Early decarbonisation of the European energy system pays off

Marta Victoriaa,b,∗, Kun Zhua , Tom Brownc , Gorm B. Andresena,b , Martin Greinera,b
a Department of Engineering, Aarhus University, Inge Lehmanns Gade 10, 8000 Aarhus, Denmark
b iCLIMATE Interdisciplinary Centre for Climate Change, Aarhus University
c Institute for Automation and Applied Informatics (IAI), Karlsruhe Institute of Technology (KIT), Forschungszentrum 449, 76344,

Eggenstein-Leopoldshafen, Germany

Abstract
arXiv:2004.11009v1 [physics.soc-ph] 23 Apr 2020

In the context of increasing public climate change awareness and plummeting costs for wind and solar photovoltaics,
discussions on increasing CO2 reduction targets for Europe have started. Here, we model alternative transition paths
with strict carbon budget for the sector-coupled networked European energy system. We show that up-to-date costs for
wind and solar and the inclusion of highly resolved time series for balancing make climate action with renewables more
cost-effective than previously seen. Ambitious CO2 reductions in the short term not only trigger a cheaper transition
but also incentivise more stable CO2 prices and build rates for the required new capacities which could be beneficial
from the point of view of investors, social acceptance, local economies, and jobs creation.
Keywords: myopic optimisation, carbon dioxide reduction, grid integration of renewable power, sector coupling, open
energy modelling

1. Introduction when the transport sector is included.

Achieving a climate-neutral European Union in 2050

[1] requires meeting the milestones in between. Although
carbon emissions will most likely sink by 20% in 2020 rel-
ative to 1990 [2], it is unclear whether the 40% objective
settled for 2030 will be met. The national energy plans for
the coming decade submitted by member states do not add
up the necessary reduction to meet the target [3], while in
the context of a European Green Deal a more ambitious
reduction of 55% is currently under discussion [4]. At the
same time, led by young people [5], society is advocating
for more ambitious climate action.

A remaining global carbon budget of 800 Gigatons (Gt)

of CO2 can be emitted from 2018 onwards to limit the
anthropogenic warming to 1.75◦ C relative to the prein- Figure 1: Historical CO2 emissions from the European power system
dustrial period with a probability of more than 66% [6]. and heating supply in the residential and services sectors [9]. The
various future transition paths shown in the figure have the same
Different sharing principles can be used to split the global cumulative CO2 emissions, which correspond to the remaining 21 Gt
carbon budget into regions and countries [7]. Considering CO2 budget to avoid human-induced warming above 1.75◦ C with a
an equal per-capita distribution translates into a quota of probability of more than 66%, assuming current sectoral distribution
48 GtCO2 for Europe. An approach that took into ac- for Europe, and equity sharing principle among regions. Black stars
indicate committed EU reduction targets, while white stars mark
count historical emissions would lead to more ambitious targets under discussion. See also Figure S1.
targets for Europe than other regions [8]. Assuming that
sectoral distribution of emissions within Europe remains
at present values, the carbon budget for the generation of Myopic optimisation with sector coupling.
electricity and provision of heating in the residential and Electricity generation is expected to spearhead the tran-
services sectors accounts for approximately 21 GtCO2 , [9] sition spurred by the dramatic cost reduction of wind en-
and Procedure S1. The budget increases to 33 GtCO2 ergy [10] and solar photovoltaics (PV) [11, 12]. A vast
body of literature shows that a power system based on
∗ Corresponding author wind, solar, and hydro generation can supply hourly elec-
Email address: mvp@eng.au.dk (Marta Victoria) tricity demand in Europe as long as proper balancing is
Preprint submitted to April 24, 2020
provided [13–16]. This can be done by reinforcing inter- effective than previously seen. Furthermore, we find that a
connections among neighbouring countries [17] to smooth transition path with more ambitious short-term CO2 tar-
renewable fluctuations by regional aggregation or through gets reduces the cumulative system cost and requires a
temporal balancing using local storage [18–20]. Moreover, smoother increase of the CO2 price and more stable build
coupling the power system with other sectors such as heat- rates. Our research includes the coupling with heating
ing or transport could provide additional flexibilities facil- and transport sectors, which is absent in transition path
itating the system operation and simultaneously helping analyses for the European power system [28–30], as well
to abate emissions in those sectors [21–23]. as up-to-date cost assumptions for wind and solar PV to-
gether with hourly resolution, in contrast to the outdated
CO2 emissions from heating in the residential and ser- cost and low temporal resolution in Integrated Assessment
vices sectors show a more modest historical reduction trend Models (IAMs) [11, 32]. We use an open model, which en-
compared to electricity generation (Figure 1). Nordic coun- sures transparency and reproducibility of the results [33].
tries have been particularly successful in reducing carbon
emissions from the heating sector by using sector-coupling
2. Results
strategies, Figures S2 and S3. Denmark, where more than
half of the households are connected to district heating Here, we investigate the consequences of following two
systems [24], has shifted the fuel used in Central Heat and alternative transition paths. The Gentle path represents
Power (CHP) units from coal to biomass and urban waste a cautious approach in which significant emissions reduc-
incineration [25]. Sweden encouraged a large-scale switch tions are attained in the early years. In the Sudden path,
from electric resistance heaters to heat pumps [24] which the low initial reduction targets quickly deplete the car-
are now supported by high CO2 prices [26] and low elec- bon budget, requiring a sharp reduction later. As in Ae-
tricity taxes. sop’s fable “The Tortoise and the Hare”, the tortoise wins
the race by making steady progress, whereas following the
Greenfield optimisation of the future European energy hare and delaying climate action requires a late accelera-
system, that is, building the system from scratch, shows tion that will be more expensive and might be unfeasible.
that sector-coupling decreases the system cost and reduces
the need for extending transmission lines due to the addi- Cumulative costs.
tional local flexibility brought by the heating and transport The two alternative paths arrive at a similar system
sectors [22]. Sector-coupling allows large CO2 reductions configuration in 2050, Figure 2. Towards the end of the pe-
before large capacities of storage become necessary, provid- riod, under heavy CO2 restriction, balancing technologies
ing more time to further develop storage technologies [20]. appear in the system. They include large storage capac-
Greenfield optimisation is useful to investigate the opti- ities comprising electric batteries and hydrogen storage,
mal configuration of the fully-decarbonised system, but it and production of synthetic methane. Cumulative sys-
does not provide insights on how to transition towards it. tem cost for the Gentle path represents 6,994 billion euros
Today’s generation fleet and decisions taken in interme- (Be ), while the Sudden path accounts for 7,341 Be . The
diate steps will shape the final configuration. Transition newly built conventional capacity for electricity generation
paths for the European power system have been analysed is very modest in both cases, Figure 3 and Figure S5. No
using myopic optimisation, without full foresight over the new lignite, coal or nuclear capacity is installed. Thus, at
investment horizon [27–30]. Myopic optimisation results the end of both paths, conventional technologies include
in higher cumulative system cost than optimising the en- only gas-fueled power plants, CHP and boilers. Biomass
tire transition period with perfect foresight because the contributes to balancing renewable power but plays a mi-
former leads to stranded investments [29, 31]. However, nor role. Decarbonising the power system has proven to
the myopic approach is less sensitive to the assumed dis- be cheaper than the heating sector [34]. Consequently, al-
count rate and can capture better short-sighted behaviour though CO2 allowances differ, the electricity sector gets
of political actors and investors [29, 30]. quickly decarbonised in both paths and more notable dif-
ferences appear in new conventional heating capacities,
In this work, we use an hourly-resolved sector-coupled Figure 4. Under the tortoise strategy yearly costs initially
networked model of the European energy system and my- decrease as the power system takes advantage of the low
opic optimisation in 5-years steps from 2020 to 2050 to costs of wind and solar. Removing the final emissions in
investigate the impact of different CO2 reduction paths heating causes total costs to rise again towards 2050.
with the same carbon budget. In every time step, the
expansion of generation, storage and interconnection ca- Stranded assets.
pacities in every country is allowed if it is cost-effective Part of the existing conventional capacities become stranded
under the corresponding global emissions constraint. We assets, in particular, coal, lignite, CCGT (which was heav-
show that up-to-date costs for wind and solar that take ily deployed in the early 2000s, Figure 3) and gas boilers.
into account recent capacity additions and technological As renewable capacities deploy, utilisation factors for con-
learning make climate action with renewables more cost- ventional power plants decline and they do not recover
2
Figure 2: Levelized Cost of Energy (LCOE) for the European electricity and heating system throughout transition paths Gentle and Sudden
shown in Figure 1. Conventional includes costs associated with coal, lignite, and gas power plants producing electricity as well as costs for
fossil-fueled boilers and CHP units. Power-to-heat includes costs associated with heat pumps and heat resistors. Balancing includes costs of
electric batteries, H2 storage, and methanation.

Figure 3: Age distribution of European power plants in operation [35, 36] and required annual installation throughout the Gentle path, see
also Figure S5-10.

their total expenditure via market revenues, Figures S11- retirement of electricity infrastructure has been identified
14. Up to 2035, operational expenditure for gas-fueled as one of the most cost-effective actions to reduce commit-
technologies are lower than market revenues so they are ted emissions and enable a 2◦ C-compatible future evolu-
expected to remain in operation. Unexpectedly, the sum tion of global emissions [37].
of expenditures not recovered via market revenues is sim-
ilar for both paths. In the Sudden path, high CO2 prices Transition smoothness.
justify producing up to 220 TWh/a of synthetic methane A timely transition is challenging yet feasible given his-
in 2040. This enables CCGT and gas boilers to keep op- toric build rates. Decarbonising the electricity and heating
erating allowing them to recover part of their capital ex- sectors using wind and solar PV requires duplicating the
penditure, but the consequence is a higher cumulative sys- highest historical build rates seen in individual countries,
tem cost, as previously discussed. Although closing plants Figure 3 and Figure S4. Consequently, attaining higher
early might be seen as an unnecessary contribution to a build rates to also decarbonise transport and industry sec-
higher cost of energy, it must be remarked that the early tors seems feasible. Wind and solar PV supply most of
the electricity demand in 2050, complemented by hydro
3
and with a minor biomass contribution. Previously, most
IAMs have emphasized the importance of bioenergy or car-
bon capture and storage and failed to identify the key role
of solar PV due to their unrealistically high cost assump-
tions for this technology, see [11, 32] and Procedure S4.2.

Figure 5: Historical evolution of CO2 price in the EU Emissions

Trading System [40] and required CO2 price obtained from the model
throughout transition paths shown in Figure 1. Co-benefits of reduc-
ing CO2 emissions in Europe due to avoided premature mortality,
reduced lost workdays, and increased crop yields are estimated in
the range of 125-425 e /ton CO2 [41].

At every time step, the optimal renewable mix in ev-

ery country depends on the local resources and the already
existing capacities, see Figures S16-18. Nevertheless, the
analysis of near-optimal solutions has recently shown that
Figure 4: Required expansion of heating capacities in both paths. country-specific mixes can vary significantly while keeping
Maximum heating capacities are shown for CHP plants.
the total system cost only slightly higher than the mini-
mum [42].
During the past decade, several European countries
have shown sudden increments in the annual build rate for Modelling an entire year with hourly resolution unveils
solar PV, followed by equivalent decrements one or two the strong links between renewable generation technolo-
years later. Italy, Germany, UK, and Spain show clear gies and balancing strategies. For countries and years in
peaks (Figure S4) due to the combination of a fast cost which large solar PV capacities are deployed, it is also cost-
decrease of the technology and unstable regulatory frame- effective to install large battery capacities to smooth the
works whose details are country-specific. These peaks are strong daily solar generation pattern. Conversely, onshore
lethal for local businesses. The sudden shrinkage of an- and offshore wind capacities require hydrogen storage and
nual build capacity results in companies bankruptcy and reinforced interconnections to balance wind synoptic fluc-
lost jobs. The Gentle path requires a smoother evolution tuations [14, 18, 20]. This can also be appreciated by look-
of build rates which could better accommodate the cul- ing at the dominant dispatch frequencies exposed by the
tural, political, and social aspects of the transition, [38] Fourier power spectra of the Europe-aggregated time se-
and Figure S15. The mild evolution could also facilitate ries in 2050 on the Gentle path, Figure 6.
reaching a stationary situation in which build rates offset
decommissioning. IAMs with similar spatial resolution have also been
used to investigate the sector-coupled decarbonisation of
CO2 prices much higher than those historically attained Europe [1, 11, 43]. However, IAMs typically use a much
in the ETS market are necessary at the end of the tran- lower time resolution, e.g., using a few time slices to repre-
sition, Figure 5. The Gentle path requires a smoother sent a full year [30, 43–46] or considering the residual load
evolution of CO2 price, which will be preferred by in- duration curve [11, 47], and some IAM assume very high
vestors. CO2 price is only an indicator of the price gap integration costs for renewables [48]. The hourly resolu-
between polluting and clean technologies and several poli- tion in our model reveals several effects that are critical
cies can be established to fill that gap. Among others, to the operation of highly renewable systems, such as the
sector-specific CO2 taxes [26], direct support for renew- variable, but correlated solar and wind power generation
ables that reduce investor risk, and consequently the cost smoothed by the grid, the role of long-term storage, and
of capital and LCOE of the technology [39], or regulatory the system operation during cold spells, i.e., a cold week
frameworks that incentivise the required technologies such with low wind and solar generation.
those promoting rooftop PV installations or ensuring the
competitiveness of district heating systems. Results robust under different scenarios.
District heating (DH) has proven to be extremely use-
Hourly and country resolved results. ful to decarbonise the heating sector. It allows cheaper
4
 Finally, Gentle and Sudden paths are re-run including
the coupling of road and rail transport, as described in
Procedure S3.5. For every time step, the electrification
of transport is assumed to be equal to the CO2 emissions
reduction relative to 2020. In this way, emissions in that
sector sink roughly parallel to those of heating and electric-
ity sectors. The extra electricity demand raises cumulative
system cost, but the LCOE remains similar throughout the
transition. The additional flexibility provided by EVs re-
duces the need for static batteries and incentivises a higher
solar PV penetration, as previously observed [20, 22].

3. Conclusions

When comparing alternative transition paths for the

European energy system with the same carbon budget, we
find that those including a gentle CO2 reduction are con-
Figure 6: Fourier power spectra of wind and solar PV genera- sistently around 300 Be cheaper than those paths where
tion, electricity and heating demand, as well as storage technolo- low targets in the initial period demand a sharper reduc-
gies dispatch. Time series represent the Europe-aggregated genera-
tion/demand for the Gentle path in 2050.
tion later. We found that up-to-date costs for wind and
the inclusion of highly resolved time series for balancing
make climate action with renewables more cost-effective
central technologies such as heat pumps and CHP units, than previously seen. The required renewable build rates
enables a faster conversion because it is easier to substitute to decarbonise the electricity and heating sectors corre-
one central heating unit than a myriad of individual do- spond to the highest historical country-level values, mak-
mestic systems, and facilitates long-term thermal energy ing the transition challenging yet feasible. We have shown
storage, via cheap large water pits, Figure 6, that help to that early action not only allows room for decision-making
balance the large seasonal variation of heating demand, later but it is also pays off.
Figure S23. So far, we have assumed that DH penetra-
tion remains constant at 2015 values. When DH is as- 4. Experimental Procedures
sumed to expand linearly so that in 2050 it supplies the
entire urban heating demand in every country, cumula- The system configuration is optimised by minimising
tive system cost for the Gentle path reduces by 238 Be . annualised system cost in every time step (one every 5
This roughly offsets the cost of extending and maintaining years), under the global CO2 emissions cap imposed by
the DH networks and avoids the additional expansion of the transition path under analysis (Figure 1). This can be
gas distribution networks. When a 2% reduction of space considered a myopic approach since the optimisation has
heating demand per year is assumed due to renovations no information about the future. The cumulative CO2
of the building stock, cumulative system cost decreases emissions for the Gentle and Sudden transition paths is
by 760 Be compared to paths with constant heating de- equal to a carbon budget of 21 GtCO2 . In every time step,
mand, significantly offsetting costs of renovations. When generation, storage, and transmission capacities in every
the model is allowed to optimise transmission capacities country are optimised assuming perfect competition and
after 2030, together with the generation and storage as- foresight as well as long-term market equilibrium. Besides
sets, the optimal configuration at the end of the paths the global CO2 emission cap, other constraints such as the
includes a transmission volume approximately three times demand-supply balance in every node, and the maximum
higher than that of 2030. Although the cumulative sys- power flowing through the links are imposed to ensure the
tem cost is 93 Be lower, it is unclear to what extent it feasibility of the solution, Procedures S2.
compensates the social acceptance issues associated with We use a one-node-per-country network, including 30
extending transmission capacities. Neither of the paths in- countries corresponding to the 28 European Union mem-
stall new nuclear capacity. This technology is only part of ber states as of 2018 excluding Malta and Cyprus but
the optimal system in 2050 when nuclear costs are lower by including Norway, Switzerland, Bosnia-Herzegovina, and
15% compared to the reference cost and no transmission Serbia (Figure S16). Countries are connected by High
capacity expansion is allowed. In all the previous scenar- Voltage Direct Current (HVDC) links whose capacities
ios, the difference in cumulative system cost for the Gentle can be expanded if it is cost-effective. In the power sec-
and Sudden path is roughly the same, Table S1. tor, electricity can be supplied by onshore and offshore
wind, solar photovoltaics (PV), hydroelectricity, Open Cy-
Transport. cle Gas Turbines (OCGT), Combined Cycle Gas Turbines
5
(CCGT), Coal, Lignite, and Nuclear power plants, and
Combined Heat and Power (CHP) units using gas, coal or
biomass. Electricity can be stored using Pumped Hydro
Storage (PHS), static electric batteries, and hydrogen stor-
age. Hydrogen is produced via electrolysers and converted
back into electricity using fuel cells. Methane can be pro-
duced by combining Direct Air Captured (DAC) CO2 and
electrolysed-H2 in the Sabatier reaction. Heating demand
is split into urban heating, corresponding to regions whose
population density allows district heating and rural heat-
ing where only individual solutions are allowed. Heating
can be supplied via central heat pumps, heat resistors, gas
boilers, solar collectors, and CHP units for urban regions,
while only individual heat pumps, electric boilers, and gas Figure 7: Model diagram representing the main generation and stor-
boilers can be used in rural areas. Central and individual age technologies in every country.
thermal energy storage can also be installed. A detailed
description of all the sectors is provided in Procedure S3. 6. Authors contribution
Costs assumed for the different technologies depend on
time (Procedure S4) but not on the cumulative installed M. Victoria designed the analysis, drafted the manuscript
capacity since we assume that they will be influenced by and contributed to the data acquisition, analysis and in-
the forecast global installation rates and learning curves. terpretation of data. K. Zhu contributed to the data ac-
The financial discount rate applied to annualise costs is quisition, modelling, analysis and interpretation of data.
equal to 7% for every technology and country. Although T. Brown, G. B. Andresen and M. Greiner contributed
it can be strongly impacted by the maturity of a technol- to the initial idea and made substantial revisions of the
ogy, including the country-specific experience on it, and manuscript.
the rating of a country [49], we assumed European coun-
tries to be similar enough to use a constant discount rate. 7. Acknowledgements
For decentral solutions, such as rooftop PV or small water
tanks, a discount rate equal to 4% is assumed. The al- M. Victoria, K. Zhu, G. B. Andresen and M. Greiner
ready installed capacities, i.e., existing capacities in 2020 are fully or partially funded by the RE-INVEST project,
or capacities installed in a previous year whose lifetime which is supported by the Innovation Fund Denmark un-
has not concluded, are exogenously included in the model. der grant number 6154-00022B. T.B. acknowledges fund-
For every time step, the total system cost includes annu- ing from the Helmholtz Association under grant no. VH-
alised and running cost for newly installed assets and for NG-1352. The responsibility for the contents lies solely
exogenously fixed capacities. To estimate the cumulative with the authors.
cost of every transition path, the annualised cost for all
year are added assuming a social discount rate of 2%. This References
rate represents the value at which we, as European society,
discount investments in far-future years when comparing [1] In-depth analysis in support of the Comission Communication
COM(2018) 773 A Clean Planet for all. A European long-term
them with present investments. We have selected a social strategic vision for a prosperous, modern, competitive and
discount rate of 2%, which is similar to the inflation rate climate neutral economy, Tech. rep. (Nov. 2018).
in the European Union, that averaged 2.4% in the past URL https://ec.europa.eu/clima/news/
20 years. The CO2 price is not an input to the model, commission-calls-climate-neutral-europe-2050_en
[2] Total greenhouse gas emissions, trends and projections, EEA.
but a result that is obtained via the Lagrange/Karush- URL https://www.eea.europa.eu/data-and-maps/
Kuhn-Tucker multiplier associated with the global CO2 indicators/greenhouse-gas-emission-trends-6/
constraint. assessment-2
[3] EU comission appraisal on national energy and climate plans,
2019.
5. Data and code availability URL https://europa.eu/rapid/press-release_IP-19-2993_
en.htm
[4] The European Green Deal.
The model is implemented in the open-source frame- URL https://ec.europa.eu/info/sites/info/files/
work Python for Power System Analysis (PyPSA) [50]. european-green-deal-communication_en.pdf
The model and data used in this paper can be retrieved [5] M. Warren, Thousands of scientists are backing the kids striking
from the repository https://github.com/martavp/pypsa- for climate change, Nature 567 (2019) 291–292. doi:10.1038/
d41586-019-00861-z.
eur-sec-30-path. URL http://www.nature.com/articles/d41586-019-00861-z
[6] Global Warming of 1.5◦ C, Intergovernmental Panel on Climate
Change (IPCC), Tech. rep. (2018).
URL https://www.ipcc.ch/sr15/

6
 [7] M. R. Raupach, S. J. Davis, G. P. Peters, R. M. Andrew, J. G. URL http://www.sciencedirect.com/science/article/pii/
Canadell, P. Ciais, P. Friedlingstein, F. Jotzo, D. P. Vuuren, S0301421512007677
C. L. Quéré, Sharing a quota on cumulative carbon emissions, [19] F. Cebulla, T. Naegler, M. Pohl, Electrical energy storage in
Nature Climate Change 4 (10) (2014) 873–879. doi:10.1038/ highly renewable European energy systems: Capacity require-
nclimate2384. ments, spatial distribution, and storage dispatch, Journal of
URL https://www.nature.com/articles/nclimate2384 Energy Storage 14 (2017) 211–223. doi:10.1016/j.est.2017.
[8] H. D. Matthews, Quantifying historical carbon and climate 10.004.
debts among nations, Nature Climate Change 6 (1) (2016) 60– URL http://www.sciencedirect.com/science/article/pii/
64. doi:10.1038/nclimate2774. S2352152X17302815
URL https://www.nature.com/articles/nclimate2774 [20] M. Victoria, K. Zhu, T. Brown, G. B. Andresen, M. Greiner,
[9] National emissions reported to the UNFCCC and to the EU The role of storage technologies throughout the decarbonisation
Greenhouse Gas Monitoring Mechanism , EEA. of the sector-coupled European energy system, Energy Con-
URL https://www.eea.europa.eu/data-and-maps/data/ version and Management 201 (2019) 111977. doi:10.1016/j.
national-emissions-reported-to-the-unfccc-and-to-the-eu-greenhouse-gas-monitoring-mechanism-15
enconman.2019.111977.
[10] E. Lantz, R. Wiser, M. Hand, The Past And Future Cost Of URL http://www.sciencedirect.com/science/article/pii/
Wind Energy, Tech. rep., NREL (2012). S0196890419309835
URL https://www.nrel.gov/docs/fy12osti/53510.pdf [21] D. Connolly, H. Lund, B. V. Mathiesen, Smart Energy Eu-
[11] F. Creutzig, P. Agoston, J. C. Goldschmidt, G. Luderer, rope: The technical and economic impact of one potential
G. Nemet, R. C. Pietzcker, The underestimated potential of 100% renewable energy scenario for the European Union, Re-
solar energy to mitigate climate change, Nature Energy 2 (9) newable and Sustainable Energy Reviews 60 (2016) 1634–1653.
(2017). doi:10.1038/nenergy.2017.140. doi:10.1016/j.rser.2016.02.025.
URL https://www.nature.com/articles/nenergy2017140 URL http://www.sciencedirect.com/science/article/pii/
[12] N. M. Haegel, H. Atwater, T. Barnes, C. Breyer, A. Burrell, S1364032116002331
Y.-M. Chiang, S. D. Wolf, B. Dimmler, D. Feldman, S. Glunz, [22] T. Brown, D. Schlachtberger, A. Kies, S. Schramm, M. Greiner,
J. C. Goldschmidt, D. Hochschild, R. Inzunza, I. Kaizuka, Synergies of sector coupling and transmission reinforcement in a
B. Kroposki, S. Kurtz, S. Leu, R. Margolis, K. Matsubara, cost-optimised, highly renewable European energy system, En-
A. Metz, W. K. Metzger, M. Morjaria, S. Niki, S. Nowak, ergy 160 (2018) 720–739. doi:10.1016/j.energy.2018.06.222.
I. M. Peters, S. Philipps, T. Reindl, A. Richter, D. Rose, URL http://www.sciencedirect.com/science/article/pii/
K. Sakurai, R. Schlatmann, M. Shikano, W. Sinke, R. Sinton, S036054421831288X
B. J. Stanbery, M. Topic, W. Tumas, Y. Ueda, J. v. d. [23] M. Child, C. Kemfert, D. Bogdanov, C. Breyer, Flexible elec-
Lagemaat, P. Verlinden, M. Vetter, E. Warren, M. Werner, tricity generation, grid exchange and storage for the transition
M. Yamaguchi, A. W. Bett, Terawatt-scale photovoltaics: to a 100% renewable energy system in Europe, Renewable En-
Transform global energy, Science 364 (6443) (2019) 836–838. ergy 139 (2019) 80–101. doi:10.1016/j.renene.2019.02.077.
doi:10.1126/science.aaw1845. URL http://www.sciencedirect.com/science/article/pii/
URL https://science.sciencemag.org/content/364/6443/ S0960148119302319
836 [24] R. Gross, R. Hanna, Path dependency in provision of do-
[13] E. H. Eriksen, L. J. Schwenk-Nebbe, B. Tranberg, T. Brown, mestic heating, Nature Energy 4 (5) 358–364. doi:10.1038/
M. Greiner, Optimal heterogeneity in a simplified highly renew- s41560-019-0383-5.
able European electricity system, Energy 133 (Supplement C) URL https://www.nature.com/articles/s41560-019-0383-5
(2017) 913–928. doi:10.1016/j.energy.2017.05.170. [25] Regulation and planning of district heating in Denmark, Tech.
URL http://www.sciencedirect.com/science/article/pii/ rep., Danish Energy Agency (2015).
S0360544217309593 URL https://ens.dk/sites/ens.dk/files/contents/
[14] D. P. Schlachtberger, T. Brown, S. Schramm, M. Greiner, The material/file/regulation_and_planning_of_district_
benefits of cooperation in a highly renewable European elec- heating_in_denmark.pdf
tricity network, Energy 134 (Supplement C) (2017) 469–481. [26] State and Trends of Carbon Pricing 2019, World Bank Group,
doi:10.1016/j.energy.2017.06.004. Tech. rep. (2019).
URL http://www.sciencedirect.com/science/article/pii/ URL https://openknowledge.worldbank.org/handle/10986/
S0360544217309969 31755
[15] H. C. Gils, Y. Scholz, T. Pregger, D. L. de Tena, D. Heide, [27] D. Bogdanov, J. Farfan, K. Sadovskaia, A. Aghahosseini,
Integrated modelling of variable renewable energy-based power M. Child, A. Gulagi, A. S. Oyewo, L. Barbosa, C. Breyer,
supply in Europe, Energy 123 (2017) 173 – 188. doi:https: Radical transformation pathway towards sustainable electricity
//doi.org/10.1016/j.energy.2017.01.115. via evolutionary steps, Nature Communications 10 (1) (2019)
URL http://www.sciencedirect.com/science/article/pii/ 1–16. doi:10.1038/s41467-019-08855-1.
S0360544217301238 URL https://www.nature.com/articles/
[16] T. W. Brown, T. Bischof-Niemz, K. Blok, C. Breyer, H. Lund, s41467-019-08855-1
B. V. Mathiesen, Response to burden of proof: A comprehensive [28] G. Pleßmann, P. Blechinger, How to meet EU GHG emission
review of the feasibility of 100% renewable-electricity systems, reduction targets? A model based decarbonization pathway for
Renewable and Sustainable Energy Reviews 92 (2018) 834–847. Europe’s electricity supply system until 2050, Energy Strategy
doi:10.1016/j.rser.2018.04.113. Reviews 15 (2017) 19–32. doi:10.1016/j.esr.2016.11.003.
URL http://www.sciencedirect.com/science/article/pii/ URL http://www.sciencedirect.com/science/article/pii/
S1364032118303307 S2211467X16300530
[17] R. A. Rodrı́guez, S. Becker, G. B. Andresen, D. Heide, [29] C. Gerbaulet, C. von Hirschhausen, C. Kemfert, C. Lorenz,
M. Greiner, Transmission needs across a fully renewable Eu- P. Y. Oei, European electricity sector decarbonization under
ropean power system, Renewable Energy 63 (2014) 467–476. different levels of foresight, Renewable Energy 141 (2019) 973–
doi:10.1016/j.renene.2013.10.005. 987. doi:10.1016/j.renene.2019.02.099.
URL http://www.sciencedirect.com/science/article/pii/ URL http://www.sciencedirect.com/science/article/pii/
S0960148113005351 S0960148119302538
[18] M. G. Rasmussen, G. B. Andresen, M. Greiner, Storage and [30] K. Poncelet, E. Delarue, D. Six, W. D’haeseleer, Myopic opti-
balancing synergies in a fully or highly renewable pan-European mization models for simulation of investment decisions in the
power system, Energy Policy 51 (2012) 642 – 651. doi:https: electric power sector, in: 13th International Conference on
//doi.org/10.1016/j.enpol.2012.09.009. the European Energy Market (EEM), 2016, pp. 1–9. doi:

7
 10.1109/EEM.2016.7521261. [44] K. Löffler, T. Burandt, K. Hainsch, P.-Y. Oei, Modeling the low-
[31] C. F. Heuberger, I. Staffell, N. Shah, N. M. Dowell, Impact carbon transition of the European energy system - A quantita-
of myopic decision-making and disruptive events in power sys- tive assessment of the stranded assets problem, Energy Strategy
tems planning, Nat Energy 3 (8) (2019) 634–640. doi:10.1038/ Reviews 26 100422. doi:10.1016/j.esr.2019.100422.
s41560-018-0159-3. URL http://www.sciencedirect.com/science/article/pii/
URL https://www.nature.com/articles/s41560-018-0159-3 S2211467X19301142
[32] V. Krey, F. Guo, P. Kolp, W. Zhou, R. Schaeffer, A. Awasthy, [45] C. McGlade, P. Ekins, The geographical distribution of fos-
C. Bertram, H.-S. de Boer, P. Fragkos, S. Fujimori, C. He, sil fuels unused when limiting global warming to 2◦ C, Nature
G. Iyer, K. Keramidas, A. C. Kberle, K. Oshiro, L. A. Reis, 517 (7533) (2015) 187–190. doi:10.1038/nature14016.
B. Shoai-Tehrani, S. Vishwanathan, P. Capros, L. Drouet, J. E. URL https://www.nature.com/articles/nature14016
Edmonds, A. Garg, D. E. H. J. Gernaat, K. Jiang, M. Kan- [46] S. Babrowski, T. Heffels, P. Jochem, W. Fichtner, Reducing
navou, A. Kitous, E. Kriegler, G. Luderer, R. Mathur, M. Mu- computing time of energy system models by a myopic approach,
ratori, F. Sano, D. P. van Vuuren, Looking under the hood: A Energy Systems 5 (1) 65–83. doi:10.1007/s12667-013-0085-1.
comparison of techno-economic assumptions across national and URL https://doi.org/10.1007/s12667-013-0085-1
global integrated assessment models, Energy 172 (2019) 1254– [47] F. Ueckerdt, R. Pietzcker, Y. Scholz, D. Stetter, A. Gian-
1267. doi:10.1016/j.energy.2018.12.131. nousakis, G. Luderer, Decarbonizing global power supply under
URL http://www.sciencedirect.com/science/article/pii/ region-specific consideration of challenges and options of inte-
S0360544218325039 grating variable renewables in the REMIND model, Energy Eco-
[33] S. Pfenninger, Energy scientists must show their workings, nomics 64 (2017) 665–684. doi:10.1016/j.eneco.2016.05.012.
Nature News 542 (7642) (2017) 393. doi:10.1038/542393a. URL http://www.sciencedirect.com/science/article/pii/
URL http://www.nature.com/news/ S014098831630130X
energy-scientists-must-show-their-workings-1.21517 [48] R. C. Pietzcker, D. Stetter, S. Manger, G. Luderer, Using the
[34] K. Zhu, M. Victoria, T. Brown, G. B. Andresen, M. Greiner, sun to decarbonize the power sector: The economic potential
Impact of CO2 prices on the design of a highly decarbonised cou- of photovoltaics and concentrating solar power, Applied Energy
pled electricity and heating system in Europe, Applied Energy 135 (2014) 704–720. doi:10.1016/j.apenergy.2014.08.011.
236 (2019) 622–634. doi:10.1016/j.apenergy.2018.12.016. URL http://www.sciencedirect.com/science/article/pii/
URL http://www.sciencedirect.com/science/article/pii/ S0306261914008149
S030626191831835X [49] F. Egli, B. Steffen, T. S. Schmidt, Bias in energy system models
[35] F. Gotzens, H. Heinrichs, J. Hrsch, F. Hofmann, Performing with uniform cost of capital assumption, Nature Communica-
energy modelling exercises in a transparent way - The issue of tions 10 (1) (2019) 1–3. doi:10.1038/s41467-019-12468-z.
data quality in power plant databases, Energy Strategy Reviews URL https://www.nature.com/articles/
23 (2019) 1–12. doi:10.1016/j.esr.2018.11.004. s41467-019-12468-z
URL http://www.sciencedirect.com/science/article/pii/ [50] T. Brown, J. Hörsch, D. Schlachtberger, PyPSA: Python for
S2211467X18301056 Power System Analysis, Journal of Open Research Software 6
[36] Renewable Capacity Statistics 2019, IRENA. (2018). doi:10.5334/jors.188.
URL https://www.irena.org/publications/2019/Mar/ URL https://doi.org/10.5334/jors.188
Renewable-Capacity-Statistics-2019
[37] D. Tong, Q. Zhang, Y. Zheng, K. Caldeira, C. Shearer, C. Hong,
Y. Qin, S. J. Davis, Committed emissions from existing en-
ergy infrastructure jeopardize 1.5 ◦ C climate target, Nature
572 (7769) 373–377. doi:10.1038/s41586-019-1364-3.
URL https://www.nature.com/articles/s41586-019-1364-3
[38] F. W. Geels, B. K. Sovacool, T. Schwanen, S. Sorrell, Sociotech-
nical transitions for deep decarbonization, Science 357 (6357)
(2017) 1242–1244. doi:10.1126/science.aao3760.
URL https://science.sciencemag.org/content/357/6357/
1242
[39] E. Vartiainen, G. Masson, C. Breyer, D. Moser, E. R. Medina,
Impact of weighted average cost of capital, capital expenditure,
and other parameters on future utility-scale PV levelised
cost of electricity, Progress in Photovoltaics: Research and
Applications (2017). doi:10.1002/pip.3189.
URL https://onlinelibrary.wiley.com/doi/abs/10.1002/
pip.3189
[40] Carbon price viewer.
URL https://sandbag.org.uk/carbon-price-viewer/
[41] T. Vandyck, K. Keramidas, A. Kitous, J. V. Spadaro, R. V.
Dingenen, M. Holland, B. Saveyn, Air quality co-benefits for
human health and agriculture counterbalance costs to meet
Paris Agreement pledges, Nature Communications 9 (1) (2018)
1–11. doi:10.1038/s41467-018-06885-9.
URL https://www.nature.com/articles/
s41467-018-06885-9
[42] F. Neumann, T. Brown, The Near-Optimal Feasible Space of a
Renewable Power System Model, arXiv:1910.01891 (2019).
URL http://arxiv.org/abs/1910.01891
[43] S. Simoes, W. Nijs, P. Ruiz, A. Sgobbi, D. Radu, P. Bolat,
C. Thiel, S. Peteves, The JRC-EU-TIMES model, assessing the
long-term role of the SET plan energy technologies.
URL https://ec.europa.eu/jrc/en/scientific-tool/
jrc-eu-times-model-assessing-long-term-role-energy-technologies

8
 1. Supplemental Tables

Table S1: Cumulative system costs (Be ) for additional analyses.

Analysis Gentle path Sudden path Difference Change relative to Base (Gentle)
Base 6,994 7,341 347
District heating expansion 6,756 7,069 313 -238
Heat savings due to building renovation 6,234 6,537 303 -760
Transmission expansion after 2030 6,901 7,205 304 -93
Including road and rail transport 7,428 7,860 432 434
arXiv:2004.11009v1 [physics.soc-ph] 23 Apr 2020
2. Supplemental Figures

Figure S1: Sectoral distribution of historical emissions in the European Union [1]. The black stars indicate committed EU reduction
targets, while white stars mark targets under discussion. LULUCF stands for land use, land-use change, and forestry.
Figure S2: Historical CO2 emissions from the supply of heating in the residential and services sector [1].
Figure S3: Historical share of technologies used to supply heating demand in the residential and services sector [2].
Figure S4: Photovoltaic annual build rates for those European countries with a prominent peak [3]. The sharp increases and subse-
quent decreases in the installation rates were caused by country-specific successive changes in the regulatory frameworks. See for
instance [4, 5].
Figure S5: Age distribution of European power plants in operation [3, 6] and required annual installation throughout the Gentle and
Sudden paths.
Figure S6: Annualised system cost for the European electricity and heating system throughout transition paths Gentle and Sudden
shown in Fig. 1 in the main text. Conventional includes costs associated with coal, lignite, and gas power plants producing electricity
as well as costs for fossil-fueled boilers and CHP units. Power-to-heat includes costs associated with heat pumps and heat resistors.
Balancing includes costs of electric batteries, H2 storage, and methanation.
Figure S7: Installed capacities for different technologies throughout transition paths shown in Fig. 1 in the main text.
Figure S8: Annual build rates for different technologies throughout transition paths shown in Fig. 1 in the main text.
Figure S9: Annual build rates for batteries and hydrogen storage throughout transition paths shown in Fig. 1 in the main text.
Figure S10: Annual synthetic methane production throughout transition paths shown in Fig. 1 in the main text.
Figure S11: Utilisation factors for lignite, coal, OCGT, CCGT, nuclear power plants and gas boilers throughout transition paths shown
in Fig. 1 in the main text.
Figure S12: Ratio of market revenues to total expenditure for lignite, coal, OCGT, CCGT, nuclear power plants and gas boilers
throughout transition paths shown in Fig. 1 in the main text. Total expenditure includes fixed and variable costs, fuel costs and cost
associated with CO2 price.
Figure S13: Ratio of market revenues to operational expenditure (OPEX) for lignite, coal, OCGT, CCGT, nuclear power plants and
gas boilers throughout transition paths shown in Fig. 1 in the main text. OPEX includes fixed and variable operation and maintenance
costs, fuel costs and cost associated with CO2 price.
Figure S14: Expenditures not recovered via market revenues for lignite, coal, OCGT, CCGT, nuclear power plants and gas boilers
throughout transition paths shown in Fig. 1 in the main text.
Figure S15: Estimated new jobs in wind, solar PV, and biomass throughout transition paths shown in Fig.1 in the main text.
Figure S16: Electricity generation in every country in 2050 for transition path Gentle. Fuel used in OCGT plants is synthetic methane
produced by combining electrolysed-H2 and direct-air-captured CO2 .
Figure S17: Evolution of the electricity generation mix in every country for the Gentle transition path.
Figure S18: Evolution of technologies used to supply heating in the residential and services sector in the Gentle path.
Figure S19: Annualised system cost for the European electricity, heating and transport system throughout transition paths Gentle and
Sudden. Conventional includes costs associated with coal, lignite, and gas power plants producing electricity as well as costs for
fossil-fueled boilers and CHP units. Power-to-heat includes costs associated with heat pumps and heat resistors. Balancing includes
costs of electric batteries, H2 storage, and methanation.
Figure S20: Levelized Cost of Energy (LCOE) for the European electricity, heating and transport system throughout transition paths
Gentle and Sudden. Conventional includes costs associated with coal, lignite, and gas power plants producing electricity as well as
costs for fossil-fueled boilers and CHP units. Power-to-heat includes costs associated with heat pumps and heat resistors. Balancing
includes costs of electric batteries, H2 storage, and methanation.
Figure S21: Installed capacities for different technologies throughout transition paths Gentle and Sudden when the electricity, heating
and transport sectors are included.
3. Supplemental Experimental Procedure

S1. CO2 restriction paths with equivalent budget

The carbon budget from now onwards for the generation of electricity and the supply of heating in the
residential and services sector in Europe accounts for 21 GtCO2 . It has been estimated based on a global
carbon budget of 800 GtCO2 to avoid temperature increments above 1.75◦ C relative to preindustrial period
with a probability of more than 66% [7]. The global budget is assumed to be split among regions according
to a constant per-capita ratio which translates into a 6% share for Europe [8]. Out of the total emissions in
Europe, the ratio corresponding to electricity and heating is considered constant and equal to present val-
ues. In 2017, electricity generation and heating in the residential and services sector emitted 1.56 GtCO2
which represents 43.5% of European emissions, [1] and Figure S1.

The B=21 GtCO2 budget can be utilised following different transition paths. One option consists in
assuming a linear CO2 restriction path. Emissions will then reach zero in tf
2B
tf = t0 + (1)
e0
where t0 =2020, and e0 represents the carbon emissions from electricity and heating sectors in 2020, which
are assumed to be the same as in 2017.

Alternatively, emissions can be assumed to follow a path defined by one minus the cumulative distribu-
tion function (CDFβ ) of a symmetric beta distribution in which β1 = β2 .

e(t) = e0 (1 − CDFβ (t))

Z t
CDFβ (t) = P DFβ (t)dt
t0 (2)
Γ(2β1 )
P DFβ (t) = (t − t0 )β1 −1 (tf − t)β1 −1
2Γ(β1 )
R∞
where Γ is the gamma function. The cumulative emissions fulfil t0 e(t)dt = B.

The third option considered for the transition path is an exponential decay, following Raupach et al. [8].
In that case, emissions evolve as:
e(t) = e0 (1 + (r + m)t)e−mt (3)
where r is the initial linear growth rate, which here is assumed to be r=0, and the decay parameter m is
determined by imposing the integral of the path to be equal to the budget.
Z ∞
B= e0 (1 + (r + m)t)e−mt dt
t0
q (4)
1 + 1 + rB e0
m= B
e0

Although the exponential decay path approaches asymptotically to zero, we assume here that e(2050) = 0.
By doing that, the final point of the different transition paths is equivalent and all of them achieve net-zero
emissions in the electricity and heating sectors by 2050.

S2. Model description

In every time step, the optimisation objective, that is, the total annualised system cost is calculated as:
 " #
X X X X
min cn,s · Gn,s + ĉn,s · En,s + c` · F` + on,s,t · gn,s,t
Gn,s ,En,s ,
F` ,gn,s,t n,s n,s ` n,s,t

where cn,s are the fixed annualised costs for generator and storage power capacity Gn,s of technology s
in every bus n, ĉn,s are the fixed annualised costs for storage energy capacity En,s , c` are the fixed an-
nualised costs for bus connectors F` , and on,s,t are the variable costs for generation and storage dispatch
gn,s,t in every hour t. Bus connectors ` include transmission lines but also converters between the buses
implemented in every country (see Figure S22), for instance, heat pumps that connect the electricity and
heating bus.

The optimisation of the system is subject to several constraints. First, hourly demand dn,t in every bus
n must be supplied by generators in that bus or imported from other buses. f`,t represents the energy flow
on the link l and αn,`,t indicates both the direction and the efficiency of flow on the bus connectors. αn,`,t
can be time-dependent such as in the case of heat pumps whose conversion efficiency depends on the
ambient temperature.
X X
gn,s,t + αn,`,t · f`,t = dn,t ↔ λn,t ∀ n, t (5)
s `

The Lagrange multiplier λn,t , also known as Karush-Kuhn-Tucker (KKT), associated with the demand
constraint indicates the marginal price of the energy carrier in the bus n, e.g., local marginal electricity price
in the electricity bus.

Second, the maximum power flowing through the links is limited by their maximum physical capacity F` .
For transmission links, f `,t = −1 and f¯`,t = 1, which allows both import and export between neighbouring
¯
countries. For a unidirectional converter e.g., a heat resistor, f `,t = 0 and f¯`,t = 1 since a heat resistor can
only convert electricity into heat. ¯

f `,t · F` ≤ f`,t ≤ f¯`,t · F` ∀ `, t . (6)

¯
Third, for every hour the maximum capacity that can provide a generator or storage is bounded by the
product between installed capacity Gn,s and availabilities g n,s,t , ḡn,s,t . For instance, for solar generators
g n,s,t is zero and ḡn,s,t refers to the capacity factor at time t¯
¯
g n,s,t · Gn,s ≤ gn,s,t ≤ ḡn,s,t · Gn,s ∀ n, s, t . (7)
¯
The maximum power capacity for generators is limited by potentials Ḡn,s that are estimated taking into
account physical and environmental constraints:

0 ≤ Gn,s ≤ Ḡn,s ∀ n, s . (8)

The storage technologies have a charging efficiency ηin and rate gn,s,t +
, a discharging efficiency ηout and
−
rate gn,s,t , possible inflow gn,s,t,inflow and spillage gn,s,t,spillage , and standing loss η0 . The state of charge
en,s,t of every storage has to be consistent with charging and discharging in every hour and is limited by
the energy capacity of the storage En,s . It should be remarked that the storage energy capacity En,s can
be optimised independently of the storage power capacity Gn,s .

+ −1 −
en,s,t = η0 · en,s,t−1 + ηin |gn,s,t | − ηout |gn,s,t | + gn,s,t,inflow − gn,s,t,spillage ,
0 ≤ en,s,t ≤ En,s ∀ n, s, t . (9)
 grid connection
electric bus

storage

electrolysis
discharge

fuel cell
charge
heat pump;
generators
resistive heater
transport hydrogen heat

generator/CHP
methanation

boiler/CHP
battery store hot water tank

methane

store

Figure S22: Energy flow at a single node representing a country. Within each node, there is a bus (thick horizontal line) for every
sector (electricity, transport and heating), to which different loads (triangles), energy sources (circles), storage units (rectangles) and
converters (lines connecting buses) are attached. This is equivalent to the diagram shown in Fig. 7 in the main text.

So far, Equations (5) to (9) represent mainly technical constraints but additional constraints can be imposed
to bound the solution.

The interconnecting transmission expansion can be limited by a global constraint

X
l` · F` ≤ CAPLV ↔ µLV , (10)
`

where the sum of transmission capacities F` multiplied by the lengths l` is bounded by a transmission vol-
ume cap CAPLV . In this case, the Lagrange/KKT multiplier µLV represents the shadow price of a marginal
increase in transmission volume.

The maximum CO2 allowed to be emitted by the system CAPCO2 can be imposed through the constraint
X gn,s,t X
εs + εs (en,s,t=0 − en,s,t=T ) ≤ CAPCO2 ↔ µCO2 (11)
n,s,t
ηn,s n,s

where εs represents the specific emissions in CO2 -tonne-per-MWhth of the fuel s, ηn,s the efficiency and
gn,s,t the generators dispatch. In this case, the Lagrange/KKT multiplier represents the shadow price of
CO2 , i.e., the additional price that should be added for every unit of CO2 to achieve the CO2 reduction
target in an open market.

S2.1. Icon design acknowledgement

Icons used in Fig. 7 in the main text were retrieved from the Noun Project. We acknowledge the
followings icons and authors: solar energy by fauzan akbar, wind energy by fauzan akbar, Air Conditioner
by Arthur Shlain, Radiator by Nicolas LEULIET, Gas boiler by ProSymbols, cogeneration by Fabio Rinaldi,
Water Tank by Luis Solorio, electric vehicle by Adrien Coquet, Dam by iconsmind.com, water heater by
Stepan Voevodin, nitrogen by Edwin PM, Battery by notplayink!, transmission lline by DARAYANI, Tank
by Fabio Rinaldi, Methane by Michael Wohlwend, Hydropower by Georgiana Ionescu, Rabbit by Marco
Galtarossa, tortoise by Christopher T. Howlett.

S3. Sectors description and data

S3.1. Electricity sector
Hourly electricity demand for every country corresponding to 2015 is retrieved from EU Network Trans-
mission System Operators of Electricity (ENTSO-E) via the convenient dataset prepared by the Open Power
System Data (OPSD) initiative [9]. In every country, electricity can be generated by solar PV, onshore wind,
offshore wind, Open Cycle Gas Turbines (OCGT), Combined Cycle Gas Turbines (CCGT), coal, lignite, and
nuclear power plants and CHP units using either gas, coal or biomass. Their costs, lifetimes and efficien-
cies are shown in Tables S5 and S6. To represent scheduled shut-downs, a constant 90% availability is
assumed for nuclear power plants.
Country-wise onshore and offshore wind capacity factor time series are modelled by converting wind
velocity from Climate Forecast System Reanalysis (CFSR) [10] into wind generation, following the method-
ology described in [11]. CFSR database comprises hourly resolution and spatial resolution equal to
0.3125◦ x0.3125◦ , which in Europe roughly corresponds to 40x40 km2 . For every country, a capacity layout
proportional to wind resource is assumed. Following [12], large countries are divided into up to 4 regions
sorted by wind resource. Independent classes of generators with different time series and average full load
hours are added to a single node representing a country. Their optimised capacities are later aggregated
on a country level for analysis.
Time series representing the hourly capacity factors for solar PV were obtained by converting bias-
corrected CFSR reanalysis irradiance into solar electricity generation, assuming a uniform capacity layout
across every country. Details on the conversion and aggregation methodology can be found in [13], the
complete time series dataset is available in 10.5281/zenodo.1321809. 50% of PV capacity in every country
is assumed to be utility-scale installations and 50% rooftop PV systems with the cost and characteristics
gathered in Tables S5 and S6. The discount rate is assumed to be 7% for the former and 4% for the latter.
As discussed in [14] the impact of this assumption is limited.
The maximum capacity for onshore wind, offshore wind, and solar PV that can be installed in every
country is limited by the estimated potentials. Those are determined by summing the available land in every
reanalysis grid cell, which in turn is calculated by considering only the suitable land for every technology,
according to the Corine Land Cover (CLC) database [15] and subtracting Natura 2000 protected areas [16].
For onshore and offshore wind, the potential is calculated as 20% of the available land. For solar PV, 3%
of the available land is used for estimating the potential.
X
P otentialn,P V = 0.03(ACLC,P
i
V
− AN
i
atura2000
) for i ∈ n (12)
i

where ACLC,P
i
V
is the area of the grid cell belonging to PV categories in the CLC database, Table S2, and
N atura2000
Ai is the area of the grid cell protected by the Natura 2000 network.
X
P otentialn,wind = 0.2(ACLC,wind
i − ANi
atura2000
)kn for i ∈ n (13)
i

For wind, kn is a coefficient calculated by imposing the condition that in none of the grid cells the installed
capacity surpasses the potential. This represents a conservative approach. Higher potentials could be
attained if assumed capacity layout is not proportional to the wind resource. For offshore wind, only areas
whose sea depth is lower than 50 m are considered as valid.

Table S2: Land types considered suitable for every technology. Categories in Corine Land Cover database
[15] are selected following [17].

Solar PV artificial surfaces (1-11), agriculture land except for those areas already occupied
by agriculture with significant natural vegetation and agro-forestry areas (12-20),
natural grasslands (26), bare rocks (31), and sparsely vegetated areas (32)
Onshore wind agriculture areas (12-22), forests (23-25), scrubs and herbaceous vegetation as-
sociations (26-29), bare rocks (31), and sparsely vegetated areas (32)
Offshore wind sea and ocean (44)

Reservoir hydropower and run-of-river capacities are exogenously fixed at their values in 2015. Hourly
inflow is modelled based on rainfall in the CFSR data set as described in [18]. CHP units are modelled
as extraction condensing units, the feasible space representing the possible combinations of power and
heat outputs is included as a constraint in the model, as detailed in [18]. Electricity can be stored in static
batteries, hydrogen storage and Pumped Hydro Storage (PHS). The capacity of the later in every country is
exogenously fixed at 2015 values. Alkaline electrolysers are assumed since they have a lower cost [19] and
higher cumulative installed capacity [20] than PEM electrolysers. Hydrogen can be stored in overground
steel tanks or underground salt caverns [20]. For the latter, energy capacities in every country are limited
to the potential estimation for onshore salt caverns within 50 km of shore to avoid environmental issues as-
sociated with brine solution disposal, see Figure 7 in [21]. Electricity can also be used to produce methane
by combining hydrogen and direct air captured (DAC) CO2 in the Sabatier reaction. Following [18], the
energy consumed in DAC is taken into account by reducing the efficiency of the Sabatier reaction to 60%.
Alternative CO2 sources, such as capturing industry process emissions or biomass-related emissions, are
not included.

The transmission links between countries are assumed to be high-voltage direct current (HVDC) con-
nections. The lengths l` are set by the distance between the geographical mid-points of each country so
that some of the transmission within each country is also reflected in the optimisation. A factor of 25% is
added to the line lengths to account for the fact that transmission lines cannot be placed as the crow flies
due to land use restriction. For the transmission line capacities F` , a safety margin of 33% of the installed
capacity is used to satisfy n-1 requirements [22]. For 2020 and 2030, the capacities correspond to the
values assumed in the ENTSOE Ten-Year Network Development Plan (TNYDP), see Table S3 and [23].
The values for 2025 are interpolated assuming a linear capacity expansion between 2020 and 2030 for
every link. For years from 2035 onwards, capacities are fixed at 2030 values.
 Table S3: Transmission capacities (MW) for interconnections [23].

Link 2020 2030 Link 2020 2030 Link 2020 2030

AL-GR 250 250 FI-EE 1000 1000 LU-FR 0 0
AL-ME 350 350 FI-NO 0 0 LV-EE 1600 1600
AL-MK 200 200 FI-SE 2300 2800 LV-LT 1200 1800
AL-RS 760 760 FR-BE 4300 4300 ME-AL 350 350
AT-CH 1700 1700 FR-CH 3700 3700 ME-BA 400 400
AT-CZ 1000 1000 FR-DE 3000 4800 ME-IT 1200 1200
AT-DE 5000 7500 FR-ES 5000 8000 ME-RS 1000 1000
AT-HU 1200 1200 FR-GB 5400 5400 MK-AL 200 200
AT-IT 555 1655 FR-IE 0 700 MK-BG 150 150
AT-SI 1200 1200 FR-IT 4350 4350 MK-GR 400 400
BA-HR 1344 1844 FR-LU 380 380 MK-RS 1050 1050
BA-ME 500 500 GB-BE 1000 1000 NI-GB 80 500
BA-RS 1100 1100 GB-DK 1400 1400 NI-IE 1100 1100
BE-DE 1000 1000 GB-FR 5400 5400 NL-BE 2400 2400
BE-FR 2800 2800 GB-IE 500 500 NL-DE 4450 5000
BE-GB 1000 1000 GB-IS 0 0 NL-DK 700 700
BE-LU 1080 1080 GB-NI 500 500 NL-GB 1000 1000
BE-NL 2400 2400 GB-NL 1000 1000 NL-NO 700 700
BG-GR 1728 1728 GB-NO 1400 1400 NO-DE 1400 1400
BG-MK 530 530 GR-AL 250 250 NO-DK 1640 1640
BG-RO 1400 1400 GR-BG 1032 1032 NO-FI 0 0
BG-RS 600 600 GR-CY 2000 2000 NO-GB 1400 1400
CH-AT 1700 1700 GR-IT 500 500 NO-NL 700 700
CH-DE 4700 4700 GR-MK 350 350 NO-SE 3695 3695
CH-FR 1300 1300 HR-BA 1312 1812 PL-CZ 600 600
CH-IT 6240 6240 HR-HU 2000 2000 PL-DE 3000 3000
CY-GR 2000 2000 HR-IT 0 0 PL-DK 0 0
CZ-AT 1200 1200 HR-RS 600 600 PL-LT 1000 1000
CZ-DE 2100 2600 HR-SI 2000 2000 PL-PL 5000 5000
CZ-PL 500 500 HU-AT 800 800 PL-SE 600 600
CZ-SK 2100 2100 HU-HR 2000 2000 PL-SK 990 990
DE-AT 5000 7500 HU-RO 1300 1300 PT-ES 3500 3500
DE-BE 1000 1000 HU-RS 600 600 RO-BG 1500 1500
DE-CH 3286 3286 HU-SI 1700 1700 RO-HU 1400 1400
DE-CZ 1500 2000 HU-SK 2000 2000 RO-RS 1450 1450
DE-DK 4000 4000 IE-FR 0 700 RS-AL 330 330
DE-FR 3000 4800 IE-GB 500 500 RS-BA 1200 1200
DE-LU 2300 2300 IE-NI 1100 1100 RS-BG 350 350
DE-NL 4450 5000 IS-GB 0 0 RS-HR 600 600
DE-NO 1400 1400 IT-AT 385 1385 RS-HU 600 600
DE-PL 2000 2000 IT-CH 3860 3860 RS-ME 1100 1100
DE-SE 615 1315 IT-FR 2160 2160 RS-MK 950 950
DK-DE 4000 4000 IT-GR 500 500 RS-RO 1050 1050
DK-DK 1200 1200 IT-HR 0 0 SE-DE 615 1315
DK-GB 1400 1400 IT-IT 5750 5750 SE-DK 1980 1980
DK-NL 700 700 IT-ME 1200 1200 SE-FI 2400 3200
DK-NO 1640 1640 IT-SI 1380 1380 SE-LT 700 700
DK-PL 0 0 IT-TN 0 0 SE-NO 3995 3995
DK-SE 2440 2440 LT-LV 1500 2100 SE-PL 600 600
EE-FI 1016 1016 LT-PL 1000 1000 SI-AT 1200 1200
EE-LV 1600 1600 LT-SE 700 700 SI-HR 2000 2000
ES-FR 5000 8000 LU-BE 700 700 SI-HU 2000 2000
ES-PT 4200 4200 LU-DE 2300 2300 SI-IT 1530 1530

S3.2. Heating sector

Annual heat demands for European countries are retrieved from [24]. They are converted into hourly
heat demand based on the population-weighted [25] Heating Degree Hour (HDH), that is, heating is as-
sumed to be proportional to the difference between ambient temperature and a threshold temperature.
17◦ C is assumed as threshold temperature. Ambient temperature is read from the same reanalysis
database [10] used to model wind and solar PV time series. The estimated values for total annual de-
mand in Europe are similar for electricity and heating, accounting for 2854 TWhel /a and 3562 TWhth /a
respectively but heating demand shows a much more pronounced seasonal variation, see Figure S23.
For every country, heating demand is split between low-population density areas and high-population
density areas. 44.6% of the European population is estimated to live in the latter [18] where district heating
systems can be deployed. In high-density population areas, heating can be supplied by central ground-
sourced heat pumps, heat resistors and gas boilers, as well as by CHP units. All the previous technologies
are assumed to be integrated into district heating networks. Furthermore, individual air-sourced heat pumps
are also allowed in those areas. In low-density population areas, heating can be supplied by individual
ground-sourced heat pumps, heat resistors and gas boilers. Costs, lifetimes, and efficiencies of the different
technologies are included in Tables S5 and S6.
The Coefficient of Performance (COP) of heat-pumps depends on ambient or ground temperature to
capture the lower COP in winter. COP depends on the difference between the source and the sink temper-
atures ∆T = Tsink − Tsource . For air-sourced heat pumps (ASHP), COP = 6.81 + 0.121∆T + 0.000630∆T 2 ,
for ground-sourced heat pumps (GSHP), COP = 8.77 + 0.150∆T + 0.000734∆T 2 [26]. The sink water tem-
perature is assumed to be Tsink = 55◦ C, the source temperature for air and ground is taken from the same
reanalysis database used to estimate heating demand [10]. Thermal energy can be stored in large water
pits associated with district heating systems and individual thermal energy storage (TES), i.e., small water
tanks. A thermal energy density of 46.8 kWhth /m3 is assumed, corresponding to a temperature difference
of 40 K. The decay of thermal energy 1 − exp(− 24τ 1
) is assumed to have a time constant of τ =180 days for
central TES and τ =3 days for individual TES. Charging and discharging efficiencies are 90% due to pipe
losses.
Capacities already existing for technologies supplying heat are retrieved from [27]. For the sake of
simplicity, coal, oil and gas boilers capacities are assimilated to gas boilers. Besides that, existing capacities
for heat resistors, ASHP, and GSHP are included in the model. For high-density population areas, the
penetration of district-heating is assumed to remain fixed at 2015 values, [28] and Table S4. Cooling
demand is currently supplied by electricity so it is included in the electricity demand time series. It is
assumed to remain constant throughout the paths. For a thorough discussion of the impact of changing
cooling demand, the reader is referred to [29].

Figure S23: Electricity, rural and urban heating, and cooling demands for Europe.

S3.3. Biomass
Solid biomass can be burnt in CHP or central heating plants associated with district heating systems or
in power plants to produce electricity. The model does not include biogas that could be burnt or upgraded
 Table S4: Current penetration of district
heating in European countries [28].

Country District heating penetration

AT 0.14
BA 0.0
BE 0.0
BG 0.16
CH 0.04
CZ 0.4
DE 0.14
DK 0.64
EE 0.52
ES 0.0
FI 0.39
FR 0.06
GB 0.02
GR 0.0
HR 0.07
HU 0.12
IE 0.0
IT 0.03
LT 0.56
LU 0.0
LV 0.3
NL 0.04
NO 0.03
PL 0.41
PT 0.0
RO 0.23
RS 0.27
SE 0.51
SI 0.09
SK 0.54

into biomethane. A conservative approach is followed to estimate biomass potentials in every country.
From the JRC-ENSPRESO database [30, 31], the potential estimations for 2030 in the scenario ‘medium’
are retrieved, but only the types of biomass which are not competing with crops are considered valid.
In essence, biomass potentials include only the following items: primary agricultural residues, primary
and secondary forestry energy residues including sawdust, forestry residues from landscape care, and
municipal waste.

S3.4. Existing power plants and decommissioning

For conventional technologies, i.e. OCGT, CCGT, coal, lignite, nuclear and gas CHP, installed capacities
in every country in 2020 and commissioning dates are retrieved from [6]. A two-step method was imple-
mented to fill commissioning date for power plants whose data was missing. First, for units larger than
50 MW, commissioning dates have been searched and manually added. Then, for smaller units, a Kernel
Density Estimation (KDE) approach is used. In essence, for every technology and country, the units with
available data are used to create a distribution, which is then used to assign an estimated commissioning
date for those units with missing data. For solar PV, the installed capacities in 2020 and the installation
dates were obtained by processing annual installed capacities statistics from [3]. For offshore and onshore
wind, capacities and age are retrieved from [32]. Existing power plants are assumed to be decommis-
sioned at their corresponding commissioning date plus lifetime (Table S6). When a power plant has been
retrofitted, we assume that its operating life is extended by half of its nominal lifetime. For heating capaci-
ties, 25% of existing capacities in 2015 are assumed to be decommissioned in every 5-year time step after
2020.
 Figure S24: Age distribution of European power plants in operation [3, 6].

S3.5. Transport sector

The transport sector is included only in the final analysis of the paper. In that case, road and rail trans-
port are considered to be electrified at a rate equal to the CO2 reduction in the heating and electricity
sectors relative to 2020. In this way, transport-related CO2 emissions sink in parallel to the other sectors.
Annual energy demands from road and rail transport for every country are retrieved from [33]. Aviation,
shipping, and pipe transport are not included in the model. A country-specific factor (averaging 3.5) is used
to account for the increased efficiency when electrifying transport. Country-specific factors are computed
by comparing the current car final energy consumption per km in [33] (averaging 0.7 kWh/km) to the 0.2
kWh/km value assumed for plug-to-wheels efficiency in EVs. The characteristic weakly profile provided
by the German Federal Highway Research Institute (BASt) [34] is used to obtain hourly time series for
European countries taking into account the corresponding local times. Furthermore, a temperature depen-
dence is included in the time series to account for heating/cooling demand in transport. For temperatures
below/above 15◦ C/20◦ C, temperature coefficients of 0.63%/◦ C and 0.98%/◦ C are assumed, see [18] for
more details. When fully electrified, the annual electricity demand from transport sector in Europe accounts
for 1,102 TWh/a.
At every time step, the internal-combustion vehicles transformed into battery electric vehicles (BEV)
are assumed to include a battery with a storage capacity of 50 kWh, charging capacity of 11 kW, and 90%
charging efficiency. It is considered that half of the existing BEV of them can shift their charging time as well
as discharge into the grid to facilitate the operation of the system and reduce its total cost. Furthermore,
it is assumed that, at every time step, 25% of the existing BEV can provide vehicle-to-grid (v2g) services.
The BEV state of charge is forced to be higher than 75% at 5 a.m. every day, through en,s,t in Equation
(9), to ensure that the batteries are full in the morning peak usage. This also restricts BEV demand to be
shifted within a day and prevent EV batteries from becoming seasonal storage. The percentage of BEV
connected to the grid at any time is inversely proportional to the transport demand profile, which translates
into an average/minimum availability of 80%/62%. This approach is conservative compared to most of the
literature. For instance, in [35] the average parking time of the European fleet of vehicles is estimated at
92%. The cost of the EV batteries is not included in the model since it is assumed that EV owners buy
them to satisfy their mobility needs.
S3.6. Levelised Cost of Energy (LCOE)
The Levelised Cost of Energy is defined as the total system cost per unit of consumed energy, that is,
including supplied electricity and heating demand.

S4. Cost assumptions

Table S5: Overnight investment cost assumptions per technology and year. All costs are given in real 2015 money.

Technology Unit 2020 2025 2030 2035 2040 2045 2050 source
Onshore Wind e /kWel 1118 1077 1035 1006 977 970 963 [19]
Offshore Wind e /kWel 2128 2031 1934 1871 1808 1792 1777 [19]
Solar PV (utility-scale) e /kWel 398 326 254 221 188 169 151 [36]
Solar PV (rooftop) e /kWel 1127 955 784 723 661 600 539 [37]
OCGT e /kWel 453 444 435 429 423 417 411 [19]
CCGT e /kWel 880 855 830 822 815 807 800 [19]
Coal power plant e /kWel 3845 3845 3845 3845 3845 3845 3845 [38]
Lignite e /kWel 3845 3845 3845 3845 3845 3845 3845 [38]
Nuclear e /kWel 7940 7940 7940 7940 7940 7940 7940 [38]
Reservoir hydro e /kWel 2208 2208 2208 2208 2208 2208 2208 [39]
Run of river e /kWel 3312 3312 3312 3312 3312 3312 3312 [39]
PHS e /kWel 2208 2208 2208 2208 2208 2208 2208 [39]
Gas CHP e /kWel 590 575 560 550 540 530 520 [19]
Biomass CHP e /kWel 3500 3400 3300 3224 3150 3075 3000 [19]
Coal CHP e /kWel 1900 1880 1860 1841 1822 1803 1783 [19]
Biomass central heat plant e /kWel 890 865 840 820 800 780 760 [19]
Biomass power plant e /kWel 3500 3400 3300 3224 3150 3075 3000 [19]
HVDC overhead e /MWkm 400 400 400 400 400 400 400 [40]
HVDC inverter pair e /MW 150000 150000 150000 150000 150000 150000 150000 [40]
Battery storage e /kWh 232 187 142 118 94 84 75 [19]
Battery inverter e /kWel 270 215 160 130 100 80 60 [19]
Electrolysis e /kWel 600 575 550 537 525 512 500 [19]
Fuel cell e /kWel 1300 1200 1100 1025 950 875 800 [19]
H2 storage underground e /kWh 3.0 2.5 2.0 1.8 1.5 1.4 1.2 [19]
H2 storage tank e /kWh 57 50 44 35 27 24 21 [19]
DAC (direct-air capture) e /(tCO2 /a) 250 250 250 250 250 250 250 [41]
Methanation e /kWH2 1000 1000 1000 1000 1000 1000 1000 [42]
Central gas boiler e /kWth 70 65 60 60 60 60 60 [19]
Decentral gas boiler e /kWth 312 304 296 289 282 275 268 [19]
Central resistive heater e /kWth 70 65 60 60 60 60 60 [19]
Decentral resistive heater e /kWth 100 100 100 100 100 100 100 [42]
Central water tank storage e /kWh 0.6 0.6 0.5 0.5 0.5 0.5 0.5 [19]
Decentral water tank storage e /kWh 18 18 18 18 18 18 18 [19, 43]
Decentral air-sourced heat pump e /kWth 940 894 850 827 804 782 760 [19]
Central ground-sourced heat pump e /kWth 657 625 592 577 562 547 532 [19]
Decentral ground-sourced heat pump e /kWth 1500 1450 1400 1349 1299 1250 1200 [19]
 Table S6: Efficiency, lifetime and FOM cost per technology (values shown corresponds to 2020).

Technology FOMa Lifetime Efficiency Source

[%/a] [a]
Onshore Wind 1.3 27 [19]
Offshore Wind 1.9 27 [19]
Solar PV (utility-scale) 3.0 30 [36]
Solar PV (rooftop) 2.0 30 [37]
OCGT 1.8 25 0.42 [19]
CCGT 3.3 25 0.59 [19]
Coal power plant 1.6 40 0.33 [38]
Lignite 1.6 40 0.33 [38]
Nuclear 1.4 40 0.33 [38]
Reservoir hydrob 1.0 80 0.9 [39]
Run of riverb 2.0 80 0.9 [39]
PHSb 1.0 80 0.75 [39]
Gas CHPc 3.3 25 [19]
Biomass CHPc 3.6 25 [19]
Coal CHPc 1.6 25 [19]
Biomass central heat plant 5.8 25 1.0 [19]
Biomass power plant 3.6 25 0.31 [19]
HVDC overhead 2.0 40 [40]
HVDC inverter pair 2.0 40 [40]
Battery storage 0.0 20 [19]
Battery inverter 0.2 20 0.9 [19]
Electrolysis 5.0 25 0.8 [19, 44]
Fuel cell 5.0 10 0.58 [19, 44]
H2 storage underground 2.0 100 1.0 [19]
H2 storage tank 1.1 25 [19]
DAC (direct-air capture)e 4.0 30 [41]
Methanation 3.0 25 0.6 [42]
Central gas boiler 2.8 25 1.0 [19]
Decentral gas boiler 0.1 20 0.97 [19]
Central resistive heater 1.5 20 0.99 [19]
Decentral resistive heater 2.0 20 0.9 [42]
Central water tank storage 0.5 20 [19]
Decentral water tank storage 1.0 20 [19, 43]
Water tank charger/discharger 0.9
Decentral air-sourced heat pumpd 0.0 18 [19]
Central ground-sourced heat pumpd 0.3 25 [19]
Decentral ground-sourced heat pumpd 0.0 20 [19]
a Fixed Operation and Maintenance (FOM) costs are given as a percentage of the overnight cost per
year.
b Hydroelectric facilities are not expanded in this model and are considered to be fully amortized.
c Efficiency for Combined Heat and Power (CHP) plants depends on the electricity/heat output and it is

modelled as described in the text.

d Coefficient of performance (COP) of heat pumps is modelled as a function of temperature, as described

in the text.
e Investments in methanation and DAC are not allowed independently, only together as ‘Methana-

tion+DAC’, see text.

 Table S7: Costs and emissions coefficient of fuels.

Fuel Cost Source Emissions Source

[e /MWhth ] [tCO2 /MWhth ]
coal 8.2 [45] 0.336 [46]
lignite 2.9 [39] 0.407 [46]
gas 20.1 [45] 0.201 [46]
oil 50.0 [47] 0.266 [46]
nuclear 2.6 [38] 0
solid biomassa 25.2 [30, 48] 0
a
Raw biomass fuel cost is assumed as the middle value of the range
provided in the references for different European countries and types of
sustainable biomass.

Figure S25: Evolution of annualised costs, relative to 2020, for some selected technologies.

S4.1. Distribution networks

Half of the solar PV capacity is assumed to be installed in rooftops. Consequently, the PV deploy-
ment that we observe will require extending and increasing the capacity of electricity distribution networks.
The deployment of battery electric vehicles (BVEs) also requires enhancing the distribution networks. Dis-
tributed PV generation will contribute to charging BEVs and this could reduce the need for an increased
distribution capacity. The cost of extending distribution networks is estimated in 140 e /kWP V , with 30 years
lifetime and 3% FOM [19, 49]. In the model, the cost of expanding and maintaining distribution networks are
not included in the optimisation but calculated based on the optimal PV capacity and added ex-post. For
the Gentle path, distribution networks costs represent 140 Be , that is, 2.6% of the cumulative system cost.
District heating and gas distribution networks are not included in the optimisation, but the cost of expanding
those networks is estimated and compared with the system costs in the scenarios with and without district
heating expansion, see Table S1.

S4.2. Up-to-date cost assumptions for solar PV

The combination of rapid learning with faster-than-expected capacity deployment has led to lower-than-
expected costs for solar PV. The potential of this technology has been repeatedly underestimated by the
International Energy Agency [50], Greenpeace [51], and PV scientists [52]. The investment cost for utility-
scale solar PV in 2020 is estimated in the range of 398-423 e /kW [19, 37]. For rooftop PV installations in
Europe, a wider range is found, 1070-1127 e /kW [19, 37, 53], due to the higher impact of local experience
and labour costs. Solar PV is forecast to achieve an investment cost between 151 e /kW [36] and 241
e /kW [19] for large installations in 2050.

Assuming outdated costs for solar PV is also known to be a flaw of most Integrated Assessment Models
(IAMs) and it can have a huge impact on the results. Creutzig et al. already pointed out this problem in
[51] where they found similar solar PV penetration to ours when up-to-date costs are assumed for solar PV.
Breyer and co-workers have also emphasized the key role that solar PV plays in decarbonisation paths in
Europe [54] and globally [55] when proper costs are assumed. However, the problem persists. For instance,
the PRIMES model used in the report supporting the Clean Planet for All strategy of the EU Commission
[56] assumes 407-495 e /kW in 2050 [57], which is higher than the lower range value for today’s costs. Even
more worrying are the findings by Krey et al. [58]. The authors review the techno-economic assumptions
in the electricity sector among fifteen different global and national IAMs. Figure 4 in [58] shows that most
of the reviewed IAMs include cost assumptions for solar PV in 2050 close to 1000 e /kW. Although they do
not specify if the cost refers to utility-scale or rooftop installations, the values are twice as high as the cost
already achieved by this technology in large installations.

S4.3. Discount rate

We use a financial discount rate to annualise the cost of every asset including generation, storage, and
transmission. The financial discount rate is equal to 7% for all the investment, except decentral assets
such as rooftop PV or individual water tanks for which 4% is assumed. Discount rates are influenced by
the economic situation in a country and this could strongly impact the results [59]. Moreover, the accumu-
lated experience with a certain technology, globally and inside the country, affects the perceived risks and
the cost of capital for that technology. When investigating transition paths, other authors have considered
sector-specific discount rates although they do not update them as the technology becomes more mature
[56]. It is extremely difficult to meaningfully estimate country-specific discount rates for future years. Hence,
we have chosen a constant 7% discount rate for all the technologies and countries. Schyska and Kies have
assessed the impacts of different cost of capital on the optimal European power system [60].

Besides the financial discount rate, a different social discount rate is used to calculate the cumulative
system cost. This is common practice when comparing transition paths derived from IAMs and energy
models [56, 61]. We have selected a social discount rate of 2%, which is similar to the inflation rate in
the European Union, that averaged 2.4% in the past 20 years. This is in agreement with reference [62],
in which the impact of discount rates for emissions pathways and negative emissions are analysed. The
authors recommend using low social discount rates, around 2%. It is worth remarking that the cumulative
cost remains lower for the Gentle path provided that discount rates lower than 15% are assumed.

S4.4. Jobs creation

Currently, the total number of renewable energy jobs in member states of the European Union is esti-
mated at 1.2 million [63]. In [64], a systematic review of literature is conducted and the average number of
full-time-equivalent (FTE) jobs associated with every renewable technology is provided. FTE jobs include
both installation as well as operation and maintenance jobs. Using that data, and assuming 40-year work
life, we have estimated the newly created jobs associated with the expansion of solar PV, wind and biomass
capacities. Cumulative new jobs represent 1.9 and 1.8 million for the Gentle and Sudden path respectively.
This is in agreement with the analysis included in the Clean Energy for All strategy [56] which estimates
the creation of 2.1 million jobs under the 1.5◦ C scenario. As other technological sectors, renewables suffer
from a major gender imbalance. The transition is expected to have a positive impact on this aspect since
women represent 32% of the renewable energy workforce while they only account for 22% of the workforce
in the oil and gas industry [65].
 Table S8: Technology assumptions and average jobs creation based on the systematic literature review in [64].

technology lifetime (years) annual CF full-time-equivalent jobs (jobs/GWh)

solar PV (rooftop) 25 0.11 2
solar PV (utility-scale) 25 0.11 0.5
wind 20 0.33 0.5
biomass 30 0.8 0.2

Supplemental References
[1] National emissions reported to the UNFCCC and to the EU Greenhouse Gas Monitoring Mechanism , EEA.
URL https://www.eea.europa.eu/data-and-maps/data/national-emissions-reported-to-the-unfccc-and-to-the-eu-greenhouse-gas-monito
[2] L. Mantzos, T. Wiesenthal, N. Matei, S. Tchung-Ming, M. Rzsai, H. P. Russ, A. Soria, JRC-IDEES: Integrated Database of the
European Energy Sector (2017). doi:10.2760/182725.
URL http://www.sciencedirect.com/science/article/pii/S0360544216310295
[3] Renewable Capacity Statistics 2019, IRENA.
URL https://www.irena.org/publications/2019/Mar/Renewable-Capacity-Statistics-2019
[4] Photovoltaics Report, Tech. rep., Fraunhofer ISE (2019).
URL https://www.ise.fraunhofer.de/content/dam/ise/de/documents/publications/studies/Photovoltaics-Report.
pdf
[5] M. Victoria, C. Gallego, I. Anton, G. Sala, Past, Present and Future of Feed-in Tariffs in Spain: What are their Real Costs?, 27th
European Photovoltaic Solar Energy Conference and Exhibition (2012) 4612–4616doi:10.4229/27thEUPVSEC2012-6CV.3.49.
URL http://www.eupvsec-proceedings.com/proceedings?paper=17736
[6] F. Gotzens, H. Heinrichs, J. Hrsch, F. Hofmann, Performing energy modelling exercises in a transparent way - The issue of data
quality in power plant databases, Energy Strategy Reviews 23 (2019) 1–12. doi:10.1016/j.esr.2018.11.004.
URL http://www.sciencedirect.com/science/article/pii/S2211467X18301056
[7] Global Warming of 1.5◦ C, Intergovernmental Panel on Climate Change (IPCC), Tech. rep. (2018).
URL https://www.ipcc.ch/sr15/
[8] M. R. Raupach, S. J. Davis, G. P. Peters, R. M. Andrew, J. G. Canadell, P. Ciais, P. Friedlingstein, F. Jotzo, D. P. Vuuren,
C. L. Quéré, Sharing a quota on cumulative carbon emissions, Nature Climate Change 4 (10) (2014) 873–879. doi:10.1038/
nclimate2384.
URL https://www.nature.com/articles/nclimate2384
[9] Open Power System Data. 2018. Data Package Time series. Version 2018-03-13. (Primary data from various sources, for a
complete list see URL).
URL https://data.open-power-system-data.org/time_series/2018-03-13/.
[10] S. Saha, S. Moorthi, X. Wu, J. Wang, S. Nadiga, P. Tripp, D. Behringer, Y.-T. Hou, H. Chuang, M. Iredell, M. Ek, J. Meng, R. Yang,
M. P. Mendez, H. van den Dool, Q. Zhang, W. Wang, M. Chen, E. Becker, NCEP climate forecast system version 2 (CFSv2)
selected hourly time-series products (2011).
URL https://doi.org/10.5065/D6N877VB
[11] G. B. Andresen, A. A. Söndergaard, M. Greiner, Validation of Danish wind time series from a new global renewable energy atlas
for energy system analysis, Energy 93 (Part 1) (2015) 1074 – 1088. doi:https://doi.org/10.1016/j.energy.2015.09.071.
URL http://www.sciencedirect.com/science/article/pii/S0360544215012815
[12] D. P. Schlachtberger, T. Brown, S. Schramm, M. Greiner, The benefits of cooperation in a highly renewable European electricity
network, Energy 134 (Supplement C) (2017) 469–481. doi:10.1016/j.energy.2017.06.004.
URL http://www.sciencedirect.com/science/article/pii/S0360544217309969
[13] M. Victoria, G. B. Andresen, Using validated reanalysis data to investigate the impact of the PV system configurations at high
penetration levels in european countries, Progress in Photovoltaics: Research and Applications 27 (7) 576–592. doi:10.1002/
pip.3126.
URL https://onlinelibrary.wiley.com/doi/full/10.1002/pip.3126
[14] M. Victoria, K. Zhu, T. Brown, G. B. Andresen, M. Greiner, The role of photovoltaics in a sustainable European energy system
under variable CO2 emissions targets, transmission capacities, and costs assumptions, Progress in Photovoltaics: Research
and Applicationsdoi:10.1002/pip.3198.
URL https://onlinelibrary.wiley.com/doi/abs/10.1002/pip.3198
[15] Corine land cover 2006, Tech. rep., EEA (2014).
URL https://www.eea.europa.eu/data-and-maps/data/clc-2006-raster-4
[16] Natura 2000 data - the European network of protected sites.
URL https://www.eea.europa.eu/data-and-maps/data/natura-8
[17] Y. Scholz, Renewable energy based electricity supply at low costs - development of the REMix model and application for Europe,
Ph.D. thesis, University of Stuttgart (2012).
URL https://elib.uni-stuttgart.de/handle/11682/2032
[18] T. Brown, D. Schlachtberger, A. Kies, S. Schramm, M. Greiner, Synergies of sector coupling and transmission reinforcement in a
cost-optimised, highly renewable European energy system, Energy 160 (2018) 720–739. doi:10.1016/j.energy.2018.06.222.
URL http://www.sciencedirect.com/science/article/pii/S036054421831288X
[19] Technology Data for Generation of Electricity and District Heating, update November 2019, Tech. rep., Danish Energy Agency
and Energinet.dk (2019).
URL https://ens.dk/en/our-services/projections-and-models/technology-data/technology-data-generation-electricity-and
[20] I. Staffell, D. Scamman, A. V. Abad, P. Balcombe, P. E. Dodds, P. Ekins, N. Shah, K. R. Ward, The role of hydrogen and fuel cells
in the global energy system, Energy & Environmental Science 12 (2) (2019) 463–491. doi:10.1039/C8EE01157E.
URL https://pubs.rsc.org/en/content/articlelanding/2019/ee/c8ee01157e
[21] D. Caglayan, N. Weber, H. U. Heinrichs, J. Linen, M. Robinius, P. A. Kukla, D. Stolten, Technical Potential of Salt Caverns for
Hydrogen Storage in Europe (Oct. 2019). doi:10.20944/preprints201910.0187.v1.
URL https://www.preprints.org/manuscript/201910.0187/v1
[22] T. Brown, P. Schierhorn, E. Tröster, T. Ackermann, Optimising the european transmission system for 77% renewable electricity
by 2030, IET Renewable Power Generation 10 (1) 3–9. doi:10.1049/iet-rpg.2015.0135.
[23] Ten-Year Network Development Plan 2016, ENTSOE.
URL https://tyndp.entsoe.eu/maps-data/
[24] Deliverable 3.1: Profile of heating and cooling demand in 2015. Data Annex. Heat Roadmap Europe.
URL www.heatroadmap.eu
[25] Population density by NUTS 3 region.
URL https://data.europa.eu
[26] I. Staffell, D. Brett, N. Brandon, A. Hawkes, A review of domestic heat pumps, Energy & Environmental Science 5 (11) 9291–
9306. doi:10.1039/C2EE22653G.
URL https://pubs.rsc.org/en/content/articlelanding/2012/ee/c2ee22653g
[27] Mapping and analyses of the current and future (2020 - 2030) heating/cooling fuel deployment (fossil/renewables).
URL https://ec.europa.eu/energy/en/studies/mapping-and-analyses-current-and-future-2020-2030-heatingcooling-fuel-deployment
[28] Euro Heat and Power.
URL https://www.euroheat.org/knowledge-hub/country-profiles/
[29] K. Zhu, M. Victoria, G. B. Andresen, M. Greiner, Impact of climatic, technical and economic uncertainties on the optimal design
of a coupled fossil-free electricity, heating and cooling system in Europe, Applied Energy 262 (2020) 114500. doi:10.1016/j.
apenergy.2020.114500.
URL http://www.sciencedirect.com/science/article/pii/S030626192030012X
[30] P. Ruiz, A. Sgobbi, W. Nijs, C. Thiel, F. Dalla, T. Kobert, B. Elbersen, G. H. Alterra, The JRC-EU-TIMES model. bioenergy
potentials for EU and neighbouring countries.
URL https://setis.ec.europa.eu/sites/default/files/reports/biomass_potentials_in_europe.pdf
[31] ENSPRESO Biomass database, JRC.
URL https://data.jrc.ec.europa.eu/dataset/74ed5a04-7d74-4807-9eab-b94774309d9f/resource/
94aca7d6-89af-4969-a74c-2c7ab4376788
[32] Wind energy database.
URL https://www.thewindpower.net/
[33] ODYSSEE database on energy effciency data & indicators, Tech. rep. (2016).
URL http://www.indicators.odyssee-mure.eu
[34] Verkehrszahlung - Stundenwerte, Tech. Rep., Bundesanstalt für Straßenwesen.
[35] The circular economy A powerful force for climate mitigation, Material Economics, Tech. rep. (2018).
URL https://www.sitra.fi/en/publications/circular-economy-powerful-force-climate-mitigation/
[36] E. Vartiainen, G. Masson, C. Breyer, D. Moser, E. R. Medina, Impact of weighted average cost of capital, capital expenditure,
and other parameters on future utility-scale PV levelised cost of electricity, Progress in Photovoltaics: Research and Applications
(2017). doi:10.1002/pip.3189.
URL https://onlinelibrary.wiley.com/doi/abs/10.1002/pip.3189
[37] E. Vartiainen, G. Masson, C. Breyer, The true competitiveness of solar PV: a European case study, Tech. rep., European
Technology and Innovation Platform for Photovoltaics (ETIP) (2017).
URL http://www.etip-pv.eu/fileadmin/Documents/ETIP_PV_Publications_2017-2018/LCOE_Report_March_2017.pdf
[38] Lazard’s Levelized Cost of Energy Analysis, version 13.0.
URL https://www.lazard.com/media/451086/lazards-levelized-cost-of-energy-version-130-vf.pdf
[39] A. Schröeder, F. Kunz, F. Meiss, R. Mendelevitch, C. von Hirschhausen, Current and prospective costs of electricity generation
until 2050, Data Documentation, DIW 68. Berlin: Deutsches Institut.
URL https://www.econstor.eu/handle/10419/80348
[40] S. Hagspiel, C. Jägemann, D. Lindenberger, T. Brown, S. Cherevatskiy, E. Tröster, Cost-optimal power system extension under
flow-based market coupling, Energy 66 (2014) 654–666. doi:10.1016/j.energy.2014.01.025.
URL http://www.sciencedirect.com/science/article/pii/S0360544214000322
[41] M. Fasihi, D. Bogdanov, C. Breyer, Long-Term Hydrocarbon Trade Options for the Maghreb Region and EuropeRenewable
Energy Based Synthetic Fuels for a Net Zero Emissions World, Sustainability 9 (2) (2017) 306. doi:10.3390/su9020306.
URL https://www.mdpi.com/2071-1050/9/2/306
[42] K. Schaber, Integration of Variable Renewable Energies in the European power system: a model-based analysis of transmission
grid extensions and energy sector coupling, Ph.D. thesis, TU München (2013).
URL https://d-nb.info/1058680781/34
[43] N. Gerhardt, A. Scholz, F. Sandau, H. H., Interaktion EE-Strom, Wärme und Verkehr. Tech. rep. Fraunhofer IWES.
URL http://www.energiesystemtechnik.iwes.fraunhofer.de/de/projekte/suche/2015/interaktion_strom_waerme_
verkehr.html
[44] C. Budischak, D. Sewell, H. Thomson, L. Mach, D. E. Veron, W. Kempton, Cost-minimized combinations of wind power, solar
power and electrochemical storage, powering the grid up to 99.9% of the time, Journal of Power Sources 225 (2013) 60–74.
doi:10.1016/j.jpowsour.2012.09.054.
URL http://www.sciencedirect.com/science/article/pii/S0378775312014759
[45] BP Statistical Review of World Energy.
URL https://www.bp.com/content/dam/bp/business-sites/en/global/corporate/pdfs/energy-economics/
statistical-review/bp-stats-review-2019-full-report.pdf
[46] Development of the specific carbon dioxide emissions of the German electricity mix in the years 1990 - 2018, German
Environment Agency.
URL https://www.umweltbundesamt.de/sites/default/files/medien/1410/publikationen/2019-04-10_cc_10-2019_
strommix_2019.pdf
[47] Word Energy Outlook 2017, International Energy Agency, Tech. rep.
[48] W. Zappa, M. Junginger, M. van den Broek, Is a 100% renewable European power system feasible by 2050?, Applied Energy
233-234 (2019) 1027–1050. doi:10.1016/j.apenergy.2018.08.109.
URL http://www.sciencedirect.com/science/article/pii/S0306261918312790
[49] Wege Zu Einem Klimaneutralen Energiesystem - Anhang zur Studie, Tech. rep. (2020).
URL https://www.ise.fraunhofer.de/content/dam/ise/de/documents/publications/studies/
Anhang-Studie-Wege-zu-einem-klimaneutralen-Energiesystem.pdf
[50] H.-J. Fell, C. Breyer, M. Metayer, The Projections for the Future and Quality in the Past of the World Energy Outlook for Solar
PV and Other Renewable Energy Technologies, 31st European Photovoltaic Solar Energy Conference and Exhibition (2015)
3220–3246doi:10.4229/EUPVSEC20152015-7DV.4.61.
URL http://www.eupvsec-proceedings.com/proceedings?paper=35067
[51] F. Creutzig, P. Agoston, J. C. Goldschmidt, G. Luderer, G. Nemet, R. C. Pietzcker, The underestimated potential of solar energy
to mitigate climate change, Nature Energy 2 (9) (2017). doi:10.1038/nenergy.2017.140.
URL https://www.nature.com/articles/nenergy2017140
[52] N. M. Haegel, H. Atwater, T. Barnes, C. Breyer, A. Burrell, Y.-M. Chiang, S. D. Wolf, B. Dimmler, D. Feldman, S. Glunz, J. C.
Goldschmidt, D. Hochschild, R. Inzunza, I. Kaizuka, B. Kroposki, S. Kurtz, S. Leu, R. Margolis, K. Matsubara, A. Metz, W. K.
Metzger, M. Morjaria, S. Niki, S. Nowak, I. M. Peters, S. Philipps, T. Reindl, A. Richter, D. Rose, K. Sakurai, R. Schlatmann,
M. Shikano, W. Sinke, R. Sinton, B. J. Stanbery, M. Topic, W. Tumas, Y. Ueda, J. v. d. Lagemaat, P. Verlinden, M. Vetter,
E. Warren, M. Werner, M. Yamaguchi, A. W. Bett, Terawatt-scale photovoltaics: Transform global energy, Science 364 (6443)
(2019) 836–838. doi:10.1126/science.aaw1845.
URL https://science.sciencemag.org/content/364/6443/836
[53] Photovoltaics Report, Fraunhofer Institute for Solar Energy Systems, version November 2019, Tech. rep. (2019).
URL https://www.ise.fraunhofer.de/content/dam/ise/de/documents/publications/studies/Photovoltaics-Report.
pdf
[54] M. Child, C. Kemfert, D. Bogdanov, C. Breyer, Flexible electricity generation, grid exchange and storage for the transition to a
100% renewable energy system in Europe, Renewable Energy 139 (2019) 80–101. doi:10.1016/j.renene.2019.02.077.
URL http://www.sciencedirect.com/science/article/pii/S0960148119302319
[55] D. Bogdanov, J. Farfan, K. Sadovskaia, A. Aghahosseini, M. Child, A. Gulagi, A. S. Oyewo, L. Barbosa, C. Breyer, Radical
transformation pathway towards sustainable electricity via evolutionary steps, Nature Communications 10 (1) (2019) 1–16. doi:
10.1038/s41467-019-08855-1.
URL https://www.nature.com/articles/s41467-019-08855-1
[56] In-depth analysis in support of the Comission Communication COM(2018) 773 A Clean Planet for all. A European long-term
strategic vision for a prosperous, modern, competitive and climate neutral economy, Tech. rep. (Nov. 2018).
URL https://ec.europa.eu/clima/news/commission-calls-climate-neutral-europe-2050_en
[57] Technology pathways in decarbonisation scenarios, Tech. rep. (2018).
URL https://ec.europa.eu/energy/sites/ener/files/documents/2018_06_27_technology_pathways_-_
finalreportmain2.pdf
[58] V. Krey, F. Guo, P. Kolp, W. Zhou, R. Schaeffer, A. Awasthy, C. Bertram, H.-S. de Boer, P. Fragkos, S. Fujimori, C. He, G. Iyer,
K. Keramidas, A. C. Kberle, K. Oshiro, L. A. Reis, B. Shoai-Tehrani, S. Vishwanathan, P. Capros, L. Drouet, J. E. Edmonds,
A. Garg, D. E. H. J. Gernaat, K. Jiang, M. Kannavou, A. Kitous, E. Kriegler, G. Luderer, R. Mathur, M. Muratori, F. Sano, D. P.
van Vuuren, Looking under the hood: A comparison of techno-economic assumptions across national and global integrated
assessment models, Energy 172 (2019) 1254–1267. doi:10.1016/j.energy.2018.12.131.
URL http://www.sciencedirect.com/science/article/pii/S0360544218325039
[59] F. Egli, B. Steffen, T. S. Schmidt, Bias in energy system models with uniform cost of capital assumption, Nature Communications
10 (1) (2019) 1–3. doi:10.1038/s41467-019-12468-z.
URL https://www.nature.com/articles/s41467-019-12468-z
[60] B. U. Schyska, A. Kies, How regional differences in cost of capital influence the optimal design of power systems, Applied Energy
262 (2020) 114523. doi:10.1016/j.apenergy.2020.114523.
URL http://www.sciencedirect.com/science/article/pii/S0306261920300350
[61] Evaluating our future. The crucial role of discount rates in European Commission energy system modelling, Ecofys, Tech. rep.
(2015).
URL https://www.eurima.org/uploads/ModuleXtender/Publications/129/Ecofys_EvaluatingOurFuture-final20151019.
pdf
[62] J. Emmerling, L. Drouet, K. van der Wijst, D. van Vuuren, V. Bosetti, M. Tavoni, The role of the discount rate for emission
 pathways and negative emissions, Environmental Research Letters 14 (10) (2019) 104008, publisher: IOP Publishing. doi:
10.1088/1748-9326/ab3cc9.
URL https://doi.org/10.1088%2F1748-9326%2Fab3cc9
[63] Renewable Energy and Jobs Annual Review 2019, IRENA, Tech. rep. (2019).
URL https://www.irena.org/publications/2019/Jun/Renewable-Energy-and-Jobs-Annual-Review-2019
[64] Low carbon jobs: The evidence for net job creation from policy support for energy efficiency and renewable energy, UKERC,
Tech. rep. (2014).
URL http://www.ukerc.ac.uk/programmes/technology-and-policy-assessment/low-carbon-jobs.html
[65] Renewable Energy: a gender perspective, Tech. rep. (2019).
URL https://www.irena.org/-/media/Files/IRENA/Agency/Publication/2019/Jan/IRENA_Gender_perspective_2019.
pdf