thestate's carbon cap. Counting these emissions reduces the state'sGHG reductions to 63–70% (vs. 75–80% when only counting In-cluded emissions). Biomass is used exclusively for transportationfuels rather than electricity generation. CCS technology has aprofound effect on the technology choices in the transportationsector. The availability of low and negative-carbon transportationfuels (biofuels and hydrogen) from CCS lead to much greater use ofinternal combustion engine vehicles (and lower vehicle effi-ciencies) than scenarios without CCS. CCS is a critical enablingtechnology for low-cost GHG mitigation, because it reduces bothcost and carbon intensity of transportation fuels and electricityand displaces higher cost options.

CA-TIMES also includes detailed demand drivers and technol-ogy options in the residential and commercial sectors. All sensi-tivity scenarios all show significant improvements in efficiency inthese sectors, coupled with an increasing reliance on low-carbonelectricity (i.e. electrification of end-uses). Technology efficienciesaveraged across all service demands rise by approximately 200%from 2010 to 2050 in the commercial sector and 275% in the

Table 4Cumulative emissions, total system costs and cost and emissions differences for various sensitivity scenarios.

Cumulative Overall InstateEmissions, MMTCO2e

Dis-countedCost, $B

Cumulative Included InstateEmissions Difference

Discounted Cost Difference(% of GSP)

Avg Cost of Emissions reduction($/tCO2e), discounted

vs. BAU vs. BAU-LoVMT vs. BAU vs. BAU-LoVMT

vs. BAU vs. BAU-LoVMT

BAU 17,391 3472 – – – – – –

BAU-LoVMT 16,539 3217 – – – – – –

GHG-Step 14,278 3482 3113 2261 0.01 0.39 3 117GHG-Line 13,161 3550 4229 3378 0.11 0.49 18 99GHG-S-NucCCS 14,229 3237 3162 2310 �0.34 0.03 �75 9GHG-S-CCS 14,301 3254 3089 2238 �0.32 0.05 �71 17GHG-S-Nuclear 13,821 3337 3569 2718 �0.20 0.18 �38 44GHG-S-HiRen 14,148 3396 3242 2391 �0.11 0.26 �24 75GHG-S-HiBio 13,878 3431 3513 2661 �0.06 0.31 �12 81GHG-S-HiOilGas 14,107 3493 3283 2432 0.03 0.40 6 114GHG-S-LoOilGas 14,454 3476 2937 2085 0.01 0.38 1 124GHG-S-Elas1 14,151 3414 3240 2388 �0.09 0.29 �95 (77)a �21 (104)a

GHG-S-Elas2 14,583 3425 2808 1956 �0.07 0.30 �51 (34)a 58 (48)a

a Mitigation cost values in parentheses represent non-monetary loss of consumer utility from reduced energy services.

C. Yang et al. / Energy Policy 77 (2015) 118–130 129

residential sector. Demand reduction is another important miti-gation option that is explored in scenarios, and the results suggestthat there are large potential savings in emissions.

Finally, the CA-TIMES model was used to explore the costs ofGHG mitigation and these costs vary significantly across the pri-mary GHG and sensitivity scenarios. However, these costs are re-latively small: cumulative discounted costs relative to the BAU-LoVMT scenario range from $20 to $333 billion or 0.03% to 0.5% ofdiscounted cumulative GSP from 2010 to 2050. Also important isthe fact that interim targets (e.g. a linear reduction in emissionsfrom 2020 to 2050) do not raise average mitigation costs sig-nificantly relatively to simply having a 2050 target, but can lead tosubstantial reductions in cumulative emissions. These calculationsinclude benefits associated reductions in fuel usage from higherefficiency, but do not include direct and ancillary benefits asso-ciated with mitigation of climate damages, air pollution and healthimpacts, nor spillover effects from low-carbon investments ontechnology development, income, and jobs.

5.2. Policy implications

California's 80% GHG reduction target is an aspirational goalrather than a binding regulation. The results of this analysis canhelp inform two aspects of the state's revision of its scoping planand consideration of setting intermediate binding targets andadditional sectoral policies: (1) setting an appropriate target in2035 while balancing emissions reductions, cumulative emissionsand expected mitigation costs, and (2) developing new or updat-ing existing policy frameworks in specific sectors. Given the un-certainty in the future of low-carbon technologies, resources,consumer demands and the variations in energy system structureacross the sensitivity scenarios, policymakers should be wary ofputting too much emphasis on the outcomes of any one scenario.The robust results described in the previous section have someimportant implications for this policy development under un-certainty. Given their importance in the modeling results, there isa need to understand the supply and carbon benefits of biomass-based fuels. CCS appears to be a critical enabling technology thathas substantial mitigation potential at a fairly low cost. The stateshould explore additional policies for incentivizing efficiency im-provements across all sectors, especially transportation and elec-trification in buildings. The state should investigate policies topromote the decarbonization and expansion of the electric sector,as well as manage the challenge of integrating very high percen-tages of intermittent renewables and load balancing. Finally, an

interim target (such as one consistent with the GHG-Line scenario)is helpful to drive near-term investments in low-carbon technol-ogies, address reductions in cumulative emissions and should notraise costs of mitigation significantly.

5.3. Caveats

Like all models, CA-TIMES represent a simplification of reality,and the results of this model must be understood in the context ofthe following limitations and caveats:

�

There is a single global decision-maker rather than millions ofindividuals and firms.

�

Decisions are made with perfect foresight and no uncertainty. � Uncertainty analysis is not performed in an explicit manner. � No explicit modeling of the industrial and agricultural sectors � Decisions are driven primarily by investment and operating costs. � Non-spatial representation of energy demands and supply. � Exogenous assumptions about technology availability, perfor-

mance and costs that do not incorporate learning-by-doing.

� No macroeconomic feedback. � Only models California, with exogenous assumptions about the

rest of the world.

5.4. Future work

This work is critical for policymakers to assess the varioustechnology options and policy mechanisms for GHG emissionsreductions. These scenarios and system analyses need to continueto be refined and improved to provide a more useful tool for policyanalysis and energy planning. Ongoing work includes:

�

Incorporating consumer heterogeneity and choice. � Assessing criteria pollutant emissions and air quality con-

centration changes.

� Water use impacts of future electricity and fuels.

Other work that is being considered and planned:

�

Incorporating demand and technology details in the industrialand agricultural sectors.

�

Modeling and analysis of demand reduction as a mitigationstrategy.

C. Yang et al. / Energy Policy 77 (2015) 118–130130

�

Improved spatial representation of infrastructure. � Inclusion of social/externality costs into model optimization. � Explicit incorporation of risk and uncertainty.

Acknowledgments

The authors would like to thank the Sustainable TransportationEnergy Pathways (STEPS) Program at UC Davis and the CaliforniaAir Resources Board (ARB) for providing the primary support forthis research. Additional support from the California EnergyCommissions (CEC) is also appreciated. The authors would alsolike to acknowledge the following individuals who provided data,technical expertise and useful discussions and feedback at variousstages of the model development: Prof. David Bunch, Prof. JoanOgden, Christina Zapata, Anthony Eggert, Amit Kanudia, AnttiLehtila, and Gary Goldstein. The authors would also like to thankthe anonymous reviewers whose comments and suggestions im-proved this paper considerably.

Appendix A. Supplementary material

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.enpol.2014.12.006.

References

AEO, 2013. Annual Energy Outlook Analysis and Projections. US Department ofEnergy, Energy Information Administration. ⟨http://www.eia.gov/forecasts/aeo/data.cfm⟩ (accessed 02.10.13).

Allan, A., 2011. Renewable Transportation Fuel for California's Electric-Drive Ve-hicles, Transportation Technology and Policy. University of California, Davis.

Boeing, 2007. Commercial Airplanes: Jet Prices. ⟨http://www.boeing.com/boeing/commercial/prices/index.page⟩ (accessed 09.10.07).

BTS, 2010. Bureau of Transportation Statistics: Aviation. ⟨http://www.transtats.bts.gov/databases.asp?mode_id¼1&mode_desc¼aviation&subject_id2¼0⟩ (ac-cessed 6/14/2011).

California Department of Finance, 2013. Population Projections, 2010–2060. ⟨http://www.dof.ca.gov/research/demographic/reports/view.php⟩ (06.02.13).

CARB, 2004. Statewide Commercial Harbor Craft Survey: Final Report, EmissionsAssessment Branch. California Air Resources Board, Sacramento, CA.

CARB, 2005. Oceangoing Ship Survey. California Air Resources Board, Sacramento,CA.

CARB, 2007. OFFROAD2007, Off-Road Emissions Inventory Program. California AirResources Board, Sacramento, CA ⟨http://www.arb.ca.gov/msei/categories.htm⟩

(accessed 04.04.09).CARB, 2008. Climate Change Scoping Plan Document: a framework for change.

California Air Resources Board for the State of California. ⟨http://www.arb.ca.gov/cc/scopingplan/document/adopted_scoping_plan.pdf⟩.

CARB, 2009. Attachment B: 2050 Greenhouse Gas Emissions Analysis: Staff Mod-eling in Support of the Zero Emission Vehicle Regulation. California Air Re-sourcs Board. ⟨http://www.arb.ca.gov/msprog/zevprog/2009zevreview/2009zevreview.htm⟩. (accessed 01.10.10).

CARB, 2011. EMFAC Software. ⟨http://www.dot.ca.gov/hq/env/air/pages/emfac.htm⟩

(accessed 03.13.12).CARB, 2012. Vision for Clean Air: A Framework for Air Quality and Climate Planning.

California Air Resources Board, Sacramento.CARB, 2014. Proposed First Update to the Climate Change Scoping Plan: Building on

the FrameworkCalifornia Air Resources Board, Sacramento, California, USA.CCSRP, 2010. Findings and Recommendations by the California Carbon Capture and

Storage Review Panel California Carbon Capture and Storage Review Panel-California Energy Commission, California Public Utilityes Commission and AirResources Board, December 2010.

CCST, 2011. California's Energy Future – The View to 2050. California Council onScience and Technology, Sacramento, California, USA.

CEC, 2011. 2011 Integrated Energy Policy Report. California Energy Commission,Sacramento, California, USA.

EPA, 2014. Assumptions from EPA US 9-Region MARKAL model Database, Wa-shington DC. ⟨http://nepis.epa.gov/exe/zypdf.cgi/p100i4rx.pdf?dockey¼p100i4rx.pdf⟩.

FAA, 2007. General Aviation and Air Taxi Activity and Avionics (GAATAA) SurveysCY2005. Federal Aviation Administration, Washington, DC, USA.

GEA, 2012. Global Energy Assessment-Toward a Sustainable Future, CambridgeUniversity Press, Cambridge, UK and New York, NY, USA; International Institutefor Applied Systems Analysis, Laxenburg, Austria.

Gillingham, K., Harding, M., Rapson, D., 2012. Split incentives in residential energy

consumption. Energy J. – Int. Assoc. Energy Econ 33 (2), 37–62.Greenblatt, J., Long, J., Hannegan, B., 2012. California's Energy Future: Electricity

from Renewable Energy and Fossil Fuels with Carbon Capture and Sequestra-tion. California Council on Science and Technology, April 2012.

Greenblatt, J.B., 2013. Estimating Policy-Driven Greenhouse Gas Emissions Trajec-tories in California: The California Greenhouse Gas Inventory Spreadsheet(GHGIS) Model Lawrence Berkeley National Laboratory, Berkeley.

Greene, D.L., 2011. Uncertainty, loss aversion, and markets for energy efficiency.Energy Econ. 33, 608–616.

H2A, 2005. DOE Hydrogen Analysis (H2A). ⟨http://www.hydrogen.energy.gov/h2a_analysis.html⟩ (accessed 02.10.06).

ICF, 2010. Modeling of Greenhouse Gas Reduction Measures to Support the Im-plementation of the California Global Warming Solutions Act (AB32): Energy2020 Model Inputs and Assumptions. ICF Consulting Canada and SystematicSolutions, Inc, Toronto, ON, Canada.

IEA, 2008. Energy Technology Perspectives 2008: Scenarios and Strategies to 2050.International Energy Agency. Organization for Economic Co-Operation andDevelopment (OECD), Paris, France.

IEA, 2010. Energy Technology Perspectives – Scenarios & Strategies to 2050. In-ternational Energy Agency, Paris, France.

IEA, 2012. Energy Technology Perspectives 2012-Pathways to a Clean Energy Sys-tem. International Energy Agency, Paris, France.

ITRON, 2006. California Commercial End-Use Survey (CEUS). ⟨http://www.energy.ca.gov/2006publications/cec-400-2006-005/cec-400-2006-005.pdf⟩.

Kay, A.I., Noland, R.B., Rodier, C.J., 2014. Achieving reductions in greenhouse gasesin the US road transportation sector. Energy Policy 69, 536–545.

Kreutz, T.G., Larson, E.D., Liu, G., Williams, R.H., 2008. Fischer–Tropsch Fuels fromCoal and Biomass. In: Proceedings of the 25th Annual International PittsburghCoal Conference Pittsburg, PA.

Leighty, W., 2012. Modeling transitions in the California light-duty vehicles sectorto achieve deep reductions in transportation greenhouse gas emissions. EnergyPolicy 44, 52–67.

Loulou, R., Labriet, M., 2007. ETSAP-TIAM: the TIMES integrated assessment modelPart I: Model structure. Computational. Manag. Sci. 5, 7–40.

Loulou, R., Lavigne, D., 1996. MARKAL model with elastic demands: applicationto greenhouse gas emission control. In: Carraro, C., Haurie, A. (Eds.),Operations Research and Environmental Management. Springer, Netherlands,pp. 201–220.

McCarthy, R., Yang, C., Ogden, J., 2008. California Energy Demand Scenario Pro-jections to 2050. Institute of Transportation Studies. University of California,Davis (Research Report UCD-ITS-RR-08-11.).

McCollum, D., Yang, C., Yeh, S., Ogden, J., 2012. Deep greenhouse gas reductionscenarios for California: Strategic implications from the CA-TIMES energy-economic systems model. Energy Strategy Rev. 1, 19–32.

Nelson, J., Johnston, J., Mileva, A., Fripp, M., Hoffman, I., Petros-Good, A., Blanco, C.,Kammen, D.M., 2012. High-resolution modeling of the western North Americanpower system demonstrates low-cost and low-carbon futures. Energy Policy 43,436–447.

NEMS, 2013. Input Assumptions to the National Energy Modeling System (NEMS).U.S. Department of Energy. Washington, DC, USA.

NRC, 2013. Transitions to Alternative Vehicles and Fuels. The National AcademisePress, Washington, DC.

NTD, 2010. National Transit Database. ⟨http://www.ntdprogram.gov/ntdprogram/data.htm. 2/9/2012⟩.

O’Neill, B., Kriegler, E., Riahi, K., Ebi, K., Hallegatte, S., Carter, T., Mathur, R., Vuuren,D., 2014. A new scenario framework for climate change research: the concept ofshared socioeconomic pathways. Clim. Change 122, 387–400.

Parker, N., Tittmann, P., Hart, Q., Nelson, R., Skog, K., Schmidt, A., Gray, E., Jenkins, B.,2010. Development of a biorefinery optimized biofuel supply curve for theWestern United States. Biomass Bioenergy 34, 1597–1607.

RETI, 2009. Renewable Energy Transmission Initiative: Phase 1B Final Report.Richter, B., Budnitz, R., Long, J., Peterson, P., Schori, J., 2011. California's Energy

Future-Powering California with Nuclear Energy. California Council on Scienceand Technology, July 2011.

Roland-Holst, D., 2008. Berkeley Energy And Resources (BEAR) Model. TechnicalDocumentation for a Dynamic California CGE Model for Energy and Environ-mental Policy Analysis.

Rufo, M., North, A., 2007. Assessment of Long-Term Electric Energy Efficiency Po-tential in California's Residential Sector, California Energy Commission, PIEREnergy – Related Environmental Research. CEC-500-2007-002.

Wei, M., Nelson, J.H., Greenblatt, J.B., Mileva, A., Johnston, J., Ting, M., Yang, C., Jones, C.,McMahon, J.E., Kammen, D.M., 2013. Deep carbon reductions in California requireelectrification and integration across economic sectors. Environ. Res. Lett. 8, 014038.

Williams, J.H., DeBenedictis, A., Ghanadan, R., Mahone, A., Moore, J., Morrow, W.R.,Price, S., Torn, M.S., 2012. The technology path to deep greenhouse gas emis-sions cuts by 2050: the pivotal role of electricity. Science 335, 53–59.

Yang, C., McCollum, D., McCarthy, R., Leighty, W., 2009. Meeting an 80% reduction ingreenhouse gas emissions from transportation by 2050: a case study in Cali-fornia, USA. Transp. Res. D – Transp. Environ. 14, 147–156.

Yang, C., Ogden, J.M., 2013. Renewable and low carbon hydrogen for California –

modeling the long term evolution of fuel infrastructure using a quasi-spatialTIMES model. Int. J. Hydrog. Energy 38, 4250–4265.

Yang, C., Yeh, S., Ramea, K., Zakerinia, S., McCollum, D., Bunch, D.S., Ogden, J., 2014.Modeling Optimal Transition Pathways to a Low Carbon Economy in California:California TIMES (CA-TIMES) Model. Institute of Transportation Studies, Uni-versity of California, Davis.