You are on page 1of 21


MAY 5, 2017

PNAS reviewed the attached Clack et al. opinion piece as a scientific article despite a requirement by
PNAS that articles must contain new science, which Clack et al. does not. The sole content of Clack et
al. is criticism plus false and unbalanced claims and data about the scientific article by Jacobson MZ,
Delucchi MA, Cameron MA, Frew BA (2015). A low-cost solution to the grid reliability problem with
100% penetration of intermittent wind, water, and solar for all purposes. Proc. Natl. Acad. Sci 112:

The proper format, according to PNAS guidelines, for such a submission by Clack et al. is a 500-word
letter to the editor, which Jacobson et al. (2015) could respond to. Yet, PNAS allowed this submission as a
manuscript plus supplemental information with zero new science content, maximizing the ability of Clack
et al. to smear and falsify information. PNAS offered only a 500-word response to Jacobson et al., which
Clack et al. could then respond to further according to PNAS rules. Such a format would prevent a line-by-
line response to the over 40 false statements by Clack et al. in the main text alone.

PNAS then failed to adhere to their policy of requiring each author on a paper to make a substantial
contribution, allowing Clack et al. to pile on a total of 22 authors, most of whom made little contribution to
the paper, to maximize the credibility of the Clack et al. document. Yet, many of the authors are not
experts on the subject being criticized. For example, Ken Caldeira acknowledged publicly on February 23,
2016, I am not an energy expert yet appears on this paper.

PNAS acknowledges that several of the coauthors have had historic financial conflicts of interest in what
they are arguing against by receiving funds from Exxon Mobil, EPRI, and the Deep Decarbonization
Initiative, for example, but still allows the paper to be passed off as a science paper with balance, rather
than an opinion piece written by authors, some of whom have a financial interest in what they are saying.

When PNAS first provided Jacobson et al. (2015) a copy of the Clack et al. paper, after PNAS had accepted
the paper, Jacobson et al. provided a list of around 40 false statements and requested that these be sent to
all the authors for them to correct the paper and for PNAS to withdraw the paper and convert it to a letter, a
reasonable request. PNAS, however, did not appear to provide any of the authors with the comments and
decided to publish the paper as is despite the false and defamatory claims and data listed herein. They then
provided Jacobson et al. with a copy of the article, again with the disproportionate offer of responding in
500 words or less and not guaranteeing that a response would be published side by side.

The refusal of PNAS to address any of the main false and defamatory comments by Clack et al., thereby
leaving fake data, including fake numbers, and information in the document, as illustrated below, leaves us
no choice but to correct the defamation publicly and clearly. PNAS | May 4, 2017 | vol. XXX | no. XX | 17


Evaluation of a proposal for reliable low-cost grid

power with 100% wind, water, and solar
Christopher T. M. Clack1,2,3*, Staffan A. Qvist4, Jay Apt5,6, Morgan Bazilian7, Adam Brandt8, Ken Caldeira9, Steven J. Davis10,
Victor Diakov11, Mark Handschy2,12, Paul Hines13, Paulina Jaramillo5, Daniel M. Kammen14-16, Jane C. S. Long17, M. Granger
Morgan5, Adam Reed18, Varun Sivaram19, James Sweeney20,21, George R. Tynan22, David G. Victor23,24, John P. Weyant20,21,
and Jay F. Whitacre5
Earth System Research Laboratory, NOAA, 325 Broadway, Boulder, Colorado 80305, USA; 2 Cooperative Institute for Research in Environmental Sciences, University of
Colorado, Boulder, Colorado 80305, USA; 3 Now at Vibrant Clean Energy, LLC, Boulder, CO, USA; 4 Department of Physics and Astronomy, Uppsala University,
Lagerhyddsvagen 1, Uppsala, Sweden; 5 Department of Engineering and Public Policy, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA; 6 Tepper
School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA; 7 Center for Global Energy Policy, Columbia University, New York, NY, USA;
Department of Energy Resources Engineering, Stanford University, 367 Panama Street, Stanford, CA 94305, USA; 9 Department of Global Ecology, Carnegie Institution for
Science, 260 Panama Street, Stanford, CA 94305, USA; 10 Department of Earth System Science, University of California, Irvine, Croul Hall, Irvine, CA 92697, USA; 11 Omni
Optimum, Evergreen, Colorado, USA; 12 Enduring Energy, LLC, Boulder, CO, USA; 13 University of Vermont, Electrical Engineering and Complex Systems Center, Burlington,
VT, 05405, USA; 14 Energy and Resources Group, University of California, 310 Barrows Hall, Berkeley, CA 94720 USA; 15 Goldman School of Public Policy, University of
California, 2607 Hearst Ave, Berkeley, CA 94720, USA; 16 Renewable and Appropriate Energy Laboratory, University of California, 310 Barrows Hall, Berkeley, CA 94720-3050,
USA; 17 Lawrence Livermore National Laboratory, Livermore, CA, USA (retired); 18 Renewable and Sustainable Energy Institute, University of Colorado, Boulder, Colorado
80305, USA; 19 Council on Foreign Relations, 58 E 68th St, New York, NY 10065, USA; 20 Precourt Energy Efficiency Center, Stanford University, 473 Via Ortega, Stanford, CA
94305-4206, USA; 21 Management Science and Engineering Department, Stanford University, Huang Engineering Center, 475 Via Ortega, Stanford, CA 94305, United States;
Department of Mechanical and Aerospace Engineering, Jacobs School of Engineering, UC San Diego, Voigt Drive, La Jolla, CA 92093, USA; 23 School of Global Policy &
Strategy, UC San Diego, La Jolla, CA 92093, USA; 24 Brookings Institution, 1775 Massachusetts Ave, NW, Washington DC 20036

This manuscript was compiled on May 4, 2017

A large number of analyses and assessments, including those per- investments in new nuclear power plants are currently
formed by the Intergovernmental Panel on Climate Change, the Na- generally not economically attractive within liberalized
tional Oceanic and Atmospheric Administration, the National Renew-
able Energy Laboratory, and the International Energy Agency, have
concluded that deployment of a diverse portfolio of clean energy
technologies makes a transition to a low-carbon-emission energy Similarly Freed et al. (2017,
system both more feasible and less costly than other pathways. way/is-nuclear-too-innovative-a14fb4fef41a#.qag59xnk0),
who are strong nuclear advocates, state,
(1) False. IPCC does not perform
new research; it summarizes other Indeed, there is virtually no history of nuclear
studies, and it suggests the exact construction under the economic and institutional
opposite. circumstances that prevail throughout much of Europe
and the United States.
IPCC Working Group III, Chapter 7
Section P. 534. high shares of variable As such, the premise of Clack et al. that diverse
RE power, for example, may not be ideally portfolios that include nuclear are less costly is false
complemented by nuclear, CCS, and CHP plants information.
(without heat storage).
(2) Further none of the deep
IPCC further states that nuclear is not decarbonization studies has compared
economically attractive, thus it cannot be less their results with a 100% scenario, such
costly as claimed by Clack et al: as in Jacobson et al. (2015), and it would
be impossible for them to do so because
IPCC Working Group III, Chapter 7, Section every one of these studies, including
7.8.2. P. 542. Potential (nuclear) project and MacDonald et al. (2016), NREL (2012),
financial risks are illustrated by the significant UNDDP (2015), and EMF (2014), etc.
time and cost over-runs of the two novel European failed to account for a (a) reduction in
Pressurized Reactors (EPR) in Finland and power demand of around 13% due to
France. Without support from governments, eliminating energy from mining, PNAS | May 4, 2017 | vol. XXX | no. XX | 17

clusively on wind, solar and hydroelectric power.
refining, and transporting fossil
fuels and uranium; (b) a reduction (5) False. Jacobson et al. (2015) do not
in power demand of around 23% rely exclusively on wind, solar, and
due to the higher work to energy hydroelectric power. They also relay on
ratio of electricity over combustion geothermal, tidal, wave, water storage,
resulting from electrification and ice storage, rock storage, pumped hydro
providing electricity from clean, storage, CSP storage, hydrogen storage,
renewable energy. short-and-long-distance transmission,
and demand response.
In contrast, Jacobson et al. [Jacobson MZ, Delucchi MA, Energy Systems Modelling | Climate Change | Renewable Energy |
Cameron MA, Frew BA (2015) Proc Natl Acad Sci USA Energy Costs | Grid Stability
112(49):1506015065] argue that it is feasible to provide

"low-cost solutions to the grid reliabil- ity problem with
100% penetration of WWS [wind, water and solar
number of studies, including a study by one of us, have
concluded that an 80% decarbonization of the U.S. elec-
tric grid could be achieved at reasonable cost [1, 2].
power] across all energy sectors in the continental United
States be- tween 2050 and 2055", with only electricity
and hydrogen as energy carriers. The high level of decarbonization is facilitated by an optimally config-
ured continental high voltage transmission network. There appears
to be some consensus that substantial amounts of greenhouse gas
(3) False. There are many additional emissions could be avoided with widespread deployment of solar and
studies aside from Jacobson et al. wind electric generation technologies along with supporting
(2015) that find 100% or near-100% infrastructure.
Further, it is not in question that it would be theoretically
clean, renewable energy grid possible to build a reliable energy system while excluding all
scenarios feasible for the electric bioenergy, nuclear energy, and fossil-fuel sources. Given un- limited
power grid or multiple energy resources to build variable energy production facilities, while
expanding the transmission grid and accompanying en- ergy storage
sectors (Mason et al., 2010; Hart and capacity indefinitely, one would eventually be able
Jacobson, 2011, 2012; Connolly et al., 2011;
2014; 2016 Mathiesen et al., 2011; 2015;
Denholm and Hand, 2011; Elliston et al., 2012,
2013, 2015; Rasmussen et al., 2012; Nelson et
al., 2012; Budischak et al., 2013; Steinke et al.,
2013; Connolly and Mathiesen, 2014; Becker et
al., 2014; Bogdanov and Breyer, 2016). In fact
there are many more such studies than deep
decarbonization studies, most of which are not
even grid studies.

In this paper, we evaluate that study and find significant

shortcomings in the analysis. In particular, we point out that this
work used invalid modeling tools, contained modeling errors, and
made incorrect, implausible, and inadequately supported assump-

(4) False. Clack et al. provides zero

evidence of modeling errors in
Jacobson et al. (2015), as illustrated
in responses to follow.
Policy makers should treat with caution any visions of a rapid,
reliable and low-cost transition to entire energy systems that rely ex- PNAS | May 4, 2017 | vol. XXX | no. XX | 17

to meet any conceivable load. Yet in
developing a strategy to effectively mitigate
global energy-related CO2 emissions, it is
critical that the scope of the challenge to
achieve this in the real world is accurately
defined and clearly communicated.
Wind and solar are variable energy
sources, and some way must be found to
address the issue of how to provide energy if
their immediate output cannot continuously
meet instantaneous demand. The main options
are to: (1) curtail load (i.e., modify or fail to
satisfy demand) at times when energy is not
available, (2) deploy very large amounts of
energy storage, or (3) provide supplemental
energy sources that can be dispatched when
needed. It is not yet clear how much it is
possible to curtail loads, especially over long
durations, without inducing large economic
costs. There are no electric storage systems
available today that can affordably and
dependably store the vast amounts of energy
needed over

Significance Statement
Previous analyses have found that the most feasible route to a
low-carbon energy future is one that adopts a diverse portfolio
(6) False. See Response
of technologies.
(3) for numerous studies that
assume 100% clean, renewable
energy. In contrast, Jacobson et al. (2015) consider
whether the future primary energy sources for the United States
could be narrowed to only wind, solar and hydroelectric power
(7) False. See response (5) and suggest
that this can be done at "low-cost" in a way that supplies all
power with a probability of loss of load "that ex- ceeds
electric-utility-industry standards for reliability". We find that
(8) False. The
their analysis involves errors
authors show no modeling errors. ,
inappropriate methods, and implausible assumptions. Their
study does not provide credible evidence for rejecting the
To whom correspondence should be addressed. E-mail:
conclusions of previous analyses that point to the benefits of
considering a broad portfolio of energy system options. A
policy prescription that over-promises on the benefits of
relying on a narrower portfolio of technologies options could
be counterproductive, seriously impeding the move to a cost
effective decarbonized energy system. PNAS | May 4, 2017 | vol. XXX | no. XX | 17

weeks to reliably satisfy demand using expanded wind and challenges are deepened by placing constraints on technological
solar power generation alone. These facts have led many options.
U.S. and global energy system analyses [110] to recognize
the importance of a broad portfolio of electricity generation (11) False. The energy and storage
technologies including sources that can be dispatched when portfolio in Jacobson et al. (2015)
is broader than in any of the
(9) False. This statement fails to studies referenced by Clack et al.
address the numerous studies that For example, MacDonald et al. (2016)
exclude such portfolios, as listed did not include CSP, tidal, wave, geothermal,
under Response (3). or any storage technology (water, ice, rocks,
CSP-with-storage, pumped hydro storage,
Major Faults with the Jacobson et al. analyses hydroelectric storage, hydrogen storage), nor
even demand response. Further, despite
Jacobson et al. [11], along with additional colleagues in a
companion article [12], attempt to demonstrate the feasibility Clack et als criticism here for not treating
of supplying all energy end uses [in the continental United CCS or bioenergy, MacDonald et al. (2016)
States] with only Wind, Water and Solar (WWS) power (no also did not treat CCS or bioenergy.
coal, natural gas, bioenergy, or nuclear power), while meeting
all loads, at reasonable cost. Such a scenario may be a useful
Similarly, (EMF, 2014) excluded storage and
way to explore the hypothesis that it is possible to meet the demand response and failed to calculate the
challenges associated with reliably supplying energy across all costs of the risks associated with nuclear and
sectors almost exclusively with large quantities of a narrow CCS. In fact, no grid integration study prior
range of variable energy resources.
to Jacobson et al. (2015) has included the
However, there is a differ- ence between presenting such combination of UTES storage, hot/cold
visions as thought experiments and asserting, as the authors
do, that rapid and complete conversion to a 100% wind,
storage in water, cold storage in ice, electrical
solar and hydroelectric power system storage (CSP, pumped hydro, hydropower)
and hydrogen storage. Thus, rather than
(10) False. Jacobson et al. (2015) do narrowing technologies available, Jacobson et al.
not rely exclusively on wind, solar, (2015) expand technologies relative to MacDonald
and hydroelectric power, as stated et al. (2016) and EMF (2014), among others.
under Response (5)
is feasible with little downside [12]. It is important to
In our view, to demonstrate that a proposed energy sys- tem is
understand the distinction between physical possibility and
technically and economically feasible, a study must, at a
feasibility in the real world. To be clear, the specific aim ofminimum, demonstrate through transparent analysis
Jacobson et al. [11] is to provide "low-cost solutions to the grid
reliability problem with 100% penetration of WWS [wind, wa-

ter and solar power] across all energy sectors in the continental (12) This statement falsely implies that our
United States between 2050 and 2055". analysis was not transparent. The entire
LOADMATCH model has been available on
Relying on 100% wind, solar and hydroelectric power could make
climate mitigation more difficult and more expensive than it request since the original PNAS paper was
needs to be. For example, the analysis by Jacobson et al. [11, published, and multiple people have requested
12] exclude from consideration several commercially available and obtained the model. Further, the 50s-state
technologies such as nuclear and bioenergy that could potentially
contribute to decarbonization of the global energy system, while paper spreadsheet to this day is posted on the
also helping assure high levels of reliability in the power grid. internet, available for anyone to see. Every
Further, Jacobson et al. [11, 12] exclude carbon capture and number in both papers is transparent.
storage technologies for fossil fuel generation. An additional
option not considered in the 100% wind, solar and hydroelectric
and validated modeling that the required technologies have
studies is bioenergy coupled with carbon capture and storage
(CCS) to create negative emissions within the system, which been commercially demonstrated at scale at a cost
could help with emissions targets. With all available comparable to alternatives; that the technologies can, at scale,
technologies at our disposal, achieving an 80% reduc- tion in provide adequate and reliable energy; that the deployment
greenhouse gas emissions from the electricity sector at rate re- quired of such technologies and their associated
reasonable costs is extremely challenging, even using a new con- infrastructure is plausible and commensurate with other
tinental scale high voltage transmission grid. Decarbonizing the historic examples in the energy sector; and the deployment
last 20% of the electricity sector, as well as decarbonizing the and operation of the technologies do not violate environmental
rest of the economy that is difficult to electrify (e.g., ce- ment regulations. We
manufacture, aviation), is even more challenging. These
2 | Clack et al.

demonstrate that ref. [11] and [12] do not meet of buildings. Right now, 60% of Denmarks heat
these crite- ria and, accordingly, do not show comes from district heating, but with water instead
the technical, practical or economic feasibility of
a 100% wind, solar and hydroelectric energy
of rocks, so the claim that this cant be done on a
vision. As we detail below and in the large scale is false. Further, the UTES system in
Supporting Information, ref. [11] contains Drakes Landing, Okotoks Canada has been
modeling errors, operating among 52 homes successfully since 2005,
(13) False. The authors demonstrate where the storage cost is less than 1/300th the cost of
no modeling errors. batteries. The grass-covered storage rocks serve as a
community park, and it is impossible to tell visually
incorrect, im- plausible and/or inadequately that the storage exists.
supported assumptions, and the application of
methods inappropriate to the task. In short, the
analysis performed in ref. [11] does not support It is not difficult to match instantaneous energy
the claim that such a system would perform at demands for all purposes with variable electricity
reasonable cost and provide reliable power. generation sources in real time as needed to assure
reliable power supply if one assumes, as the authors
The vision proposed by the studies in ref. [11, of the ref. [11] do, that there exists a nationally
12] narrows options, integrated grid, that most loads can be flexibly
shifted in time, that large amounts of multi-week
and seasonal energy storage will be readily
(14) False. See response (11) available at low cost, and that the entire econ-
omy can easily be electrified or made to use
hydrogen. But adequate support for the validity of
yet includes a wide range of currently these assumptions is lack- ing. Furthermore, the
uncosted in- novations that would have to be conclusions in ref. [11] rely heavily on free, non-
deployed at large scale (e.g., replacement of our modeled hydroelectric expansion;
current aviation system with yet-to-be-
developed hydrogen-powered planes). The system
in ref. [11] assumes the availability of multi-week (17) False. Jacobson et al. (2015)
energy storage systems that are not yet
demonstrated at scale and deploys them at a
assumed zero expansion of hydroelectric
capacity twice that of the entire U.S. power power reservoirs, only expansion of
system today. turbine capacity. The cost of turbine
expansion was not included in the paper,
(15) False. Clack et al. fail to recognize that but the cost has since been calculated for
Jacobson et al.s (2015) 100% WWS plans require the U.S. and worldwide. The mean U.S.
electrifying all energy sectors. They falsely cost is 0.2 cents/kWh, which is roughly
compare how much storage is needed with the 2% of the cost of overall energy. This
capacity of the U.S. power sector today, not even cost is derived from the fact that the
recognizing the U.S. power sector is only one-fifth cost of electrical equipment (turbines,
of all U.S. energy. Jacobson et al. (2015) are
generators, and transformers) in a
proposing a solution for 100% of all energy, not
hydropower plant ranges from
just electricity, so the storage capacity is 2/5 of the
energy modeled, not 2 times. ~$560/kW for 500 MW plants to ~$200-
$300/kW for 1000 MW plants (Figs. 4.5
and 4.7 of IRENA, 2012a). We assume
There would be underground thermal energy
storage systems deployed in nearly every home,
large 1000-MW plants but also assume
business, office building, hospi- tal, school and construction costs are generally higher
factory in the United States. than anticipated, so we assume the
additional cost per MW of hydropower
(16) False. Clack et al. misstate how UTES turbines is 16% of the hydropower
works. Underground systems themselves are not capital cost, or $468 (387-548) / kW.
deployed in nearly every home Underground When this cost is multiplied by the
rocks storage beds are centralized, district heating, fraction of total end use energy from
systems. Only solar collectors are on the rooftops hydro, the additional hydropower
2 | Clack et al.

turbine portion of total energy cost is hydroelectric power system [11], we focus on four
major issues that are explored in more detail
less than 2%. below and in the Supporting Information. (i) We
note several modeling errors presented in ref.
[11] that invalidate the results in the studies,
massive scale-up of hydrogen production and
particularly with respect to the amount of
hydropower available and the demand response of
(18) False. The scale-up of hydrogen flexible loads (SI Appendix section S1).
proposed is much less than the scale-up (22) False. The authors demonstrate no
of transmission required in modeling errors.
MacDonald et al (2016) and less than
the nuclear or coal-CCS scale-up
(ii) We examine poorly documented and
required in the other papers cited implausible assumptions, including: the cost and
herein. scalability of storage technologies; the use of
hydrogen fuels; lifecycle assessments of
unconstrained, non-modeled transmission
technologies; cost of capital and capacity factors of
existing technologies; and land use (SI Appendix
section S2). (iii) We discuss the studies lack of
electric power system modeling of transmission,
(19) False. The cost of both short reserve margins and frequency response; despite
and long-distance transmission and claims of system reliability (SI Appendix section
distribution were accounted for. S3). (iv) Lastly, we argue that the climate/weather
model used for estimates of wind and solar
; and free time-shifting of loads at large- scale in energy production has not been sufficiently peer-
response to variable energy provision. None of reviewed and has not demonstrated the ability to
these are going to be achieved without cost. accurately simulate wind speeds or solar insolation
Some assumed expan- sions, such as the at the scales needed
hydroelectric power output, imply operating
facilities way beyond existing constraints that
have been es- tablished for important (23). False. Zhang (2008), who reviewed
environmental reasons. Without these elements, coupled climate-air quality models, determined
the costs of the energy system in ref. [11] would
be substantially higher than claimed
GATOR-GCMOM to be the first fully-
coupled online model in the history that
accounts for all major feedbacks among major
(20) False. See Responses (17) and (19). The only
atmospheric processes based on first
relevant cost omitted was the hydropower turbine principles. As such, it is the most complete
expansion, and that cost is less than 2% the cost of model worldwide.
overall electricity.
Further, GATOR-GCMOM has been validated
and, more importantly, the system would fail at high and low resolution for paired-in-time-
to deliver reliable energy. and-space wind and solar data as well as global
cloud, humidity, stability, and fields in multiple
(21) False. Jacobson et al. (2015) studies, not only by Jacobson et al. (1996, 2007,
provided numerous low-cost solutions. 2014) and Jacobson and Kaufman (2006),
Clack et al. (2015) provide zero Jacobson and Archer (2012), and Jacobson
evidence that additional solution (1997, 1998, 1999a,b, 2001a,b, 2002, 2004, 2010,
cannot be obtained, with, for example, 2012, 2014), but also others (e.g., Whitt et al.,
less expansion of hydroelectric turbine 2011; Ten Hoeve et al., 2012). In addition, over
1000 researchers have used algorithms from
capacity. In fact, many solutions with
GATOR-GCMOM and dozens have either used
much less hydro turbine expansion and
or seen the inner workings of the code. Further
zero expansion have been obtained (see a textbook was written describing many
Response 24). algorithms and all other algorithms are
described in over 50 peer-reviewed papers
In evaluating the 100% wind, solar and where the model has been developed, evaluated,
2 | Clack et al.

and/or applied.

Finally, virtually every weather-climate model

has copied or adopted some or many its
techniques, including interactively coupling
aerosols, clouds, radiation, and meteorology
with feedback. In one example, the NCAR
WRF-CHEM model started using the
GATOR-GCMOM technique of online
coupling between gases, aerosols, and
meteorology only in 2005, 11 years after
GATOR-GCMOM developed that technique,
as described in Jacobson (2006, Comment on
fully coupled online chemistry within the
WRF model, by Grell et al. 2005. Atmos.
Environ, 39, 6957-6975, Atmos. Environ. 40,
4646-4648). GATOR-GCMOM also contains
hundreds of processes still not treated in any
other global model.

to assure

2 | Clack et al.

the technical reliability of an energy system relying so heavily on clearly stated on page 15,061, the annual heating and
intermittent energy sources (SI Appendix section S4). cooling loads are distributed every 30 seconds
according to the number of heating and cooling
Modeling errors
degree days, respectively, each year. Thus, the
As we detail in Supporting Information section S1 of this
paper, ref. [11] includes several modeling mistakes that call into
flexible load at any moment could be higher or lower
question the conclusions of the study. For example, the numbers than the average load in Table 1. Figure 3 is
given in the Supporting Information of ref. [11] imply that perfectly fine. The LOADMATCH code also
maximum output from hydroelectric facilities cannot exceed contains this information, which the authors of the
145.26 GW (see our Section S1.1), about 50% more than exists
in the U.S. today [13], yet Figure 4(b) of ref. [11] (our Fig. 1)
commentary could have requested but didnt.
shows hydroelectric output exceeding 1,300 GW.

(24). False. Increasing the discharge rate was not a
modeling mistake but a model assumption, and Dr.
Clack is well aware that it was not a modeling
mistake. On Monday, February 29, 2016, Dr. Clack
was informed by email that Jacobson et al. (2015) Fig. 1. This figure (Figure 4(b) from ref. [11]) shows hydropower supply rates peaking at
assumed an increase in discharge rate while keeping nearly 1,300 GW despite the fact that their proposal calls for less than 150 GW of
hydropower capacity. This indicates a major error in their analysis.
annual energy output constant, as clearly evidence by
Footnote 4 of Table S.2 of Jacobson et al. (2015). This In the analysis in ref. [11], the flexible loads can be accu-
is also obvious from the LOADMATCH code itself, mulated in eight-hour blocks; which raises a serious issue of
which was available to Dr. Clack or anyone else upon extreme excess capacity to utilize the high power for short
request. In fact, multiple people requested the code. periods of time. Under these assumptions, there would need to
be oversized facilities on both the demand and generation sides
The only omission was including the cost of the to compensate for their respective variabilities. These errors
additional turbines needed to increase the discharge are critical, as the conclusions reached by ref. [11] depend on
rate, and that has since been costed at about 2% the the availability of large amounts of dispatchable energy and a large
degree of flexibility in demand.
cost of total energy (see Response 17), thus had no
impact on the conclusions of the study. Further, (26) False. Figure S14 of Jacobson et al. (2015)
additional simulations with LOADMATCH have shows a zero-load loss, low-cost solution with zero
indicated that it is possible to keep a stable grid for the
hours of demand response. Second, the cost of the
U.S. either using more CSP and without increasing the
high discharge rate required when load builds up
hydropower discharge rate at all or with a
hydropower discharge rate down to 700 GW (rather over time was accounted for in all costs, except
than 1300 GW), without changing total annual for the additional hydropower turbines, which
hydropower energy use. In sum, adding hydropower have since been cost less than 2% of the total
turbines without changing annual hydropower energy energy cost (Response 17).
is only one way to balance load at low cost, not the
Implausible assumptions
only way.
The conclusions contained in ref. [11] rely on a number of
unproven technologies and poorly substantiated assumptions, as
Similarly, as detailed in our Section S1.2, the total amount of detailed in section S2 of our SI Appendix. In summary, the
load labeled as flexible in the figures of ref. [11] is much greater reliability of the 100% wind, solar and hydroelectric power
than the amount of flexible load represented in their supporting system proposed scheme depends centrally on a large installed
tabular data. In fact, the flexible load used by LOADMATCH
is over double the maximum possible value from their Table 1.
The maximum possible from Table 1 from ref. [11] is given as
1,064.16 GW, while Fig. 3 of [11] shows flexible load (in green)
used up to 1,944 GW (on day 912.6). Indeed, in all the figures in
[11] that show flexible load, the restrictions enumerated in their
Table 1 are not satisfied.

(25) False. As clearly stated on the second page of

Jacobson et al. (2015) p. 15,061) Table 1 is an
annual-average load, not a maximum load. As also
2 | Clack et al.
4000 3,797
3500 113x

Installed Capacity (GW)

3000 33x
2,449 inf.
2000 2,000

101 74 14 55 0 23

Hydroelectric Wind Solar Other Hydrogen


2 | Clack et al.


Storage Total Capacity

capacity of several different energy storage systems [11], which Fig. 2. Installed capacity values for 2015 (left column in each pair) and the ref.
collectively allow their model to flexibly reshape energy de- studies [11, 12]. These 100% wind, solar and hydroelectric studies propose installing
mand to match the output of variable electricity generation technologies at a scale equivalent to (or substantially greater than) the entire capacity
technologies. The study [11] assumes a total of 2,604 GW1 of of the existing electricity generation infrastructure. The other category includes coal,
natural gas and nuclear; all of which is removed by 2050.
storage charging capacity, more than double the entire current

2 | Clack et al.

capacity of all power plants in the United States [16]. papers are ready to be implemented today at scale, at
low cost, and that there are no technological or
economical hurdles to the proposed system2.
(27) False. Jacobson et al.s (2015) 100% WWS
The 100% wind, solar and hydroelectric power system
plans require electrifying all energy sectors, and study [11] also makes unsupported assumptions about
all-sector energy is 5 times electric power-sector widespread adoption of hydrogen as an energy carrier,
energy. Thus, the storage capacity proposing is including the con- version of the aviation and steel
only 2/5 of, not two times, the power demand of industries to hydrogen, and the ability to store in
hydrogen an amount of energy equivalent to more than a
the current energy infrastructure. month of current U.S. electricity consumption. Further,
in Figure S6 of ref. [11], hydrogen is being pro-
The energy storage capacity consists almost entirely duced at a peak rate consuming nearly 2,000 GW of
of two technologies that remain unproven at any electricity, nearly twice the current U.S. electricity
significant scale: 514.6 TWh of underground thermal generating capacity.
energy storage (UTES, the largest UTES facility
today is 0.0041 TWh; further discussed in (29) False. As stated on Page 2112 of Jacobson et al.
Section S2.1 of our SI Appendix), and 13.26
terawatt-hours (TWh) of phase-change materials
(2015a), we dont expect to convert fully short-haul
(PCM; effectively, in research and demonstration aircraft until 2035 and long-haul aircraft until 2040.
phase; further discussed in Section S2.2 of Given that a short-haul hydrogen fuel cell aircraft
our SI Appendix) coupled to concentrating that seats four and has a range of 1500 km already
solar thermal power (CSP). To give an idea of
scale, the 100% wind, solar and hydroelectric existed in 2016, these goals are surpassable. With
power system proposed in ref. [11] envisions regard to using hydrogen in industry, our latest U.S.
underground thermal energy storage (UTES) and world studies do not consider that option; opting
systems deployed in nearly every home, business,
office building, hospital, school and factory in
instead to electrify all industry, so the issue is moot.
the United States. The U.S. and world grid stays stable. The production
rate of hydrogen relative to the current U.S.
(28) False. UTES systems are not deployed in electrical demand is irrelevant given that we propose
nearly every home See response (16). Further, to electrify all energy sectors and electricity is
UTES has been demonstrated at the scale it needs currently only 20% of all energy.
to be deployed-neighborhood and complex scale,
As detailed in Section S2.3 of our SI Appendix, the
and it has been tested in more extreme conditions costs and feasibility of this transition to a hydrogen
(Canada seasonally) than it would be needed for economy are not appropriately accounted for by ref.
in the United States. Further, its cost is so low it [11]. To demonstrate the scale of the additional
has already far surpassed more mature capacities that are demanded in ref. [11, 12] we plot
them along with the electricity generation capacity in
technologies. With regard to CSP, molten salt has 2015 in Fig 2. References [11] and [12] cite each other
been used commercially in a number of plants, about the values of capacity in a manner that suggests
and PCM is only marginally better than molten the capacities are the same; however, if one inspects
salt, so even if PCM didnt work well, the fallback Table S2 from [11] and Table 2 from [12] it is obvious
that the values are different for every single technology.
works perfectly fine at only slightly higher cost.
(30) False. Table S2 from Jacobson et al. (2015 = Ref. 11)
Although both PCM and UTES are
promising resources, neither technology has clearly states that the capacity numbers are for the
reached the level of technological ma- turity to CONUS (continental U.S., which is 48 states); Table 2 of
be confidently employed as the main Jacobson et al. (2015b = Ref 12) clearly states in the
underpinning technology in a study aiming to
caption that the capacity numbers are for the 50 U.S.
demonstrate the technical reliability and
feasibility of an energy system. The relative states.
immaturity of these technologies cannot be
reconciled with the authors assertion that the Additionally, ref. [11]
solutions proposed in ref. [11] and companion
energy types.

Table S1 in [11] shows non-UTES storage 1,065 GW; UTES electric storage 1,072 GW; and UTES
thermal storage 467 GW. In ref. [11] there is no description of how LOADMATCH differentiates
Clack et al. PNAS | May 4, 2017 | vol. XXX | no. XX | 3

100% conversions [to WWS energy systems] are technically and
economically feasible with little downside. Numerous low-cost solutions
are found, suggesting that maintaining grid reliability upon 100%
conversion to WWS is economically feasible and not a barrier to the
conversion [to a 100% WWS system]. We do not believe a technical or
economic barrier exists to ramping up production of WWS technologies.
Based on the scientific results presented, current barriers to
implementing the [100% WWS] roadmaps are neither technical nor
economic [12]. Our goal is to get to 80 per cent by 2030 and 100 per cent
by 2050. It is certainly technically and economically practical. Mark
Jacobson, Jan 2016 [14]

Clack et al. PNAS | May 4, 2017 | vol. XXX | no. XX | 3

assumes that 63% of all energy-intensive industrial demand
is flexible, able to reschedule all energy inputs within an 8-
hour window. As discussed in Section S2.4 of our SI
Appendix, the authors of ref. [11] provide no basis for
assuming such flexibility in industrial energy demand.

Fig. 3. Historical and proposed hydroelectric generation per year. The historical data
(31) As shown in Figure S14 of Jacobson et al. ( show generation averaging
(2015), a low-cost, zero-load-loss solution was 280.9 TWh per year; generation proposed in [11] is 402.2 TWh, 13% higher than the 25-year
historical maximum of 356.5 TWh (1997) and 85% higher than the historical minimum of 217 TWh
also obtained for 0 hours of demand response. (2001).
Nevertheless, Clack et al. fail to show why
industrial processes need to run during peak To illustrate the implausibility of the assumed increase in
hydroelectric generation in the face of limited water supply, we plot in Fig.
energy times day and cant be subject to demand 3 the last 25 years of generation from hydropower in the U.S. along with
response. the average for the studies in ref. [11, 12]. Average future generation
assumed by ref. [11, 12] is 13% higher than the highest peak year in the last
25 and 85% higher than the minimum year in the last 25.
Similarly, ref. [11] assumes that the capacity factor
(i.e., actual electricity generation divided by the (32) Figure 3 of Clack et al. falsely shows that the
theoretically maxi- mum potential generation obtained by
operating continuously at full nameplate capacity) for Jacobson et al. (2015) hydropower output in 2050 is
existing energy technologies will increase dramatically in 402.2 TWh. In fact, they forgot to multiply by the
the future. As described in Section S2.5 of our SI transmission and distribution efficiency, a mean of 0.925,
Appendix, the authors of ref. [11] anticipate that individual
hydropower facilities are assumed to increase gener- ation by
so it is really 372.035 TWh, much closer to current
over 30%. They explain this by saying, Increasing the output. Regardless, the current hydro output has
capacity factor is feasible because existing dams currently nothing to do with what is possible.
provide much less than their maximum capacity, primarily
So in addition to needing
due to an oversupply of energy available from fossil fuel
sources, resulting in less demand for hydroelectricity [12]. See Excel spreadsheets from [11] and [12], Tab EIA capacity factors 2011-2075: http://web.
From [12] it is stated that hydroelectric and geothermal
capacity factors increase because For geothermal and
hydropower, which are less variable on short time scales
than wind and solar, the capacity-factor multipliers in our
analysis are slightly greater than 100% on account of these
being used more steadily in a 100% WWS system than in
the base year. In addition to con- tradicting their statement
that hydropower is used only as a last resort [11], this
explanation demonstrates a fundamental misunderstanding
of the operation of electricity markets and the factors
determining hydroelectric supply.

(33) False. It is impossible for (12) to contradict

(11) because Reference (12) was published before
(11) and the result in 11 was a new finding
With near-zero marginal costs (free fuel),
hydroelectric generators will es- sentially run whenever
they are available; in those instances where they
participate in merchant markets they underbid fossil
generators that must at least recover their coal or natural gas
costs. The primary factor limiting hydroelectric capacity
factor is water supply and environmental constraints, not
lack of demand. Further, there appears to be a mistake
with the hydroelectric capacity factor adjustment: from
EIA it should
only go up to 42% not 52.5%3.
Jacobs on et al. propos ed avera ge = 40 2.2 TWh
nual US Hydroelectric Generation (TWh)

Increa se of
375 ~13 %
356.5 TWh
Increas e of Incr ease of
~85% ~43%

4 | Clack et al.

Average o ver 25 yea rs = 28 0.9 TWh

1,300 GW of peak power from 150 GW of the feasibility of the plan set out in ref. [11] for the United
capacity, there also needs to be an extra 120 States. In Section S2.8 of our SI Appendix, we describe how
TWh of hydroelectric generation on top of the ref. [11] assumes that the U.S. will build out new solar, wind
280 TWh available. Further difficulties in and hydro facilities at a sustained rate that, on a per unit
raising hydropower capacity factors are GDP basis, is 16 times greater than the average deployment
described in our SI Appendix, Section 2.5. rate in Germanys Energiewende initiative during the years
2007 to 2014, and over 6 times greater than Germany
Most of the technologies considered in ref.
achieved in the peak year of 2011 (see our Fig. S4).
[11] have high capital costs but relatively low
operating costs. As a result, the cost of capital is In Fig. 4, we display another metric on the scale of ex-
a primary cost driver in the vision contained in pansion. It shows the rate of installation as Watts per Year
ref. [11]. As discussed in Section S2.7 of our SI per Capita. Using this metric, we can compare the scale of
Appendix, the baseline value for cost of capacity expansion in ref. [11] with historic data. Figure 4
capital in ref. [11] is one-half to one-third of shows that the plans proposed in references [11] and [12] would
that used by most other studies. require a sustained installation rate that is over 14 times the
U.S. average over the last 55 years; and over 6 times the peak
rate. For the sake of comparison, Fig. 4 includes the estimated
(34) False. The only relevant studies are those rate for a solution that decarbonizes the U.S. electric grid by
that are recent and among those, Lazard (10.0) 78% by 2030 [1], historical German data and historical Chinese
( data. Sustaining public support for this scale of investment
(and this scale of deployment of new wind turbines, power
cost-of-energy-v100.pdf) is the most detailed and lines, etc.) could prove challenging. One of the reasons this
relied upon by the energy industry, and capital buildout may prove difficult, is that the 100% wind, solar and
costs are consistent with that study and other hydroelectric system relies on energy sources with relatively
low areal power density (see Section S2.9 of our SI Appendix
contemporary studies. To the contrary, the costs
for further details). According to NREL, maximum power
from Clack et al. are based on old data and densities achievable in land-based wind farms is about 3 W/m2
inconsistent with facts on the ground. For (although at larger scale, power densities would likely be lower)
example, such studies pretend as if nuclear is [15].
cost competitive whereas, nuclear advocates
acknowlege that the costs for nuclear are (36) False. The NREL study says nothing of the
sort. Section 4.3 of that reference says that the
prohibitive (Freed et al., 2017).
maximum installed power density is 11.2 W/m2,
not 3 W/m2.
The 100% wind, solar and hydroelectric
energy system studies [11, 12] provide little
Even at these high power densities, the scale of wind power
evidence that the low cost of capital assumed in
envisioned in ref. [11] would require nearly 500,000 km2,
their study could be obtained by real investors
in the capital markets. Using more realistic
discount rates of 6 9% per year instead of the 3 (37) False. Clack et al. used an erroneous installed
4.5% in ref. [11] power density thus miscalculated areas by a factor
of 3.5.
(35) False. The Federal Discount rate on May 5,
2017 was 1.50%; the WSJ Prime rate was 4.0%.
which is roughly 6% of the continental U.S. and >1,500 m2
of land for wind turbines for each American. Much of this
could double the estimate of an 11 land could be dual use, but the challenges associated with this
cents/kWh cost of electricity to 22 cents/kWh, level of scale up should not be underestimated. Many energy
even before adding in the unaccounted-for transition proposals require implausible and unprecedented
capital costs described above. One possible
explanation of the lower discount rates used
could be that they forecast lower growth (or
negative) gross domestic product. In the case
of lower growth, there would likely be lower
interest rates; however that lower growth may
also lead to lower energy demand and
One of the global leaders of solar PV and
wind energy installation in recent years is
Germany, which through its En- ergiewende is
attempting to shift toward an 80% renewables
energy system. Germany, therefore, presents a
suitable exam- ple against which to benchmark
4 | Clack et al.

rates of technology deployment. Most models do not account plant during an assumed planning and construction period for up to 19
for effects of rapid shifts in the economy needed for such large years per nuclear plant4.
scale transitions, and how these shifts might impact energy
prices. For example, increased pressure on materials, elevated (39) False. The emissions are of the background grid,
commodity prices and high demand for wind power instal-
lations produced elevated prices for wind power deployment
not a coal plant. Clack et al. pretends that nuclear does
between 2002 and 2008 [16, 17]. not have planning to operation delays, misleading
PNAS into thinking that Jacobson et al. are stack(ing)
1,200 Historical (US)
the deck when in fact it is Clack et al. who are
100% wind, solar and hydroelectric (US)
misleading the public into thinking nuclear does not
Capacity Additions (W / y / Capita)

1,000 Historical (Germany)

> 6x increase over

80% electricity (US)
Historical (China)
US historical
maximum have opportunity cost emissions or costs associated with
Coal and Nuclear
Peak Germany
nuclear weapons proliferation or meltdown risk
although the international community knows otherwise:
2nd Nuclear Solar PV Peak > 14x increase
Peak over US
Natural Gas
Germany historical
400 Wind Peak average

IPCC Working Group III, Chapter 7

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050
Executive Summary. P. 517. Barriers to and risks
Fig. 4. The historical rate of installed electric generating capacity per Capita associated with an increasing use of nuclear energy
(W/y/Capita) for China (blue), Germany (grey) and the U.S. (black) are shown with
the estimated values for the Jacobson et al. [11, 12] (red) and MacDonald et al. [1]
include operational risks and the associated safety
(green) U.S. proposals. It shows that the 100% wind, solar and hydropower power concerns, uranium mining risks, financial and
plan requires installation of new capacity at a rate more than an order of magnitude
greater than that previously recorded in China, Germany or the United States. The
regulatory risks, unresolved waste management issues,
rate would have to be continued indefinitely due to replacing generation as it aged. nuclear weapons proliferation concerns, and adverse
public opinion (robust evidence, high agreement).
The rejection of many potential sources of low-carbon-
emission energy is based on an analysis presented by Jacobson Added to this, the effects of a nuclear war, which is assumed to
et al. in ref. [18]. A full discussion of that paper is beyond periodically re-occur on a 30-year cycle, is included in the analysis of
the scope of our current evaluation. However, one major flaw emissions and mortality of civilian nuclear power5. In contrast, those
is its failure to rely on the many already published detailed same authors do not consider emissions from coal plants associated with
studies on life-cycle greenhouse gas (GHG) emissions, land- permitting delays for offshore wind power.
use requirements and human mortality of energy production
technologies. Rather than relying on the results of the many
detailed studies available from large international bodies such
(40) False. Jacobson (2009) assumed 2-5 years between
as those surveyed by the Intergovernmental Panel on Climate planning and operation of wind farms. Based on the
Change, ref. [18] presents assessments that in many cases information at the time, that was the best estimate.
differ in method and granularity to produce results that dif-
fer markedly from those generally accepted in scientific and
While there is extensive experience outside of the U.S. with
technical communities.
developing offshore wind resources, very few offshore wind facilities have
been permitted in the U.S. territorial waters. The 100% wind, solar and
(38) False. Reference 18 is consistent with the hydroelectric power system [11] envisions more than 150,000 5-MW
literature, including IPCC, in terms of quantities turbines permitted and built offshore, without delays.
that the literature has to provide. For example,
IPCC estimates the range of lifecycle costs of 5
The five sources cited in ref. [12] give construction time estimates of 5-8 years.
In the almost 60 years of civilian nuclear power (two of the assumed war-cycles), there have been no nuclear
nuclear power as 4-110 g-CO2/kWh, which exchanges. The existence of nuclear weapons does not depend on civil power produc- tion from uranium.

compares well with 9-70 g-CO2/kWh from

Jacobson et al. (2009):

IPCC Working Group III, Chapter 7

Section 7.8.1. P. 540. The ranges of harmonized
lifecycle greenhouse gas emissions reported in the
literature are 4-110 gCO2eq/kWh for nuclear

Selective assessments of life-cycle emissions can be used

to favor or disfavor specific technologies. As an example, the
lifecycle GHG-emissions for nuclear power generation in ref.
[18] include the emissions of an equivalent-sized coal power
Clack et al. PNAS | May 4, 2017 | vol. XXX | no. XX | 5

Insufficient power system modeling failed to consider 100% WWS systems. In addition,
The study of a 100% wind, solar and MacDonald et al. (2016) included nuclear, which entails
hydroelectric power system [11] purports to greater catastrophic and/or health/ water/land risks than
report the results of agrid integra- tion 100% WWS, yet they failed to quantify costs of such risks,
model. It is important to understand the
limitations of the study with regard to what is
pretending as if they dont exist. They further failed to
usually meant by grid integration. Reliable consider the impossibility of planning plus building nuclear
operation of the grid involves a myriad of plants in any reasonable time frame.
challenges beyond just matching total
generation to total load. Its role in cascading all loads, gen- eration (sited precisely where existing
failures and blackouts illustrates the important generation resides) and storage are summed in a single place.
role of the transmission system [19]. Reliable Therefore, those authors do not perform any modeling or
grid operation is further complicated by its AC analysis of transmission. As a result, their analysis ignores
nature, with real and reactive power flows and transmission capacity expansion, power flow, and the logistics
the need to closely maintain a constant frequency of transmission constraints (see our Section S2.6). Similarly,
[20]. Margins for generator failures must be those authors do not account for operating reserves, a
provided through operational and planning fundamental constraint necessary for the electric grid. Indeed,
reserves [21]. The solution proposed by ref. LOADMATCH used in ref. [11] is a simplified representation
[11, 12] involves fundamental shifts in aspects of electric power system operations that does not capture
of grid architecture that are critical to reliable requirements for frequency regulation to ensure operating
operation. Wind generation, largely located far reliability (see further details in Section S3 of our SI
from load centers, will require new Appendix).
transmission. Solar generation and on-site Further, the model is fully deterministic, implying perfect
storage connected to the distribution grid replace foresight about the electricity demand and the variability
capa- bility currently connected to the more- of wind and solar energy resources,
centralized transmission grid. Rotating
machines whose substantial inertia is critical
for frequency stability are supplanted by (42) False. As clearly stated in Section S1.M of
asynchronous wind and solar generators. Jacobson et al. (2015), LOADMATCH simulations
While a grid integration study is detailed and here are similar to those of a pure stochastic
complex, the model of ref. [11] is spatially zero- model Thus, it is not a deterministic model. In
addition, unlike optimization models such as in
(41) False. There are two models used in Jacobson MacDonald et al. (2016), LOADMATCH is entirely
et al (2015), a 3-D global climate-air-pollution- prognostic. It marches forward in time with no
weather forecast model (GATOR-GCMOM) and knowledge of the wind or solar fields or load the
the LOADMATCH grid integration model. Only next time step. This is why it requires trial and
LOADMATCH is spatially 0 dimensional but it error.
uses the aggregate of 3-D wind fields. Despite the
0-dimensional spatial dimension of neglecting the effect of forecast errors on reserve
LOADMATCH, it includes more variables and requirements [22]. In a system where variable renewable
takes shorter times steps (30 s) than any grid resources make up over 95% of U.S. energy supply, renewable
integration model used to study high penetrations energy forecast errors would be a significant source of
uncertainty in the daily operation of power systems. The
of renewables. For example, MacDonald et al. LOADMATCH model does not demonstrate the technical
(2016) took 1-hour times steps, with 120 times ability of the proposed system from ref. [11] to operate reliably
lower resolution that LOADMATCH. It also given the magnitude of the architectural changes to the grid
considered only 3 years rather than 6 years of and the degree of uncertainty imposed by renewable resources.
data in Jacobson et al. (2015). Further, that study
failed to include storage, allowing excess energy to Inadequate scrutiny of input climate model
be unrealistically shed. Further, it failed to The climate model used to generate weather data used by ref.
electrify all energy sectors, looking only at electric [11] has never been adequately evaluated.
power. It also failed to treat realistic placement of
wind turbines or solar (Jacobson, 2016). In (43) False. See Response (23).
addition, it failed to electrify all energy sectors
state by state or country by country, and failed to For example, results from this model have not been made
available to the Climate Model Intercomparison Project [23]
examine changes in demand due to electrification, or been opened to public inspection in ways similar to the
Clack et al. PNAS | May 4, 2017 | vol. XXX | no. XX | 5

results for major reanalysis projects [24]. As
detailed in section S4 of our SI Appendix, the
fragmentary results that have been made
available show poor correlation with reality in
terms of resolution and accuracy. Since the
conclusions from ref. [11] depend on the weather
data used, their conclusions cannot be
considered to be adequate without an
appropriate evaluation of the weather data used.

Clack et al. PNAS | May 4, 2017 | vol. XXX | no. XX | 5

Conclusions supported assumptions, by using simpler models that do not
consider important features, and then performing an analysis
Many previous studies of deep decarbonization of electric
that contains critical mistakes, the anomalous conclusion can-
power illustrate that much can be done with wind and solar
not be heralded as a new discovery. The conclusions reached
power, but that it is extremely difficult to achieve complete
by the study contained within ref. [11] about the performance
decarbonization of the energy system even when employing
and cost of a system of "100% penetration of intermittent
every current technology and tool available, including energy
wind, water and solar for all purposes" are not supported by
efficiency and wind, hydro, and solar energy, but also carbon
adequate and realistic analysis and do not provide a reliable
capture and storage, bioenergy, and nuclear energy [16, 810].
guide to whether, and at what cost, such a transition might
In contrast, ref. [11] asserts that it is cost-effective to fully
be achieved. In contrast, the weight of the evidence strongly
decarbonize the U.S. energy system primarily using just three
suggests that a broad portfolio of energy options will help fa-
inherently variable generating technologies: solar PV, solar
cilitate an affordable transition to a near zero-emission energy
CSP, and wind, which supply more than 95% of total energy
in the proposal presented in ref. [11]. Such an extraordinarily
constrained conclusion demands a standard of proof that ref. Supporting Information (SI) Appendix. The Supporting Informa-
[11] does not meet. tion document contains the details of this evaluation.
The scenarios of ref. [11] can at best be described as a poorly
executed exploration of an interesting hypothesis. The studys SI Appendix Dataset. An Excel file containing all the calculations
numerous shortcomings and errors render it unreliable as a made for this article is available.
guide about the likely cost, technical reliability, or feasibility
Detailed Conflict of Interest and Funding Disclosure. The authors
of a 100% wind solar and hydroelectric power system. It is one
declare no conflict of interest, and with the exception of SQ,
thing to explore the potential use of technologies in a clearly
none received support from sources other than normal salary
caveated hypothetical analysis; it is quite another to claim that
from their employers for work on the preparation of this paper.
a model employing these technologies at an unprecedented
With the exception of MB and JL, all of the authors have
scale conclusively demonstrates the feasibility and reliability
recently received outside support for more general research on
of the modeled energy system.
energy systems and renewable energy. CC received support
From the information given by ref. [11], it is clear that both
in the past from NOAA. SQ was supported for analysis that
hydroelectric power and flexible load have been modeled in
supported this paper from the Rodel Foundation of Delaware
erroneous ways, and these errors alone invalidate the study and
and has received more general faculty funding from Uppsala
its results. The study of 100% wind, solar and hydroelectric
University. JA and GM have received support from NSF,
power systems [11] extrapolate from a few small-scale instal-
EPRI, the Doris Duke Charitable Foundation, and members
lations of relatively immature energy storage technologies
of the Carnegie Mellon Electricity Industry Center. AB has
to assume ubiquitous adoption of high-temperature phase-
received support from the California Air Resources Board,
change materials for storage at concentrating solar power
the Carnegie Endowment for International Peace, Argonne
plants, underground thermal energy storage for heating, cool-
National Laboratory, Sandia National Laboratory, NREL,
ing, and refrigeration at every building in the United States,
Ford Motor Company, and Saudi Aramco. KC has received
and widespread use of hydrogen to fuel airplanes, rail, ship-
support from the Carnegie Institution for Science endowment
ping, and most energy-intensive industrial processes. For the
and the Fund for Innovative Climate and Energy Research.
critical variable characteristics of wind and solar resources,
SD has received support from NSF. VD has received support
they rely on a climate model that has not been independently
from NREL. MH has received support from NSF and DOE.
PH has received support from NSF and DOE. PJ has received
The authors of ref. [11] claim to have demonstrated that support from NSF, EPA and NOAA. DK has received support
their proposed system would be low-cost and that there are no from NSF and the Zaffaroni and Karsten Family Foundations.
economic barriers to the implementation of their vision [12]. AR has received support from NSF. VS has received support
However, the modeling errors described above, the speculative from the Sloan Foundation. JS has received funding from
nature of the TW-scale storage technologies envisioned, the Jay Precourt, Bloom Energy, EPA, ExxonMobil Corporation,
theoretical nature of the solutions proposed to handle critical California Energy Commission, and DOE. GT has received
stability aspects of the system and a number of unsupported support from DOE and UCSD Deep Decarbonization Initiative.
assumptions, including a cost of capital that is a third to a DGV has received support from EPRI, the UC San Diego
half lower than is used in practice in the real world, undermine Deep Decarbonization Initiative, and the Brookings Institution.
that claim. Their LOADMATCH model does not consider JPW has received support from DOE, EPA and industry
aspects of transmission power flow, operating reserves or of affiliates of the Energy Modeling Forum. JFW has received
frequency regulation that would typically be represented in a support from NSF, DOE, DOD, Toyota and Aquion Energy.
grid model aimed at assessing reliability. Further, as detailed
above and in the SI Appendix, a large number of costs and References.
barriers have not been considered in ref. [11]. 1. MacDonald AE, Clack CTM, Alexander A, Dunbar A, Wilczak J, Xie Y (2016) Future cost-
Many researchers have been examining energy system tran- competitive electricity systems and their impact on US CO2 emissions. Nature Climate
Change 6:526531.
sitions for a long time. Previous detailed studies have generally 2. NREL (2012) Renewable Electricity Futures Study (NREL/TP-6A20-52409), (National Renew-
found that energy system transitions are extremely difficult, able Energy Laboratory), Technical report.
and that a broad portfolio of technological options eases that 3. UNDDP (2015) Pathways to deep decarbonization 2015 report (United Nations Deep Decar-
bonization Pathways Project), (SDSN - IDDRI), Technical report.
transition. If one reaches a new conclusion by not addressing 4. EMF (2014) The EMF24 Study on U.S. Technology and Climate Policy Strategies (Energy
factors considered by others, by making a large set of un- Modeling Forum). Special issue of The Energy Journal 35:1229.

6 | Clack et al.

5. Krey V, Luderer G, Clarke L, Kriegler E (2013) Getting from here to there energy technology 17. The Globe and Mail (2016) Can the world convert to total renewable energy by 2050?
transformation pathways in the EMF27 scenarios. Climatic Change 123(3-4):369382. 18. Denholm P, Hand M, Jackson M, Ong S (2009) Land-use requirements of modern wind power
6. Williams JH et al. (2012) The Technology Path to Deep Greenhouse Gas Emissions Cuts by plants in the United States, (National Renewable Energy Laboratory), Technical report.
2050: The Pivotal Role of Electricity. Science 335(6064):5359. 19. Isabel BM (2009) The economics of wind energy. Renewable and Sustainable Energy Re-
7. Mileva A, Johnston J, Nelson JH, Kammen DM (2016) Power system balancing for deep views 13(6-7):13721382.
decarbonization of the electricity sector. Applied Energy 162:10011009. 20. Bolinger M, Wiser R (2011) Understanding Trends in Wind Turbine Prices Over the Past
8. IEA (2015) Energy Technology Perspectives 2015, (International Energy Agency), Technical Decade, (Lawrence Berkeley National Laboratory, Berkeley, CA), Technical report.
report. 21. Jacobson MZ (2009) Review of solutions to global warming, air pollution, and energy security.
9. Energy and Research Partnership (2015) Managing Flexibility Whilst Decarbonising the GB Energy Environ. Sci. 2(2):148173.
Electricity System, Technical report. 22. Dobson I, Carreras BA, Lynch VE, Newman DE (2007) Complex systems analysis of series
10. IPCC and Working and Group and III (2014) Contribution of Working Group III to the Fifth of blackouts: Cascading failure, critical points, and self-organization. Chaos 17(2).
Assessment Report of the Intergovernmental Panel on Climate Change. 23. Backhaus S, Chertkov M (2013) Getting a grip on the electrical grid. Physics Today 66(5):42
11. Jacobson MZ, Delucchi MA, Cameron MA, Frew BA (2015) Low-Cost Solution to the Grid 48.
Reliability Problem with 100% Penetration of Intermittent Wind, Water, and Solar for All Pur- 24. Hirst E, Kirby B (1996) Electric-Power Ancillary Services, (Oak Ridge National Laboratory,
poses. Proceedings of the National Academy of Sciences of the United States of America Oak Ridge, TN), Technical report.
112(49):1506015065. 25. Mauch B, Apt J, Carvalho PMS, Jaramillo P (2013) What day-ahead reserves are needed in
12. Jacobson MZ et al. (2015) 100% clean and renewable wind, water, and sunlight (WWS) electric grids with high levels of wind power? Environ. Res. Lett. 8(3):034013.
all-sector energy roadmaps for the 50 United States. Energy & Environmental Science 26. Taylor KE, Stouffer RJ, Meehl GA (2012) An Overview of CMIP5 and the Experiment Design.
8(7):20932117. Bulletin of the American Meteorological Society 93(4):485498.
13. Energy Information Agency (2014) EIA Electricity Data, Table 4.3. Existing Capacity by En- 27. Dee D, Fasullo J, Shea D, Walsh J, NCAR Staff (2016) The Climate Data Guide Atmospheric
ergy Source, 2014 (Megawatts). Reanalysis Overview & Comparison Tables.

V References

W Clarke LE, et al. (2014) Technology and U.S. emissions

reductions goals: Results of the EMF 24 modeling exercise.
Energy Journal 35: 931.
Y IRENA (International Renewable Energy Agency) (2012a)
Renewable energy technologies: Cost analysis series, Vol.
_technologies_cost_analysis-hydropower.pdf, Accessed
February 10, 2014.
AA Freed, J., T. Allen, T. Nordhaus, and J. Lovering (2017) Is
nuclear too innovative, Third Way,
a14fb4fef41a#.qag59xnk0, Accessed Feb. 28, 2017.
BB Jacobson, M. Z., R. Lu, R. P. Turco, and O. B. Toon,
Development and application of a new air pollution
modeling system. Part I: Gas-phase simulations, Atmos.
Environ., 30B, 19391963, 1996
CC Jacobson, M. Z., Development and application of a new
air pollution modeling system. Part III: Aerosol-phase
simulations, Atmos. Environ., 31A, 587608, 1997
DD Jacobson, M. Z., Studying the effects of aerosols on
vertical photolysis rate coefficient and temperature
profiles over an urban airshed, J. Geophys. Res., 103,
10,593-10,604, 1998
EE Jacobson, M. Z., Isolating nitrated and aromatic aerosols
and nitrated aromatic gases as sources of ultraviolet light
absorption, J. Geophys. Res., 104, 3527-3542, 1999a,
FF Jacobson, M. Z., Effects of soil moisture on temperatures,
winds, and pollutant concentrations in Los Angeles, J.
Appl. Meteorol., 38, 607-616, 1999b.
GG Jacobson, M.Z. (2001a) Global direct radiative forcing due
to multicomponent anthropogenic and natural aerosols, J.
Geophys. Res., 106, 1551-1568.
Clack et al. PNAS | May 4, 2017 | vol. XXX | no. XX | 7

HH Jacobson, M. Z. (2001b) GATOR-GCMM: 2. A 9002, doi:10.1002/2014JD021861, 2014.
study of day- and nighttime ozone layers aloft,SS
ozone in national parks, and weather during TT Jacobson MZ, et al. (2015b) 100% clean and renewable
the SARMAP field campaign, J. Geophys. Res., wind, water, and sunlight (WWS) all-sector energy
106, 5403-5420. roadmaps for the 50 United States. Energy Environ Sci 8:
II Jacobson, M. Z. (2002) Control of fossil-fuel 2093-2117.
particulate black carbon plus organic matter, UU Jacobson MZ, Delucchi MA, Cameron MA, Frew BA
possibly the most effective method of slowing (2015). A low-cost solution to the grid reliability problem
global warming, J. Geophys. Res., 107 (D19), with 100% penetration of intermittent wind, water, and
4410, doi:10.1029/ 2001JD001376. solar for all purposes. Proc. Natl. Acad. Sci 112: 15,060-
JJ Jacobson, M.Z. (2004) The climate response of 15,065.
fossil-fuel and biofuel soot, accounting for VV Jacobson, M.Z. (2016) Clean grids with current
soots feedback to snow and sea ice albedo and technology, Nature Climate Change, 6, 441-442.
emissivity, J. Geophys. Res., 109, D21201, WW MacDonald A.E., C.T.M. Clack, A. Alexander, A.
doi:10.1029/2004JD004945. Dunbar, J. Wilczak, Y. Xie (2016) Future cost competitive
KK Jacobson, M.Z. (2006) Comment on fully electricity systems and their impact on U.S. CO2.
coupled online chemistry within the WRF Emissions. Nature Climate Change, 6, 526-531.
model, by Grell et al. 2005. Atmos. Environ, 39,
XX Ten Hoeve, J.E., M.Z. Jacobson, and L. Remer (2012)
6957-6975, Atmos. Environ. 40, 4646-4648.
Comparing results from a physical model with satellite
LL Jacobson, M.Z., and Y.J. Kaufmann (2006)
and in situ observations to determine whether biomass
Wind reduction by aerosol particles, Geophys.
burning aerosols over the Amazon brighten or burn off
Res. Lett., 33, L24814,
clouds, J. Geophys. Res., 117, D08203,
MM Jacobson, M.Z., Y.J. Kaufmann, Y.
YY Whitt, D.B., J.T. Wilkerson, M.Z. Jacobson, A.D. Naiman,
Rudich, Examining feedbacks of aerosols to
and S.K. Lele (2011) Vertical mixing of commercial
urban climate with a model that treats 3-D
aviation emissions from cruise altitude to the surface,
clouds with aerosol inclusions, J. Geophys. Res.,
Journal of Geophysical Research., 116, D14109,
112, D24205, doi:10.1029/2007JD008922, 2007
NN Jacobson, M.Z., Short-term effects of
ZZ Zhang, Y., Online coupled meteorological and chemistry
controlling fossil-fuel soot, biofuel soot and
models: history, current status, and outlook (2008), Atmos.
gases, and methane on climate, Arctic ice, and
Chem. Phys., 8, 2895-2932.
air pollution health, J. Geophys. Res., 115,
D14209, doi:10.1029/2009JD013795, 2010
OO Jacobson, M.Z., Investigating cloud absorption
effects: Global absorption properties of black
carbon, tar balls, and soil dust in clouds and
aerosols, J. Geophys. Res., 117, D06205, DDD
doi:10.1029/2011JD017218, 2012
PP Jacobson, M.Z., and C.L. Archer, Saturation
wind power potential and its implications for
wind energy, Proc. Nat. Acad. Sci., 109, 15,679-
15,684, doi:10.1073/pnas.1208993109, 2012.
QQ Jacobson, M.Z., C.L. Archer, and W. Kempton,
Taming hurricanes with arrays of offshore
wind turbines, Nature Climate Change, 4, 195-
200, doi: 10.1038/NCLIMATE2120, 2014,
RR Jacobson, M.Z., Effects of biomass burning on
climate, accounting for heat and moisture
fluxes, black and brown carbon, and cloud
absorption effects, J. Geophys. Res., 119, 8980-
Clack et al. PNAS | May 4, 2017 | vol. XXX | no. XX | 7