See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283975603

Impact of solar panels on global climate

Article in Nature Climate Change · November 2015

DOI: 10.1038/nclimate2843

CITATIONS READS

157 19,859

9 authors, including:

Aixue Hu Samuel Levis

NSF National Center for Atmospheric Research SLevis Consulting LLC
157 PUBLICATIONS 13,281 CITATIONS 126 PUBLICATIONS 25,053 CITATIONS

SEE PROFILE SEE PROFILE

Gerald A. Meehl Weiqing Han

NSF National Center for Atmospheric Research University of Colorado Boulder
345 PUBLICATIONS 100,140 CITATIONS 191 PUBLICATIONS 8,617 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Aixue Hu on 19 November 2015.

The user has requested enhancement of the downloaded file.

 LETTERS
PUBLISHED ONLINE: 2 NOVEMBER 2015 | DOI: 10.1038/NCLIMATE2843

Impact of solar panels on global climate

Aixue Hu1*, Samuel Levis1†, Gerald A. Meehl1, Weiqing Han2, Warren M. Washington1,
Keith W. Oleson1, Bas J. van Ruijven1, Mingqiong He3 and Warren G. Strand1

Regardless of the harmful effects of burning fossil fuels Four idealized simulations are carried out. The first simulation is
on global climate1,2 , other energy sources will become more a control simulation (hereafter Control) with the climatic boundary
important in the future because fossil fuels could run out condition from RCP2.6 (2006–2100; ref. 5)—the lowest emission
by the early twenty-second century3 given the present rate scenario for the Coupled Model Intercomparison Project phase 5
of consumption4 . This implies that sooner or later humanity (CMIP5). The second is the same as the Control but with solar
will rely heavily on renewable energy sources. Here we panels installed in the desert areas and in all urban regions
model the effects of an idealized large-scale application of (Supplementary Fig. 1). This simulation tests the impact of the solar
renewable energy on global and regional climate relative to power production alone on regional and global climate (hereafter
a background climate of the representative concentration SPDU). The third simulation is the same as the second, but we
pathway 2.6 scenario (RCP2.6; ref. 5). We find that solar panels further test the climate impact of consuming the power produced
alone induce regional cooling by converting incoming solar by solar panels in urban areas by hypothetically making interior
energy to electricity in comparison to the climate without solar building thermostat settings globally equal to those used in the
panels. The conversion of this electricity to heat, primarily United States in the CCSM urban module (hereafter SPDU+UH)
in urban areas, increases regional and global temperatures (see Supplementary Information)10 . The last experiment is the same
which compensate the cooling effect. However, there are as SPDU, but the solar panel installation is limited to part of
consequences involved with these processes that modulate Egypt only (Supplementary Fig. 1 green stippling region). This tests
the global atmospheric circulation, resulting in changes in the climate impact of a more realistic projection of future energy
regional precipitation. demand as outlined below (hereafter SPDLess).
Solar power is the most abundant available renewable energy For its Fifth Assessment Report, the IPCC collected a large set
source6,7 . The solar power reaching the Earth’s surface is about of global energy and climate scenarios11 . The scenarios in this
86,000 TW (1 TW = 1012 J s−1 ; refs 6,8), but the harvestable solar database comprise a wide range of different futures with respect to
power is much less than this7 . Recent estimates of achievable solar population, economic growth, energy use, technology development
power in the world range from ∼400 to 8,800 TW, given the current and availability, and climate policies. We have sampled this database
system performance, topographic limitations and environmental to identify the potential range of future solar energy demand. The
and land-use constraints7 . In 2010, the average global power con- upper bound of solar power production in these scenarios increases
sumption was about 17.5 TW (ref. 4), so harvesting a few percent of from 0.5 EJ yr−1 in 2010 (0.015 TW) to 525 EJ yr−1 in 2100 (17 TW).
the achievable solar power would provide enough energy for all hu- However, the upper bound of total primary energy use increases
mans today. Here we apply the Community Climate System Model from 523 EJ yr−1 in 2010 (17 TW) to 1,980 EJ yr−1 in 2100 (63 TW).
version 4 (CCSM4; ref. 9) to investigate how the required large- Assuming final energy use in all sectors would change from fossil
scale solar panel installations might affect the global climate. This fuels to electricity (see Supplementary Methods), having to be
was achieved through a set of idealized climate model sensitivity transported over long distances, the upper bound for solar electricity
experiments where all future energy is derived from solar power demand would be around 1,420 EJ yr−1 (45 TW) by 2100.
alone. (A climate model sensitivity experiment is a standard climate There are three main ways to convert solar power to electricity:
modelling methodology that employs an idealized large forcing in photovoltaic (PV) panels that convert light directly to electricity,
the model to produce a high-amplitude response with a significant thermophotovoltaic (TPV) panels that convert radiant heat
and unambiguous signal. The results from such sensitivity experi- differentials to electricity via photons, and concentrated solar
ments are used to provide insights into processes and mechanisms in power (CSP) using mirrors or lenses to concentrate sunlight to heat
the climate system and to help interpret responses from experiments a fluid to drive a turbine and generate power (see Supplementary
with smaller amplitude and more realistic forcings.) Information). The present efficiency of these panels ranges from
Ideally, solar panels should be installed in regions with little cloud less than 20% (PV) to over 40% (TPV and CSP; refs 12,13), and
cover to maximize electricity production. We emphasize the climate concentrated PV panels (CPV) using multi-junctions can also reach
signal, by hypothetically installing the solar panels in all the major an efficiency of ∼40% (ref. 14). However, potential solar panel
desert regions of the world (Northern Sahara desert and the desert efficiency could reach 60% (ref. 15). Here we conservatively assume
areas of Asia, North America and Australia) in our simulations this efficiency to be 30% by assuming a combination of CPV and
(Supplementary Fig. 1a). However, energy demand centres are not CSP panels, but excluding TPV panels because these panels are too
always collocated with the best locations for solar panels, so we expensive to be installed at large scale. On this basis, we assume
also test the decentralized installation of solar panels in urban areas 10% of the incident solar radiation is either reflected by the solar
around the globe (Supplementary Fig. 1b). panels, as a result of panels’ glare and glint, or lost owing to the

1 Climate and Global Dynamics Division, National Center for Atmospheric Research, Boulder, Colorado 80305, USA. 2 Department of Atmospheric and

Oceanic Sciences, University of Colorado, Boulder, Colorado 80301, USA. 3 Meteorological Bureau of Hubei Province, Wuhan, Hubei Province 430074,
China. † Present address: The Climate Corporation, San Francisco, California 94103, USA. *e-mail: [email protected]

NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange 1

© 2015 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE2843
Temperature
a Control b SPDU−Control
288.23 K −0.34 K

255 265 275 285 295 305 −1.0 −0.5 −0.1 0.1 0.5 1.0

c SPDU+UH−SPDU d SPDLess−Control
0.09 K −0.04 K

−1.0 −0.5 −0.1 0.1 0.5 1.0 −1.0 −0.5 −0.1 0.1 0.5 1.0

Figure 1 | Surface temperature. a, Surface temperature in the Control experiment. b, Surface temperature anomaly relative to the Control experiment in the
SPDU experiment. c, Surface temperature anomaly relative to the SPDU experiment in the SPDU+UH experiment. d, Surface temperature anomaly relative
to the Control experiment in the SPDLess experiment. The numbers at the upper right corner of each panel represent the global average. Stippling indicates
the changes are statistically significant at the 95% level using a double-sided Student’s t-test.

conversion from direct current to alternating current and the local by 2100, suggesting that solar power in these experiments has the
wire thermal loss before the electricity feeds into the main grid potential to satisfy human demand now and in the future. However,
(effectively this is parameterized as reflection in the model). The in these idealized sensitivity experiments, solar panels cover 100%
remainder (90%) is partitioned as 30% (of the remaining 90%) of the urban and desert regions, as shown in Supplementary
absorbed by the panels and converted to electricity, and the other Fig. 1. In reality, this coverage would be at most 40%. Thus the
70% (of the remaining 90%) transmitted through the panels and actual solar power production in our simulations would be about
absorbed by the underlying surface. Thus the effective solar panel 60% less than the numbers mentioned above (see Supplementary
efficiency in our simulations is 27% (90% × 30%). Information; ref. 16).
It takes about five years for the surface climate to reach a Climate change may affect the amount of solar radiation reaching
quasi-equilibrium state in the three sensitivity simulations (SPDU, the Earth’s surface17 . For example, reduced sea ice, snow and ice
SPDU+UH, SPDLess; Supplementary Fig. 2). Thus we analyse the sheet coverage will increase the absorption of solar radiation at
last 90 years from each of these simulations. The results discussed the surface, but the increased cloudiness induced by an enhanced
below are the 90-year means in each of the simulations and can hydrologic cycle may reflect more solar radiation. Here we find
be considered as representing conditions for the mid-twenty-first that solar panel electricity generation will redistribute the energy
century. In the following analysis, the changes of climate properties from the sun, thus affecting regional and global climates. Without
in these sensitivity simulations relative to Control and relative to the solar panels, solar radiation reaching the surface is partitioned
each other are discussed. into absorption and reflection. The transmission part of the solar
First we examine whether the solar panels in these idealized radiation is eventually either reflected or absorbed by the Earth’s
experiments could produce enough power to satisfy human surface in the annual mean, thus it is not explicitly considered
demand. Power production by solar panels is ∼740 ± 5 TW here. With the solar panels, a portion of absorbed solar radiation
(uncertainty values here and throughout the text are ±1 s.d.) in is diverted to electricity generation. In the regions with solar panels
desert regions and 48 ± 1 TW in the urban areas in both the installed, the direct shortwave radiation incident on the solar panels
SPDU and SPDU+UH simulations (Supplementary Table 1). Even increases slightly in all experiments relative to the Control owing
after the solar panel installation is scaled back in the SPDLess to a reduction of cloudiness (Supplementary Tables 1 and 2).
simulation, the power production is still about 59 ± 1 TW, roughly However, local absorption of direct shortwave radiation decreases
30% more than the upper bound of a fully solar-based energy system by up to 19% in the SPDU and SPDU+UH experiments, with an

2 NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange

© 2015 Macmillan Publishers Limited. All rights reserved
NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE2843 LETTERS
Precipitation
a Control b SPDU−Control
1.103 m yr−1 −0.006 m yr−1

0.1 0.5 1.5 2.5 3.5 4.5 −0.25 −0.15 −0.05 0.05 0.15 0.25

c SPDU+UH−SPDU d SPDLess−Control
0.001 m yr−1 −0.001 m yr−1

−0.25 −0.15 −0.05 0.05 0.15 0.25 −0.25 −0.15 −0.05 0.05 0.15 0.25

Figure 2 | Precipitation. a, Annual total precipitation in the Control experiment. b, Precipitation anomaly relative to the Control experiment in the SPDU
experiment. c, Precipitation anomaly relative to the SPDU experiment in the SPDU+UH experiment. d, Precipitation anomaly relative to the Control
experiment in the SPDLess experiment. The numbers at the upper right corner of each panel represent the global average. Stippling indicates the changes
are statistically significant at 95% level using a double-sided Student’s t-test.

increase of 4% in the SPDLess experiment (Supplementary Table 2). reduced by over 20% for the SPDU and SPDU+UH simulations
The reflected direct solar radiation is reduced by 44% in the (Supplementary Table 2) in contrast to a minor 2–4% increase in the
SPDU and SPDU+UH experiments, but by 77% in the SPDLess RCP scenarios (Supplementary Tables 4 and 6). The precipitation
experiment. Therefore, the total solar panel power production in changes in the SPDLess simulation are also large (∼20%), but statis-
the SPDU and SPDU+UH experiments is from the reduction of tically insignificant owing to large internal variability. In the urban
both reflected and absorbed direct incident solar (about 50% each) regions, solar panels induce a moderate cooling of about −0.26 ◦ C
in comparison to the Control. In the SPDLess experiment, this in the SPDU experiment, agreeing with previous studies18–20 .
power production is entirely from reduced reflection, because the The above local cooling in desert regions generates significant
absorption is slightly increased. climate responses in remote areas (Fig. 1b). In contrast to a projected
In general, the changes in the reflected solar radiation do not warming almost everywhere from the CMIP5 future climate change
directly affect the regional and global climate, but the changes in experiments21 (see Supplementary Fig. 5 for CCSM4), hemispheric-
absorbed solar radiation do. Reduced absorption of solar radiation scale cooling of up to 1 ◦ C relative to the Control in the SPDU
leads to a significant local cooling by more than −2 ◦ C relative to simulation occurs to the east of the solar panels or downstream from
Control averaged in the desert regions with installed solar panels the prevailing westerly winds, except in a few places, such as India
in the SPDU and SPDU+UH experiments (Fig. 1 and Supplemen- and the west coast of North America, which actually warm by up to
tary Tables 1 and 2). In contrast, the temperature in these regions 1 ◦ C. In the Southern Hemisphere, the cooling effect is considerably
is projected to increase by 1∼2.5 ◦ C in the four RCP scenarios less owing to a much smaller area with solar panels installed. Along
(RCP2.6, RCP4.5, RCP6.0 and RCP8.5) in CCSM4. Projected global with the temperature changes, the global precipitation pattern also
and regional temperature changes in CCSM4 are comparable to changes significantly (Fig. 2b). The annual mean precipitation is
the multi-model ensemble temperature changes for CMIP5 mod- reduced by up to 0.25 m yr−1 in Central, East and Southeast Asia,
els (Supplementary Figs 3 and 4 and Supplementary Tables 4–6). parts of the Middle East, Australia and part of the tropical Pacific,
Therefore, hereafter, we compare the results in the sensitivity simu- increasing by up to 0.25 m yr−1 in the SPDU experiment in Europe,
lations only with the RCP scenarios using CCSM4. In SPDLess, a the North Pacific, western North America, tropical Africa, and the
slight increase of absorbed solar radiation induces little warming southeast Indian Ocean. These precipitation changes are, in general,
(Supplementary Table 1). Precipitation in these desert regions is opposite to the CMIP5 projections21 (Supplementary Fig. 5).

NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange 3

© 2015 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE2843

The warming in India and eastern Australia is associated with a in all experiments, but that more realistic ranges of solar power
precipitation feedback on land. Precipitation is reduced in these two production would result in a negligible impact on global mean
regions when the solar panels are installed (Fig. 2b). This reduces temperature. In fact, our sensitivity experiments with solar panels in
the evaporative cooling on land (Supplementary Fig. 6a) and leads deserts only and solar panels in both desert and urban areas show a
to warming. In response to the reduced rainfall, plant leaf area smaller impact on temperature and precipitation for the latter than
declines in these regions (Supplementary Fig. 6c), contributing to for the former, suggesting a more distributed solar panel installation
a reduced transpiration cooling effect, which also leads to warming. could reduce the impact of the solar panels on regional and global
The warming in northwest North America is caused by enhanced climate (see Supplementary Information and Supplementary Fig. 9).
southerly flow related to the deepening of the Aleutian Low (Supple- By using RCP2.6 as the background reference climate, our results
mentary Fig. 6d). This southerly flow brings warmer ocean air into suggest that idealized massive-scale installations of solar panels
northwest North America, leading to higher temperatures there. can generate enough power for human usage now and in the
The cooling shown in Fig. 1b also induces robust changes future, although there are consequences that involve impacts on the
in atmospheric circulation. Corresponding to the surface climate system. Such a large number of solar panels redistributes the
cooling, there are positive surface pressure anomalies which incoming solar radiation and changes the local radiation balance,
produce a divergent flow that drives cold air to surrounding resulting in changes in atmospheric circulation, thus affecting
regions (Supplementary Fig. 6d). At 500 hPa, there are five regional and global climate. Overall, regardless of its capacity (as
negative geopotential height anomaly centres and cyclonic wind large as ∼800 TW or a more realistic projection of 45 TW), the
anomalies in mid-latitudes, located over northwest Africa, the potential global mean climate changes induced by the use of solar
Middle East, eastern China, North Pacific and North America panels are small in comparison to the expected climate change
(Supplementary Fig. 7b). These changes in wind patterns signify owing to fossil fuel consumption, which could raise the global mean
a strengthening of the jet stream south of 45◦ N and a weakening temperature by a few degrees by 2100 relative to pre-industrial
north of 45◦ N, leading to an equatorward shift of the jet stream climate21 . However, some of the regional climate changes induced
(Supplementary Fig. 8b). In the Southern Hemisphere, the changes by solar panels could be much greater than the global mean.
of the jet stream are less significant in comparison to those in the
Northern Hemisphere owing to the smaller surface temperature Received 5 May 2015; accepted 27 September 2015;
change. These changes of the jet location are consistent with the published online 2 November 2015
thermal wind relationship. The larger cooling in the mid to high
latitudes increases the temperature contrast between low and mid References
to high latitudes, leading to a strengthening of the jet. 1. IPCC in Climate Change 2014: Impacts, Adaptation, and Vulnerability
An earlier study22 suggests that heat released to the environment (eds Field, C. B. et al.) (Cambridge Univ. Press, 2014).
due to human energy usage will not change the global mean surface 2. IPCC in Climate Change 2013: The Physical Science Basis
temperature much, but could generate regional temperature changes (eds Stocker, T. F. et al.) (Cambridge Univ. Press, 2013).
3. Shafiee, S. & Topal, E. When will fossil fuel reserves be diminished? Energy
of up to 1 ◦ C with waste heat added to the system at the rate of Policy 37, 181–189 (2009).
about 42% (6.7 TW) of the energy consumption in 2006. In the 4. Key World Energy Statistics (International Energy Agency, 2013).
SPDU+UH experiment, about 110 TW of heat is added to the 5. Van Vuuren, D. P. et al. RCP2.6: Exploring the possibility to keep global mean
climate system as waste heat in urban regions. As a result, global temperature increase below 2 ◦ C. Clim. Dynam. 109, 95–116 (2011).
mean temperature rises by 0.09 ± 0.12 ◦ C, about nine times more 6. International Energy Outlook 2013 DOE/EIA-0484 (US Energy Information
than in the earlier study22 . If we linearly scale up this temperature Administration, US Department of Energy, 2013).
7. Rogner, H.-H. et al. Global Energy Assessment-Toward a Sustainable Future
change to consume all the power produced in the SPDU+UH Ch. 7, 423–512 (Cambridge Univ. Press and the International Institute for
experiment, the global mean temperature could rise by 0.63 ◦ C, Applied Systems Analysis, 2012).
which will not only compensate the cooling induced by solar panel 8. Kamat, P. V. Meeting the clean energy demand: Nanostructure architectures for
power production, but also lead to a few tenths of a degree warming solar energy conversion. J. Phys. Chem. 111, 2834–2860 (2007).
relative to the Control (0.63 − 0.25 = 0.38 ◦ C). However, with a 9. Gent, P. R. et al. The Community Climate System Model version 4. J. Clim. 24,
power consumption of ∼800 TW (about 45 times higher than 4973–4991 (2011).
10. Oleson, K. W., Bonan, G. B., Feddema, J. & Jackson, T. An examination of
current global energy consumption), the resulting slight warming urban heat island characteristics in a global climate model. Int. J. Clim. 31,
is much less than the greenhouse-gas-induced surface temperature 1848–1865 (2010).
increase in any RCP scenario (for example, Supplementary Fig. 5). In 11. Clarke, L. et al. in Climate Change 2014: Mitigation of Climate Change
the urban areas, the power consumption induces a mean surface air (eds Edenhofer, O. et al.) 413–510 (IPCC, Cambridge Univ. Press, 2014).
temperature increase of 1.1 ± 0.2 ◦ C in the SPDU+UH experiment 12. Rephaeli, E. & Fan, S. Absorber and emitter for solar thermo-phtovoltaic
relative to the SPDU experiment. Globally, as shown in Fig. 1c, systems to achieve efficiency exceeding the Shockley–Queisser limit. Opt.
Express 17, 15145–15159 (2009).
warming occurs almost everywhere due to this energy usage, with 13. Bermel, P. et al. Design and global optimization of high-efficiency
a warming of up to a few degrees in India, equatorial Africa and thermophotovoltaic systems. Opt. Express 18, A314–A334 (2010).
Midwestern America. Consistent with ref. 22, this surface warming 14. Chukwula, C. & Folly, K. A. Overview of concentrated photovoltaic (CPV)
generates opposite changes of the subtropical jet stream to those in cells. J. Power Energy Eng. 2, 1–8 (2014).
the SPDU experiment (see Supplementary Fig. 7c). 15. Lenert, A. et al. A nanophotonic solar thermophotovoltaic device. Nature
Because the total energy production is far beyond the expected Nanotech. 9, 126–130 (2014).
16. Kurokawa, K. et al. (eds) Energy from the Desert: Very Large Scale Photovoltaic
maximum usage (45 TW by 2100) in the SPDU and SPDU+UH System (Earthscan, 2007).
simulations, we use the SPDLess experiment to further evaluate 17. Meehl et al. in Climate Change 2007: The Physical Science Basis
the impact of solar panels on regional and global climate. The (eds Solomon, S. et al.) 747–845 (IPCC, Cambridge Univ. Press, 2007).
solar panel installed area in the SPDLess experiment is only about 18. Millstein, D. & Menon, S. Regional climate consequences of large scale cool
10% of that in the SPDU experiment (Supplementary Fig. 1). The roof and photovoltaic array deployment. Environ. Res. Lett. 6, 034001 (2011).
energy production is about 59 ± 1 TW, with an associated global 19. Taha, H. The potential for air-temperature impact from large scale deployment
of solar photovoltaic arrays in urban areas. Sol. Energy 91, 358–367 (2013).
cooling of 0.04 ± 0.13 ◦ C. The pattern of surface temperature 20. Masson, V. et al. Solar panels reduce both global warming and urban heat
change is similar to that in the SPDU experiment, but with a much island. Front. Environ. Sci. 2, 14 (2014).
reduced magnitude (Fig. 1d). This suggests that the impact of solar 21. Collins, M. et al. in Climate Change 2013: The Physical Science Basis
panels on regional and global climate is qualitatively the same (eds Stocker, T. F. et al.) 1029–1136 (Cambridge Univ. Press, 2013).

4 NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange

© 2015 Macmillan Publishers Limited. All rights reserved
NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE2843 LETTERS
22. Zhang, G. J., Cai, M. & Hu, A. Energy consumption and the unexplained winter Author contributions
warming over northern Asia and North America. Nature Clim. Change 3, A.H. designed and led the study. A.H., S.L., G.A.M., W.H., W.M.W., K.W.O., B.J.v.R.,
466–470 (2013). M.H. and W.G.S. contributed to the model simulations, data analysis, and all authors
actively contributed towards writing the manuscript.
Acknowledgements
A portion of this study was supported by the Regional and Global Climate Modelling Additional information
Program (RGCM) of the US Department of Energy’s Office of Science (BER), Supplementary information is available in the online version of the paper. Reprints and
Cooperative Agreement No. DE-FC02-97ER62402. This research used computing permissions information is available online at www.nature.com/reprints.
resources of the Climate Simulation Laboratory at the National Center for Atmospheric Correspondence and requests for materials should be addressed to A.H.
Research (NCAR), which is sponsored by the National Science Foundation; the Oak
Ridge Leadership Computing Facility, which is supported by the Office of Science of the
US Department of Energy under Contract DE-AC05-00OR22725. The National Center Competing financial interests
for Atmospheric Research is sponsored by the National Science Foundation. The authors declare no competing financial interests.

NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange 5

© 2015 Macmillan Publishers Limited. All rights reserved
 SUPPLEMENTARY INFORMATION
DOI: 10.1038/NCLIMATE2843

Impact of solar panels on global climate

1 Impact of Solar Panels on global climate

2 Aixue Hu1*, Samuel Levis1,2, Gerald A. Meehl1, Weiqing Han3, Warren M. Washington1, Keith

3 W. Oleson1, Bas J. van Ruijven1, Mingqiong He4, Warren G. Strand1

1
4 Climate and Global Dynamics Division, National Center for Atmospheric Research, Boulder,

5 CO 80305, USA

2
6 Now at the Climate Corporation, San Francisco, California

3
7 Department of Atmospheric and Oceanic Sciences, University of Colorado, Boulder, CO 80301,

8 USA

4
9 Meterological Bureau of Hubei Province, Wuhan, Hubei Province, 430074, China

10 *Correspondence to: [email protected]

NATURE CLIMATE CHANGE | www.nature.com/natureclimatechange 1

© 2015 Macmillan Publishers Limited. All rights reserved
11 Supplementary material:

12 CCSM4 and its urban model:

13 The National Center for Atmospheric Research (NCAR) Community Climate System

14 Model version 4 (CCSM4) is a fully coupled climate model which is developed under the

15 collaboration among NCAR scientists, Scientists from US Department of Energy Laboratories,

16 and university scientists. The version of the model used here is the default version which

17 contains the Community Atmospheric Model version 4 (CAM4) with 26 vertical levels and a

18 horizontal resolution of 1 degree, the Parallel Ocean Program (POP) version 2 with 60 levels

19 vertically and nominal 1 degree resolution horizontally, the Community Land Model version 4

20 (CLM4), and the Community Ice Code version 4 (CICE4)1. The equilibrium climate sensitivity

21 to a CO2 doubling is 3.2oC and the transient climate response from a 1% CO2 simulation around

22 the time of CO2 doubling is 1.73oC for CCSM4, lying near the mid-way values among the

23 CMIP5 models2.

24 The land component of CCSM4 is the Community Land Model (CLM4)3,4. Included in

25 the CLM4 is an urban canyon parameterization that is modeled as a separate land unit within

26 each model grid cell [Community Land Model Urban (CLMU)]. A full technical description of

27 CLMU can be found in Oleson et al.5. Here a brief description of the parameterizations for

28 anthropogenic heat flux (urban space heating, air conditioning, and waste heat) is provided.

29 The urban land unit has the following five components denoted as columns in the CLM4

30 subgrid framework: roof, sunlit wall, shaded wall, and pervious (e.g., to represent lawns and

31 parks) and impervious (e.g., to represent roads, parking lots, sidewalks) canyon floor. Each

32 column is divided into 15 below-ground layers for temperature calculations. A one-dimensional

33 heat conduction equation is solved numerically for each column to determine conduction fluxes

© 2015 Macmillan Publishers Limited. All rights reserved

34 into and out of each surface. The lower (internal) boundary conditions for roofs and walls are

35 determined by an approximation of internal building temperature held between maximum and

36 minimum temperatures as prescribed by the urban properties dataset6. The amount of energy

37 required to be added to bring the interior building temperature up to the minimum temperature

38 and the amount of energy required to be removed from the building interior to reduce the interior

39 building temperature to the maximum temperature are proxies for the space heating and air

40 conditioning fluxes, respectively. The heat removed by air conditioning is added as waste heat

41 (sensible heat) to the canyon floor, in proportion to pervious and impervious surface fraction.

42 Waste heat from inefficiencies in the heating and air conditioning equipment and from energy

43 lost in the conversion of primary energy sources to end use energy is also added as sensible heat

44 to the canyon floor5. The total amount of anthropogenic heat flux added to the climate system is

45 the sum of the energy due to the nonzero internal boundary condition for roofs and walls, the air

46 conditioning flux, and the waste heat7. This energy is distributed in urban areas and depends on

47 the local urban climate simulated in the model.

48 It is worth noting that the current CCSM urban model takes into account only energy

49 consumed for building space heating and cooling which is about 1/5 of the total current energy

50 consumption. The other two major sectors, transportation and factories, use roughly the other 4/5

51 of the total energy consumption which is not included in the CCSM simulation.

52 In the SPDU+UH simulation, we have set the living standard in the whole world to the

53 same as in the US. For example, air conditioning is used world-wide the same as in the US. The

54 purpose for this assumption is not meant to make the assumption realistic, but to consume as

55 much power as possible for our climate sensitivity analysis. By applying this assumption, heat

56 removed by air conditioning increases from only 0.14±0.01 TW in the Control to 27±1.4 TW in

© 2015 Macmillan Publishers Limited. All rights reserved

57 the SPDU+UH run, and the total waste heat increases from about 6±0.3 TW to 104±5 TW.

58 Overall, the world-wide power consumption in our model simulations increases dramatically

59 from 5.4±0.5 TW in the Control to 109±5 TW in the SPDU+UH which is equivalent to a change

60 of the global mean radiative forcing from 0.01±0.001 W/m2 to 0.21±0.01 W/m2. In response to

61 this increased energy consumption, the global mean temperature rises by 0.09±0.12oC with

62 global mean urban temperature increasing by 1.1±0.2oC in comparison to SPDU (Table S1-S6).

63 Overall, one can clearly see that the impact of the energy consumption itself. The release of

64 waste heat into the environment does not affect the global mean temperature much, but the

65 greenhouse gases produced by burning of the fossil fuels can induce much more significant

66 global mean temperature changes.

67 Choice of background climate forcing and their potential impact to our conclusions:

68 In our simulations, we choose the representative concentration pathway (RCP) 2.6 as our

69 background climate forcing. This is a future greenhouse gas emission scenario which limits the

70 global mean surface temperature change by 2100 to be less than 2oC higher than the pre-

71 industrial level. In this scenario, it assumes a 70% reduction of the greenhouse gas emission from

72 2010 to 2100. We define this climate scenario as our control climate. All of our sensitivity

73 simulations with solar panels are compared with this control simulation. In other words, we focus

74 on the anomalies in the sensitivity simulations relative to the CONTROL, or the changes, not the

75 absolute values. Therefore, our conclusions shown in this manuscript do not depend on the

76 choices of the climate background because we discuss here the potential impact of the solar

77 panels on regional and global climate against a background climate. If different climate

78 backgrounds were chosen, the absolute changes induced by solar panels on regional and global

79 climate may be different, but the overall impact will remain the same.

© 2015 Macmillan Publishers Limited. All rights reserved

 80 Limitations and constraints of solar panel installation and production:

81 In our simulations the solar panels cover 100% of the urban and desert regions as shown in Fig.

82 S1. This large coverage of course would not be feasible in the real world, but these are designed

83 as sensitivity experiments to provide a large forcing so that the climate system response can be

84 unambiguously detected. Such experiments are standard practice in the field of climate

85 modeling. The idea behind such sensitivity experiments is that smaller forcing (i.e. smaller areas

86 covered by solar panels) would produce a similar but smaller amplitude climate signal. In reality

87 solar panels can only cover a small portion of the urban area if the panels are only installed on

88 rooftops. Roof area occupies about 42% of total urban areas averaged globally in CCSM4. This

89 will reduce the urban energy production from 48 TW to about 20 TW which will still provide

90 enough energy for the short term. Of course, it is not possible that all roofs are suitable for solar

91 panel installation8. If solar panels are installed only on 50% of the roofs, the energy production in

92 urban areas reduces to about 10 TW, making it necessary to install solar panels outside of the

93 urban regions as well. Solar panel installation in the desert areas cannot be 100% either. Spacing

94 is needed between rows/strings of the panels in order to avoid shading and to maximize the solar

95 panel production. To properly maintain the panels, access roads are also needed. Corridors for

96 wild-life access and habitat preservation also need to be considered for large scale panel

97 installation. Given these limitations, solar panels in desert areas normally would only cover

98 about 40% or less of the land surface9. Therefore, the actual solar power production in our model

99 would be reduced by about 60%, i.e. about 296 TW for desert areas and 10 TW for urban areas

100 in SPDU and SPDU+UH experiments, and only about 24 TW for SPDLess experiment. By

101 comparing the three sensitivity simulations, we can expect that if a more realistic deployment of

102 the solar panels is used, such as reduced percentage coverage of solar panels in a certain area,

© 2015 Macmillan Publishers Limited. All rights reserved

103 this not only will lead to reduced total production of solar power, but also result in reduced

104 impacts of the solar panels on local radiation balance, thus producing less of an influence on the

105 regional and global climate. For example, the changes of radiation budget in solar panel areas as

106 shown in Table S1 could reduce by 60%, and the corresponding temperature changes in these

107 areas would be reduced as well.

108 Currently we assume that the future solar panel installation is mainly in desert areas. The lack of

109 vegetation in these areas makes the impact of solar panel installation on evapotranspiration and

110 its feedback on local precipitation be very small. However, when solar panels are installed in

111 urban or other areas with dense vegetation, there is a potential that the ways on how the solar

112 panels are installed could influence the local evapotranspiration and its feedback on precipitation.

113 To assess these potential impacts, a specific model module, which is capable to simulation the

114 detailed processes of the solar panel-environment interactions10 (such as explicitly modeling the

115 mass, momentum, and energy balances of a large solar farm to more realistically represent these

116 processes), is needed which is beyond the scope of current model simulations, but is planned for

117 future model development work.

118 A comparison of the sensitivity simulations with CCSM4 and CMIP5 model ensemble

119 As shown in Table S4-6, the surface temperature changes in CCSM4, in general, agree with the

120 CMIP5 multi-model ensemble mean changes. Therefore, we only compare our sensitivity

121 simulations with the ensemble mean of the RCP scenarios using CCSM4. The global mean

122 temperature changes in the RCP scenarios are higher than that in any of our sensitivity

123 simulations, indicating that the greenhouse gas induced climate change signal is much larger than

124 the global and regional climate change induced by solar panel installation and consumption of

© 2015 Macmillan Publishers Limited. All rights reserved

125 the power produced by solar panels. It becomes even clearer if we compare the regional

126 temperature and precipitation change patterns in Figures 1 and 2 with Figure S5. In Figure S5, all

127 RCP scenarios show a larger warming almost everywhere in comparison to the RCP2.6 than the

128 warming shown in Figure 1. Although the magnitude of the precipitation changes in the RCP

129 scenarios is comparable to the sensitivity simulations, the regional patterns are quite different.

130 For the desert regions with solar panels installed, the reduction of precipitation is much larger in

131 the sensitivity experiments than that in the RCP scenarios using CCSM4 (Table S6). In the RCP

132 scenarios, the precipitation changes in the desert regions are mostly insignificant, but they are

133 significant in SPDU and SPDU+UH experiments.

134 Future energy demands:

135 To gain some insight into plausible future ranges of demand for solar energy, we have analyzed

136 the IPCC WG3 AR5 scenarios database11,12. This database includes 1184 scenarios from the

137 peer-reviewed literature, generated by 31 different models. We have used this scenario database

138 because it represents the current state of the science on future scenarios for energy use and

139 emissions. The scenarios in this database comprise a wide range of different futures with respect

140 to population growth, economic growth, energy use, technology development and availability,

141 and climate policies. About 95% of the scenarios in the database were developed as part of nine

142 model comparison exercises: ADAM (Adaptation and Mitigation Strategies— Supporting

143 European Climate Policy)13; AME (Asian Modeling Exercise)14; AMPERE (Assessment of

144 Climate Change Mitigation Pathways and Evaluation of the Robustness of Mitigation Cost

145 Estimates)15,16; EMF 22 (Energy Modeling Forum 22)17; EMF 27 (Energy Modeling Forum

146 27)18-20; LIMITS (Low Climate Impact Scenarios and the Implications of required tight

147 emissions control strategies)21,22; POeM (Policy Options to engage Emerging Asian economies

© 2015 Macmillan Publishers Limited. All rights reserved

148 in a post‐Kyoto regime)23,24; RECIPE (Report on Energy and Climate Policy in Europe)25;

149 RoSE (Roadmaps towards Sustainable Energy futures)26-30. Other scenarios in this database

150 originate from the Global Energy Assessment and explore a range of possible future

151 development pathways that meet policy goals on climate change, with energy access and air

152 quality31-33.

153 From this database of scenarios, we extracted two key indicators to assess plausible future ranges

154 of solar energy production (see Table S7): solar electricity production and total primary energy

155 use. We have extracted these values at the global level, for two subsets of scenarios: 1) all

156 scenarios and 2) low emission scenarios (i.e. scenarios that stabilize GHG emissions at 450 or

157 550 ppm CO2-eq, or 2.6 or 3.7 W/m2 increase in radiative forcing).

158 The first indicator (solar electricity production) directly indicates what the global energy models

159 in the database deem a plausible range for solar power in scenarios where it competes with other

160 energy technologies such as wind power, hydropower, nuclear, or carbon capture and storage.

161 The highest value for solar energy production is 525 EJ/yr (or 17TW) by 2100, which comes

162 from the MESSAGE model scenario from the Global Energy Assessment, in which emissions

163 stabilize at 450 ppm while most of the mitigation takes place on the supply side of the energy

164 system and nuclear energy is phased out. This scenario implies a total solar electricity production

165 by the end of the century that is comparable to the scale of the present day total global energy

166 system. The maximum value for 2050 (131 EJ or 4 TW) comes from a comparable scenario (low

167 climate stabilization with limited technology availability) with the GCAM model.

168 The second indicator, total primary energy, provides insight into the total scale of the energy

169 system. The maximum values for this indicator by 2050 and 2100 are from the IMACLIM model,

© 2015 Macmillan Publishers Limited. All rights reserved

170 for a baseline scenario where nuclear energy is phased out, and the total energy system increases

171 from 523 EJ (or 17TW) currently to 1980 EJ (or 63TW) by the end of the century. In scenarios

172 with limited greenhouse gas emissions, the scale of the energy system is expected to be

173 considerably smaller (max 1421 EJ, 45TW) due to the implementation of energy efficiency

174 measures that reduce the demand for energy.

175 Besides these indicators that are directly provided by the global energy models, we have

176 calculated a third indicator that quantifies what the demand for solar electricity would be if it

177 were to supply all global final energy use. Because the IPCC database does not provide sectorial

178 detail on final energy use in the scenarios, calculating this indicator is based on several

179 assumptions. This indicator is supposed to give a rough idea of the ultimate maximum scale of

180 demand for solar electricity.

181 First, a large portion of final energy use that is currently provided by fuels (such as natural gas,

182 oil, coal or biofuels) is expected to shift to electricity over the century in many of the scenarios.

183 For this, we rely on the global energy models behind the IPCC database and directly use the

184 variable “final energy – electricity”.

185 Second, in virtually all economic sectors the remaining use of fuels can theoretically be

186 substituted by electricity or hydrogen (which can be produced from solar electricity through

187 electrolysis). However, for most energy services, there are large differences in efficiency

188 between fuels or electricity. Therefore, we make several assumptions for each sector.

189 In the transport sector, the ‘tank-to-wheel’ efficiency of an internal combustion engine to convert

190 fuel into movement is currently around 25% (and could be expected to increase to 35% over
34,35
191 time), while the efficiency of battery electric vehicles is around 90% . Not all energy

© 2015 Macmillan Publishers Limited. All rights reserved

192 functions in the transport sector can theoretically be supplied by battery powered vehicles, but

193 hydrogen-powered fuel cells can ultimately supply energy for larger equipment, such as trucks,

194 ships or aircraft36,37. The chain efficiency of such a hydrogen system (i.e. from solar power to
38-41
195 hydrogen to movement) is assessed to be around 65% . Hence, assuming a 50% share for

196 both battery-electric and hydrogen powered vehicles, only 45% of the final energy fuel use

197 would be needed in the transport sector if solar electricity had to provide all primary energy

198 production.

199 In the buildings sector, fuels are mainly used for space heating and cooling. These energy

200 functions have an efficiency of around 90%, which we assumed to be similar whether fuels or

201 electricity is used42,43. Hence, if electricity had to provide all final energy in the buildings sector,

202 100% of the current fuel use would be needed.

203 In the industry sector, we assumed that most fuels are used for high temperature processes in the

204 heavy industry (steel production, cement production) at an efficiency of 90%. These high

205 temperatures could also be delivered by hydrogen produced from solar electricity44. However,

206 since electrolytic hydrogen production has an efficiency of 80% (and we assume that the final

207 application of hydrogen would be as efficient as other fuels), a shift to solar-electricity would

208 lead to an increase of energy use for the industry sector, requiring 112.5% of fuel use if it were to

209 be replaced by solar-electricity based alternatives.

210 Since no sector-level information is provided in the IPCC database, we have to make an

211 assumption about the shares of these sectors in final energy by the end of the century. Currently,

212 the shares of transport, building and industry in final energy use are roughly equal, around 30-35%

213 each, and we have assumed that this remains the case by the end of the century. Averaging the

© 2015 Macmillan Publishers Limited. All rights reserved

214 changes in final energy use between these three sectors leads to a total efficiency gain of 14%.

215 Hence, if all remaining final energy use in fuels were to be replaced by solar-electricity based

216 technologies, only 86% of the final energy use of fuels would be needed as solar –electricity

217 demand. This is mostly due to efficiency gains in the transport sector, where electricity and

218 hydrogen are more efficiently applied than fuels.

219 Based on these assumptions, we derived that the scale of electricity demand in a fully solar-

220 electricity powered energy system by the end of the 21st century would be 789-1138 EJ/yr (or

221 25-36TW) depending on the level of final energy use. This paper looks into centrally produced

222 solar-power, which has to be transported over long distances. Generally, transmission losses are
45,46
223 assumed to be around 10-20% . Taking the upper end of this range, we conclude that the

224 maximum plausible range of a fully solar-powered energy system by the end of the 21st century

225 would be around 1420 EJ/yr or 45TW.

226 Solar panels:

227 Based on the ways electricity is generated from solar power, there are three major types of solar

228 panels, namely photovoltaic (PV), thermophotovoltaic (TPV), and concentrated solar power

229 (CSP). The PV panels convert sunlight directly into electricity. When sunlight is absorbed by PV

230 panels, the solar energy knocks electrons loose in these PV panels, thus electricity can flow. A

231 TPV system converts radiant heat differentials directly into electricity via photons. This system

232 includes a thermal emitter and a PV diode cell. CSP is a system using mirrors and lenses to focus

233 the sunlight to a small area and converting this focused sunlight to heat, then this heat drives a

234 steam turbine to generate electricity. The concentrated PV (CPV) system uses lenses and curved

235 mirrors to focus sunlight onto small, multi-junction solar cells to improve the efficiency of the

© 2015 Macmillan Publishers Limited. All rights reserved

236 PV panels. Except the conventional PV system, CPV, TPV and CSP systems all can reach an

237 efficiency of 40% or above and are suitable for large scale installation47-51.

238 Sensitivity simulations:

239 To investigate the impact of the more centralized solar panel installation in desert areas versus

240 the more distributed solar panel installation in urban areas, two additional simulations are carried

241 out under the same assumption as SPDU. The area for the desert where the solar panels are

242 installed is shown in Figure S10 which is about 100% more area than that in SPDU simulation.

243 The urban area is the same as in bottom panel of Figure S1. The total area with solar panels is the

244 same in these two simulations. In other words, the shaded area of bottom panel in Figure S10

245 plus the shaded area in Figure S1 bottom panel is equal to the shaded area of top panel in Figure

246 S10. These two experiments are named “solar panels in large desert areas” (SPDL) and “solar

247 panels in large desert areas and urban” (SPDUL). Figure S9a shows that since the area where

248 solar panels are installed is larger in SPDL (Figure S10) than SPDU (Figure 1), the regional and

249 global cooling effect is also larger in SPDL (global mean -0.52±0.15oC) than in SPDU (-

250 0.34±0.12oC). However, the patterns of the surface temperature change in these two experiments

251 are almost the same.

252 On the other hand, although the total area where solar panels are installed is exactly the same in

253 SPDL and SPDUL, the global mean cooling effect in SPDUL reduces by 0.04±0.13oC (Figure

254 S9a, S9c). There are two reasons for this: 1. The impact of more distributed installation of solar

255 panels would reduce the overall impact on regional and global climate; 2. Many cities are located

256 at higher latitudes where less solar radiation reaches the surface, leading to a reduction of solar

257 energy production by 5%. This also reduces the impact on local and global precipitation (Figure

258 S9b and S9d), thus resulting in an overall reduced climate impact.

© 2015 Macmillan Publishers Limited. All rights reserved

259 References

260 1. Gent, P. R., et al. The Community Climate System Model version 4, J. Climate, 24,

261 4973–4991 (2011).

262 2. Meehl, G. A. et al. Climate system response to external forcings and climate change

263 projections in CCSM4, J. Climate, 25, 3661-3683 (2012).

264 3. Lawrence, D. M. et al. Parameterization Improvements and Functional and Structural

265 Advances in Version 4 of the Community Land Model, J Adv Model Earth Sys, 3, Artn

266 M03001, doi:10.1029/2011ms000045 (2011).

267 4. Oleson, K. W. et al. Technical description of version 4.0 of the Community Land Model

268 (CLM), NCAR Technical Note NCAR/TN-478+STR, 257 pp (2010).

269 5. Oleson, K. W., Bonan, G. B., Feddema, J. J., Vertenstein, M., and Kluzek, E.: Technical

270 description of an urban parameterization for the Community Land Model (CLMU),

271 NCAR Technical Note NCAR/TN-480+STR, 169 pp (2010).

272 6. Jackson, T. L., Feddema, J. J., Oleson, K. W., Bonan, G. B., and Bauer, J. T.

273 Parameterization of Urban Characteristics for Global Climate Modeling, Ann Assoc Am

274 Geogr, 100, 848-865 (2010).

275 7. Oleson, K.W., Bonan, G.B., Feddema, J., and Jackson, T. An examination of urban heat

276 island characteristics in a global climate model, Int. J. Clim., 31, 1848-1865,

277 doi:10.1002/joc.2201 (2010).

278 8. European Photovoltaic Industry association Report. Solar Generation Solar generation

279 6[R]. EPIA, 1-200pp (2011).

280 9. Energy from the Desert: Very Large Scale Photovoltaic system. Ed. K. Kurokawa et al.,

281 published by Earthscan in the UK and USA (2007).

© 2015 Macmillan Publishers Limited. All rights reserved

282 10. Fthenakis V. and Yu Y., Analysis of the Potential for a Heat Island Effect in Large Solar

283 Farms, Proceedings 39th IEEE Photovoltaic Specialists Conference, June 17-21, 2013,

284 Tampa, FL; pp. 3362 -3366 (2013).

285 11. Clarke, L. et al. Assessing Transformation Pathways. In: Edenhofer, O. et al. (Eds),

286 Climate Change 2014: Mitigation of Climate Change. Contribution of Working Group III

287 to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.

288 Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA

289 (2014).

290 12. Krey, V. et al. Annex II: Metrics & Methodology. In: Edenhofer, O. et al. (Eds), Climate

291 Change 2014: Mitigation of Climate Change. Contribution of Working Group III to the

292 Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge

293 University Press, Cambridge, United Kingdom and New York, NY, USA (2014).

294 13. Edenhofer, O. et al. The Economics of Low Stabilization: Model Comparison of

295 Mitigation Strategies and Costs. The Energy Journal, 31, 11-48 (2010).

296 14. Calvin, K. et al. The role of Asia in mitigating climate change: Results from the Asia

297 modeling exercise. Energy Economics, 34, Supplement 3, S251-S260 (2012).

298 15. Kriegler, E. et al. Making or breaking climate targets: The AMPERE study on staged

299 accession scenarios for climate policy. Technological Forecasting and Social Change, 90,

300 Part A, 24-44 (2015).

301 16. Riahi, K. et al. Locked into Copenhagen pledges — Implications of short-term emission

302 targets for the cost and feasibility of long-term climate goals. Technological Forecasting

303 and Social Change, 90, Part A, 8-23 (2015).

© 2015 Macmillan Publishers Limited. All rights reserved

304 17. Clarke, L., Edmonds, J., Krey, V., Richels, R., Rose, S., Tavoni, M. International climate

305 policy architectures: Overview of the EMF 22 International Scenarios. Energy

306 Economics, 31, S64-S81 (2009).

307 18. Blanford, G., Kriegler, E., Tavoni, M. Harmonization vs. fragmentation: overview of

308 climate policy scenarios in EMF27. Climatic Change, 123; 383-396 (2014).

309 19. Krey, V., Luderer, G., Clarke, L., Kriegler, E. Getting from here to there – energy

310 technology transformation pathways in the EMF27 scenarios. Climatic Change, 123, 369-

311 382 (2014).

312 20. Kriegler, E. et al. The role of technology for achieving climate policy objectives:

313 overview of the EMF 27 study on global technology and climate policy strategies.

314 Climatic Change, 123, 1-15 (2014).

315 21. Kriegler, E. et al. What does the 2°C targen imply for a global climate agreement in 2020?

316 The LIMITS study on Durban Platform Scenarios. Climate Change Economics, 4,

317 1340008 (2013).

318 22. Tavoni, M. et al. The distribution of the major economies' effort in the Durban platform

319 scenarios. Climate Change Economics, 4, 1340009 (2013).

320 23. Johansson, D. A. et al. Multi-model comparison of the economic and energy implications

321 for China and India in an international climate regime. Mitigation and Adaptation

322 Strategies for Global Change, 1-25, DOI:10.1007/s11027-014-9549-4, in press.

323 24. Lucas, P. L. et al. Implications of the international reduction pledges on long-term energy

324 system changes and costs in China and India. Energy Policy, 63, 1032-1041 (2013).

325 25. Luderer, G. et al. The economics of decarbonizing the energy system—results and

326 insights from the RECIPE model intercomparison. Climatic Change, 114, 9-37 (2012).

© 2015 Macmillan Publishers Limited. All rights reserved

327 26. Bauer, N. et al. Global fossil energy markets and climate change mitigation – an analysis

328 with REMIND. Climatic Change, 1-14, DOI 10.1007/s10584-013-0901-6, in press.

329 27. Calvin, K. et al. Implications of uncertain future fossil energy resources on bioenergy use

330 and terrestrial carbon emissions. Climatic Change, 1-12, DOI:10.1007/s10584-013-0923-

331 0, in press.

332 28. Chen, W., Yin, X., Zhang, H. Towards low carbon development in China: a comparison

333 of national and global models. Climatic Change, 1-14, DOI 10.1007/s10584-013-0937-7,

334 in press.

335 29. De Cian, E., Sferra, F., Tavoni, M. The influence of economic growth, population, and

336 fossil fuel scarcity on energy investments. Climatic Change, 1-17, DOI 10.1007/s10584-

337 013-0902-5, in press.

338 30. Luderer, G., Bertram, C., Calvin, K., De Cian, E., Kriegler, E. Implications of weak near-

339 term climate policies on long-term mitigation pathways. Climatic Change, 1-14, DOI

340 10.1007/s10584-013-0899-9, in press.

341 31. McCollum, D.L., Krey, V., Riahi, K. An integrated approach to energy sustainability.

342 Nature Clim. Change, 1, 428-429 (2011).

343 32. Pachauri, S. et al. Pathways to achieve universal household access to modern energy by

344 2030. Environmental Research Letters, 8, 024015, doi:10.1088/1748-9326/8/2/024015

345 (2013).

346 33. Riahi, K. et al. Chapter 17 - Energy Pathways for Sustainable Development, Global

347 Energy Assessment - Toward a Sustainable Future, Cambridge University Press,

348 Cambridge, UK and New York, NY, USA and the International Institute for Applied

349 Systems Analysis, Laxenburg, Austria, pp. 1203-1306 (2012).

© 2015 Macmillan Publishers Limited. All rights reserved

350 34. Ogden, J.M., Steinbugler, M.M., Kreutz, T.G. A comparison of hydrogen, methanol and

351 gasoline as fuels for fuel cell vehicles: implications for vehicle design and infrastructure

352 development. Journal of Power Sources, 79, 143-168 (1999).

353 35. Williams, J.H. et al. The Technology Path to Deep Greenhouse Gas Emissions Cuts by

354 2050: The Pivotal Role of Electricity. Science, 335, 53-59 (2012).

355 36. Ogden, J.M. Prospects for building a hydrogen energy infrastructure. Annual Reviews of

356 Energy and the Environment, 24, 227-279 (1999).

357 37. van Ruijven, B., Lamarque, J-F, van Vuuren, D.P., Kram, T., Eerens, H. Emission

358 scenarios for a global hydrogen economy and the consequences for global air pollution.

359 Global Environmental Change, 21, 983-994 (2011).

360 38. Balachandra, P., Reddy, B.S. Hydrogen Energy for the Transport Sector: A Well-To-

361 Wheel Techno-Economic and Environmental Feasibility Analysis. IGIDR, Mumbai, p.

362 35 (2007).

363 39. de Wit, M.P., Faaij, A.P.C. Impact of hydrogen onboard storage technologies on the

364 performance of hydrogen fuelled vehicles: A techno-economic well-to-wheel assessment.

365 International Journal of Hydrogen Energy, 32, 4859-4870 (2007).

366 40. GM Well-to-wheel analysis of energy use and greenhouse gas emissions of advanced fuel

367 / vehicle systems - A European Study. L-B systemtechnik, Ottobrun, Germany, p. 130

368 (2002).

369 41. Joint Research Centre. Well-to-Wheels analysis of future automotive fuels and

370 powertrains in the European context (2007).

© 2015 Macmillan Publishers Limited. All rights reserved

371 42. Isaac, M., van Vuuren, D. P. Modeling global residential sector energy demand for

372 heating and air conditioning in the context of climate change. Energy Policy, 37, 507-521

373 (2009).

374 43. Weiss, M., Junginger, M., Patel, M.K. Learning Energy Efficiency - Experience Curves

375 for Household Appliances and Space Heating, Cooling and Lighting Technologies.

376 Copernicus Institute, Utrecht, p. 223 (2008)

377 44. Ouellette, N., Rogner, H.H., Scott, D.S. Hydrogen-based industry from remote excess

378 hydroelectricity. Int. J. Hydrogen Energy, 22, 397-403 (1997).

379 45. Delucchi, M.A., Jacobson, M.Z. Providing all global energy with wind, water, and solar

380 power, Part II: Reliability, system and transmission costs, and policies. Energy Policy, 39,

381 1170-1190 (2011).

382 46. Fthenakis, V., Mason, J.E., Zweibel, K. The technical, geographical, and economic

383 feasibility for solar energy to supply the energy needs of the US. Energy Policy, 37, 387-

384 399 (2009).

385 47. Chukwuka, C. Folly, K. A. Overview of concentrated photovoltaic (CPV) cells. Journal

386 of Power and Energy Engineering, 2, 1-8 (2014)

387 48. Kurtz, S. Opportuniies and challenges for development of a mature concentrated

388 photovolraic power industry. Technical Report, NREL/TP-5200-432008 (2012)

389 49. Rephaeli, E. and Fan, S. Absorber and emitter for solar thermo-phtovoltaic systems to

390 achieve efficiency exceeding the Shockley-Queisser limit. Opt. Express, 17, 15145-

391 15159 (2009).

392 50. Bermel, P. et al. Design and global optimization of high-efficiency thermophotovoltaic

393 systems. Opt. Express, 18, A314-A334 (2010).

© 2015 Macmillan Publishers Limited. All rights reserved

394 51. International Renewable Energy Agency, Renewable energy technologies: cost analysis

395 series – Concentrating solar power (2012).

396

© 2015 Macmillan Publishers Limited. All rights reserved

397 Table:

398 Table S1 | Global totals for solar panel production, and changes in climate variables in solar
399 panel installed desert regions

Control Changes from Control

SPDU area SPDLess area SPDU SPDU+UH SPDLess
Energy production 0 0 739±5 740±5 59±1
(TW)
Energy production 0 0 48±1 48±1 0
urban only (TW)
SPD incident direct 2703±13 216±2 35±19 49±18 1±2
solar radiation (TW)
SPD total cloud cover 21.5±1 9.8±1.6 -1.1±1.3 -1.5±1.3 -0.3±2.2
(%)
SPD absorbed direct 1955±9 125±5 -374±3 -367±3 5±1
solar radiation (TW)
SPD reflected direct 748±5 89±1 -330±3 -330±3 -63±1
solar radiation (TW)
SPD T in desert solar 16.24±0.25 22.24±.40 -2.35±0.20 -2.17±0.15 0.04±0.64
panel region (oC)
SPD P in desert solar 291±4.1 29±14 -61±17 -68±18 -7±11
panel region (mm/yr)
SPD Albedo 0.295±0.003 0.410±0.001 .114±0.003 .113±0.003 -.026±0.001
400 SPD represents solar panel installed desert area. T denotes temperature and P denotes
401 precipitation. The values for the Control are either area mean or area sum, and changes in SPDU,
402 SPDU+UH and SPDLess are with respect to the same area mean/sum in the Control.
403 Negative/positive values represent a decrease/increase relative to the Control. The numbers after
404 the ± sign represent the uncertainty which is represented by one standard deviation of that
405 variable. TW is Terawatts, 1 TW=1012 Joules/second. mm/yr is millimeter per year

406

© 2015 Macmillan Publishers Limited. All rights reserved

407 Table S2 | Percentage changes of the radiation, cloud cover and precipitation in solar panel
408 installed desert area relative to Control run

Percentage Changes from Control

SPDU SPDU+UH SPDLess
SPD incident direct solar radiation (TW) 1.3±0.7% 1.8±0.7% 0.5±1.0%
SPD total cloud cover (%) -5±6% -7±6% -3±200%
SPD absorbed direct solar radiation (TW) -19±0.2% -19±0.2% 4±0.8%
SPD reflected direct solar radiation (TW) -44±0.4% -44±0.4% -77±1%
SPD P in desert solar panel region (mm/yr) -21±6% -23±6% -21±38%
409 Definitions of the variable in this table are the same as in Table S1, but for the percentage change
410 in the SPDU, SPDU+UH, and SPDLess relative Control.

411

412 Table S3 | Temperature, precipitation, and radiation on global and regional scales

Control Changes from Control

SPDU SPDU+UH SPDLess
Global1 mean temperature (oC) 15.08±0.13 -0.34±0.12 -0.25±0.12 -0.04±0.13
Land mean temperature (oC) 12.11±0.21 -0.58±0.19 -0.41±0.18 -0.04±0.20
Urban mean temperature (oC) 21.10±0.20 -0.26±0.19 0.84±0.21 0.00±0.20
Global mean Precipitation 1131±11 -6±4 -4±4 1±4
(mm/yr)
Land mean precipitation (mm/yr) 901±29 -8±4 0±6 1±6
Global incident solar radiation 97394±151 231±178 177±166 18±164
(TW)
Land incident solar radiation (TW) 31709±112 111±113 110±125 16±107
Ocean incident solar radiation 65685±118 120±151 67±144 2±125
(TW)
Global absorbed solar radiation 84801±140 -274±135 -284±143 0±142
(TW)
Land absorbed solar radiation 23936±92 -320±88 -315±96 13±89
(TW)
Ocean absorbed solar radiation 60865±113 46±97 31±109 -13±117
(TW)
1
413 Includes land and ocean
414 The values for the Control are either area mean or area sum, and changes in SPDU, SPDU+UH
415 and SPDLess are with respect to the same area mean/sum in the Control. Negative/positive
416 values represent a decrease/increase relative to the Control. The numbers in parenthesis are the
417 percentage changes relative to the Control. The numbers after the ± sign represent the
418 uncertainty which is represented by one standard deviation of that variable. TW is Terawatts, 1
419 TW=1012 Joules/second. m/yr represents meter per year.
420

© 2015 Macmillan Publishers Limited. All rights reserved

421 Table S4 | Global and regional mean temperature of 1986-2005 for CMIP5 model ensemble
422 and for CCSM4, and the global and regional mean precipitation of 1986-2005 for CCSM4
423 only
Temperature (oK) Precipitation (mm/year)
CMIP5 global 287.7±0.14
CCSM4 global 287.4±0.20 1085±4
desert 288.4±0.40 284±6
desertLess 294.5±0.31 30±6
424
425 Table S5 | Global mean temperature change relative to 1985-2005 mean for CMIP5 model
426 ensemble (oC)
427
Periods RCP2.6 RCP4.5 RCP6.0 RCP8.5
2041-2060 1.02±0.02 1.28±0.01 1.15±0.02 1.81±0.01
2081-2100 1.08±0.02 1.78±0.01 2.21±0.02 3.74±0.01
2011-2100 0.96±0.11 1.28±0.09 1.33±0.06 2.13±0.09
428
429 Table S6 | Global and regional mean temperature and precipitation changes averaged over
430 2011-2100 relative to climatological mean of 1985-2005 for CCSM4 and solar panel
431 simulations
432
Temperature (oC) Precipitation (mm/year)
global Desert desertLess global desert desertLess
RCP2.6 0.79±0.09 0.97±0.16 0.90±0.20 18.7±2.4 5.9±10.4 -1.7± 7.0
RCP4.5 1.20±0.10 1.54±0.17 1.38±0.21 24.4±2.1 4.7± 9.4 -0.6± 8.2
RCP6.0 1.30±0.06 1.66±0.13 1.59±0.20 24.5±2.1 4.7± 9.1 -0.7± 7.8
RCP8.5 2.02±0.10 2.58±0.15 2.54±0.20 36.9±2.4 11.3± 7.9 1.8± 8.0
Control 0.79±0.11 0.98±0.13 0.89±0.41 18.4±4.1 6.7± 4.1 -0.6±14.2
SPDU 0.45±0.10 -1.36±0.20 ̶ 12.5±4.1 -54.5±17.2 ̶
SPDU+UH 0.54±0.12 -1.19±0.15 ̶ 13.8±4.4 -61.0±17.5 ̶
SPDLess 0.75±0.12 ̶ 0.93±0.45 17.5±4.1 ̶ -6.83±11.4
433

© 2015 Macmillan Publishers Limited. All rights reserved

434

435 Table S7 | Maximum values for solar electricity production and total primary energy use

436 from the IPCC AR5 scenarios database and derived maximum values for solar electricity

437 demand

TW EJ

2010 2050 2100 2010 2050 2100

All scenarios Electricity production, solar 0.015 4 17 0.462 131 525
Total primary energy 17 41 63 523 1281 1980
If all final energy were solar-
electricity 15 30 45 472 957 1422

Low Electricity production, solar 0 4 17 0 131 525

scenarios
Total primary energy 17 33 45 523 1046 1421
If all final energy were solar-
electricity 14 25 31 450 797 986

438

© 2015 Macmillan Publishers Limited. All rights reserved

439 Figures:

440

441 Figure S1 | Areas where solar panels are installed. a desert areas and b urban areas. Green

442 stippling in panel a indicates where solar panels are installed for experiment SPDLess. Panel b

443 shows the percentage of the urban area in each model grid cell.

444

© 2015 Macmillan Publishers Limited. All rights reserved

445

446 Figure S2 | Global mean temperature time series. Upper panel is the absolute values of the
447 global mean temperature for the four simulations, and lower panel is the global mean
448 temperature anomaly for the three sensitivity simulations relative to the Control.

© 2015 Macmillan Publishers Limited. All rights reserved

449

450 Figure S3 | Global mean temperature evolution for coupled model intercomparison project
451 phase 5 (CMIP5).

© 2015 Macmillan Publishers Limited. All rights reserved

452

453 Figure S4 | Time evolving global mean temperature and precipitation for RCP2.6, RCP4.5,
454 RCP6.0, RCP8.5 from CCSM4 and for solar panel sensitivity simulations. Left panels are
455 the temperature and right panels are the precipitation. Top panels are the global means, mid-
456 panels are the mean of solar panel installed desert areas in Control, SPDU and SPDU+UH
457 experiments, and bottom panels are the mean of solar panel installed small desert areas in
458 Control and SPDLess experiments.

© 2015 Macmillan Publishers Limited. All rights reserved

459

460 Figure S5 | Ensemble mean temperature and precipitation anomaly relative to the
461 ensemble mean of RCP2.6 for RCP4.5, RCP6.0 and RCP8.5 averaged over 2011-2100.
462 Contour interval for temperature is 0.2oC and for precipitation is 0.025m/yr. Stippling indicates
463 changes are significant at the 95% level using a double sided student t-test.

© 2015 Macmillan Publishers Limited. All rights reserved

464

465 Figure S6 | Surface property anomalies relative to Control in SPDU. a the latent heat flux; b
466 sensible heat flux; c total leaf area index; d sea level pressure and surface wind. The units for
467 latent heat, sensible heat, and sea level pressure are given at the top-right corner of each panel.
468 The unit for surface wind is m/s and the leaf area index is m2 of leaf area per m2 of ground area.
469 Contour interval for latent and sensible heat flux is 2 W/m2, for total leaf area index is 0.2, and
470 for sea level pressure is 0.2 hPa. Stippling indicates changes are significant at the 95% level
471 using a double sided student t-test.

472

© 2015 Macmillan Publishers Limited. All rights reserved

473

474 Figure S7 | 500hPa geopotential height (shading) and wind (vector). a Control; b
475 geopotential height and wind anomaly relative to the Control in SPDU; c geopotential height and
476 wind anomaly relative SPDU in SPDU+UH; d the same as b but in SPDLess. The unit for
477 geopotential height is meters and for wind is m/s. Contour interval is 100 hPa for a, 4 hPa for b,
478 and 2 hPa for c and d. Stippling indicates changes are significant at the 95% level using a double
479 sided student t-test.

© 2015 Macmillan Publishers Limited. All rights reserved

480
481 Figure S8 | Mean zonal wind. a Control; b zonal wind anomaly relative to the Control in SPDU;
482 c zonal wind anomaly relative to SPDU in SPDU+UH; d the same as b but for SPDLess. The
483 unit is m/s. Contour interval is 5 m/s for Control, but 0.2 m/s for panels b-d. The stippling
484 indicates changes are significant at the 95% level using a double sided student t-test.

485

© 2015 Macmillan Publishers Limited. All rights reserved

486

487 Figure S9 | Surface air temperature (left panels) and precipitation (right panels) changes in
488 the solar panel production sensitivity experiments. a/b temperature/precipitation difference
489 between experiments SPDL and control; c/d temperature/precipitation difference between
490 experiments SPDUL and SPDL. The numbers at upper right corner of each panel represents the
491 global average difference. The unit is oK for temperature and meter/year (m/yr) for precipitation.
492 Contour interval for temperature is 0.1oC, and for precipitation is 0.05 m/yr. Stippling indicates
493 changes are significant at the 95% level using a double sided student t-test. These simulations are
494 discussed in the main text and the supplementary material.

© 2015 Macmillan Publishers Limited. All rights reserved

495

496 Figure S10 | A sensitivity experiment which is similar to SPDU, but with solar panel
497 installation expanded to including the entire Sahara Desert and deserts in the Middle East,
498 China and Mongolia. Panel a shows the regions where solar panels are installed in desert
499 regions only, and Panel b shows a reduced desert area where the solar panels are installed. The
500 area reduction in the bottom panel is equivalent to the total urban area in the model. Thus the
501 total areas where solar panels are installed are exactly the same in experiments SPDL and
502 SPDUL.

503

View publication stats

© 2015 Macmillan Publishers Limited. All rights reserved