On choosing problems to work on

This rant is from an email sent to my research group. It seems that some of us have been asking questions like, “What can I do with a climate model that has not already been done?” If this is the question we are asking, then we are asking the wrong question.

IMG_4424_crop

Carnegie postdocs Clara Garcia-Sanchez and Anna Possner using fluid dynamical models to investigate geophysical limits to wind power.

Most people in the world are focused on solving pressing problems (how to provide for their families, how to get access to health care, etc). Most people are faced with pressing problems that they have to solve, not problems they choose to solve.

Some people approach their scientific or technical work choosing to focus on pressing problems (“What can I be doing to most effectively help a transition to a clean energy system?”) but other people approach their work thinking, “I have a hammer; are there any nails around that I might be able to hammer on?” — Or even worse, “Are there any nails around that other people have already whacked at, but that I might be able to give another whack or two?”

If you are not working on a problem that you feel is important and pressing, then you are probably working on the wrong problem. (The reason the problem is important could be for fundamental scientific understanding, and not necessarily utilitarian concern.)

It is important to start with the problem, not the tool.

Once you have identified the problem, then your experience with specific tools might inform how you can most effectively contribute to problem solution, but the starting point should be the problem, not the tool.


An intermediate position is to ask: What are the important problems that this tool could contribute to solving? Realistically, this is where we are with much of our work.

The main point is: If you are having trouble finding important problems to address with the tools you already know how to use, that is probably a sign that it is time to learn to use new tools. (This is why I have been learning about economics and energy system modeling.)

You should not just address ever more arcane and irrelevant problems using the tools you already know how to use.


In short:

The world is replete with pressing problems. If you are not working on at least one of these problems, there is a good chance you are wasting your time and you should be doing something else.

swatch-white_8page-divider-clipart-line-dividers-clipart-superb-decorative-divider-2-cartoon-clipart-collection (1)

If you have recently gotten your PhD or will get your PhD within the next year or two, and are interested in trying to address important problems using new tools or approaches, please apply for a postdoc job in my group.

page-divider-clipart-line-dividers-clipart-superb-decorative-divider-2-cartoon-clipart-collection flip

swatch-white_8

What do you do when a poorly expressed idea is quoted out of context by people with a political agenda?

I woke up this morning to read The Federalist quoting me out of context, putting words in my mouth that I did say but wished I had worded more carefully. For those not familiar with The Federalist, they are a right wing online magazine.

The paragraph in question was:

This opens up the possibility that we could stabilize the climate for affordable amounts of money without changing the entire energy system or changing everyone’s behavior,” Ken Caldeira, a senior scientist at the Carnegie Institution for Science, told The Atlantic.

Here is the full email I sent to Robinson Meyer, writer for The Atlantic:

Rob,

I am no expert in systems costing, but I read the paper as saying that Direct Air Capture of carbon dioxide would cost somewhere in the range of $100 to $250 per ton.

If these costs are real, it is an important result.

If you look at this paper (and this is what I could find quickly on the web)

https://static1.squarespace.com/static/54ff9c5ce4b0a53decccfb4c/t/592bd365414fb5ddd39de548/1496044396189/Guivarch%2C+Rogelj+-+Carbon+prices+2C.pdf

Carbon prices projected for this century look like this for 2 C stabilization from a business-as-usual scenario:

unnamed (1)

If you notice, by the end of the century, these integrated assessment models project carbon prices of many hundreds if not thousands of dollars per ton CO2.

The IPCC estimated that these levels of carbon prices could shave 5% off of global GDP.

The result of David Keith and colleagues suggest that carbon prices could never go above the $100 to $250 range per ton CO2, because it would be economic to capture CO2 from air at that price.

This suggests that the hardest to decarbonize parts of the economy (e.g., steel, cement manufacture, long-distance air travel, etc) might continue just as they are now, and we just pay for CO2 removal.

To put these prices in context, $100 per ton CO2 works out to about $1 per gallon of gasoline. This suggests that a fee of somewhere between $1 and $2.50 per gallon would allow people to drive their ordinary cars, and we could just suck the CO2 out of the atmosphere later.

This opens up the possibility that we could stabilize climate for affordable amounts of money without changing the entire energy system or changing everyone’s behavior.

To give more context, global CO2 emissions is something like 36 GtCO2 per year. If we were to remove all of that with air capture at $100 per tonCO2, that works out to $3.6 trillion dollars.

Depending on how you count things, global GDP is somewhere in the neighborhood of $75 to $110 trillion. So, to remove all of this CO2 would be something like 3 to 5% of global GDP (if the $100 per ton number is right). This puts an upper bound on how expensive it could be to solve the climate problem, because there are lots of ways to reduce emissions for less than $100 per ton.

In any case, it makes it much easier to deal with the hardest to decarbonize parts of the economy.

Again, this is all with the caveat that I am no expert in costing of engineering systems. But, if this paper is correct, the result seems important to me.

Best,
Ken

My colleagues and I have been spending a lot of time thinking about how we are to decarbonize the hardest parts of the energy system to decarbonize. We have a paper in press on this very topic, which we expect out later this month.

My positions are fairly well known. In MIT Technology Review, I wrote in 2015:

It is always going to be easier and cheaper to avoid making a mess than to clean up one we have already made. It is easier to remove carbon dioxide from a smokestack, where the exhaust is 10 percent carbon dioxide, than from the atmosphere, which is 0.04 percent carbon dioxide.

In that piece, I went on to write:

When the Constitution of the United States of America was written, it seemed inconceivable that people would be released from slavery or that women would vote. Just a few years before gay marriage became the law of the land, it would have been impossible to predict such a sweeping change in social attitudes. For us to even have a chance of addressing the climate problem, we’ll need another huge change in public attitudes. It will need to be simply unacceptable to build things with smokestacks or tailpipes that dump waste into the air. This change could happen.

The point with my poorly worded quote was not that we don’t need revolutionary changes in our energy system, but that there are some very hard-to-deal-with sources of CO2 emission, like long-distance aviation, that could be addressed by using hydrocarbon fuels coupled with contemporaneous capture of CO2 by devices like that being investigated by David Keith and colleagues.

As recently as 1 June 2018, I wrote an email to Peter Frumhoff of the Union of Concerned Scientists, urging that organization to put out a statement saying:

Today’s emissions policies should be based on the assumption that most [of] our CO2 emissions will remain in the environment for hundreds of thousands of years. Emissions policies should not be made on the assumption that future generations will clean up our mess using carbon dioxide removal technologies and approaches.

There is a big difference in using direct air capture of CO2 to offset contemporaneous emissions and using direct air capture of CO2 to argue that we can continue emitting CO2 today in the hopes that someone else will clean up our mess in the future.


As a little egomaniacal side note, I would like to point out that Caldeira and Rampino (1990) may be the first paper to point out the approximately 300,000 year time scale for removal of atmospheric CO2 concentration perturbations by silicate rock weathering. This estimate has held up pretty well over the last decades.

Caldeira_Rampino2-1

Clipboard01


What are the lessons learned?

When speaking or writing an email to a journalist, think about how each sentence can be read taken out of context. Even if you trust the journalist to represent your views well (and I think Robinson Meyer did an excellent job), somebody later can take a carelessly worded statement and use it out of context.

Also, we are busy, and when requests come in, we often try to respond with something quickly so we can get back to our day jobs (which in my case happens to be scientific and technical research). I should slow down a little bit and take the time needed to write more careful prose.


So, what do you do when a poorly expressed idea is quoted out of context by people with a political agenda?

My answer: “Write a blog post about it, and then Tweet and move on.”

swatch-white_8

Steps to writing a scientific paper based on model results

My postdocs and I are having a discussion about how to be more efficient in producing high-impact papers in quality peer-reviewed journals. I sent the steps in my preferred process to them, which are repeated below.

CLab animation
Photo by Jess Barker

Steps are similar for the observationally-based work we do. The main difference is that obtaining additional observations is usually much harder than performing additional model simulations.

Steps to writing a scientific paper

1. Play until you stumble on something of interest. Obtain initially promising results. Alternatively, think about what paper people would find useful that you could write but has not yet been written.

2. Write a provisional draft abstract for the proposed paper. This defines the problem, the scope of work, the expected results, and why it is important or interesting. What is the main point of the study and why should anyone care? This is a good time to start thinking about the target journal.

3. Write the introduction of the proposed paper. This forces you to do a literature review and understand what else is out there. It also forces you to write up the problem statement while you still think the problem is important. Usually, by the end of the study, the result seems trivial and obvious, and the problem unimportant.

4. Do additional simulations, measurements, analyses, etc, needed to test out the basic hypothesis and produce data for tables and figures. Attempt to get enough of a mechanistic understanding so that the central result starts to seem trivial and obvious.

5. Create rough drafts of figures. Make an abundance of figures, assuming that some will be in the main paper, some in the supporting material, some for talks, and some not used at all. Make preliminary decision of what figures will be in the main paper.

6. Write first draft of paper around figures. Do this before iterating on figure improvement. The standard outline is: Abstract, Introduction, Methods, Results, Discussion, Conclusions. The Results section should describe the results produced by the model. Usually, the Discussion section should discuss the relevance of those model results to the real world. Sometimes, the exposition is smoother if results in a sequence are each in turn presented and then discussed. This is OK if care is taken to be clear about when you are referring the model and when you are referring to the real world.

7. Write figure captions. Figure captions are often among the parts of the paper read by the broadest audience. Place in figure caption a one sentence statement of the main point you expect the reader to derive from looking at the figure. Sometimes editors pull this sentence out, but they often leave it in. In any case, you should understand the main point of each figure.

8. Iterate improvement of the draft of the paper and main paper figures until the process starts to asymptote. Do additional simulations and make additional figures as necessary. Take care to make your figures beautiful. Beautiful figures not only communicate scientific content well to a broad audience, but also communicate that you care about your work and strive for a high level of excellence. Consider target journal guidelines and what should go in the supporting material and what should be in the main body of the paper.

9. Wherever possible, replace jargon and acronyms with ordinary English. Insofar as it is possible, improve felicity of expression. Write good prose. This is especially important in the abstract, first and last paragraphs, and figure captions.

10. Before submission, double check that the main story of the paper can be obtained by reading (1) the abstract, (2) the first paragraph, (3) the last paragraph, and (4) the figure captions. This is already more than what most ‘readers’ of your paper will actually read. Only experts will read the entire paper. Most readers will just want the idea of the paper and the basic results.

11. Make sure all codes, intermediate data, etc, are packaged up in a single directory. This is done both to facilitate making modifications later, and also so as to provide maximum transparency into and reproducibility of the scientific process.

12. Write cover letter to editor and submit. Stress the new finding and to whom this finding will be of interest. Suggest knowledgeable reviewers who you have not collaborated with recently. If you have written papers on related topics, people who have cited your previous papers would be good candidate reviewers.


Key is to have rough figures and a rough draft on paper early. It is much easier to improve existing text and figures than to start with a blank page.

Also key is recognizing when your manuscript is beginning to asymptote. A sloppy error-filled manuscript will give reviewers the feeling that your work is sloppy. However, perfectionism can mean low productivity. Striking the correct balance is hard.

Another thing is to do Step One 20 times. If you have 20 ideas for papers you can pick the best one. If you have only one idea, it is unlikely to be a great idea. People who have only one idea at a time tend to write papers that are footnotes to their previous papers, and then have careers that descend into meaningless detail that nobody cares about.

You might also want to take a look at this advice on writing scientific papers from George M. Whitesides, and this advice on the 5 most pivotal paragraphs in a scientific paper by Brian McGill.

swatch-white_8

How much ice is melted by each carbon dioxide emission?

I am refining and extending  a back-of-envelope calculation here that I did for an interesting discussion on the Carbon Dioxide Removal google group about Marzeion et al. (2018), which concluded that mountain glaciers contribute about 15 kg of ice melt for each kg of CO2 released.  

F2.large

Figure 2 from Winkelmann et al. (2015) indicating how much Antarctic ice loss is projected to occur as a result of different amounts of cumulative carbon dioxide emission, over the next one, three and ten millennia. Note that 10,000 GtC of cumulative emissions results in about 60 m (about 200 ft) of sea-level rise over the long term (taking additional contributions from Greenland and mountain glaciers into account).

According to the USGS, there 24,064,000 km3 of ice and snow in the world.

According to Winkelmann et al. (2015), it would take about 10,000 GtC to melt (nearly) all of this ice.

If we divide 24,064,000 km3  by 10,000 GtC, assume the density of the ice is 1 kg per liter, and do the appropriate unit conversions, we can conclude that each kg of carbon emitted as CO2 will ultimately melt about 2,400 kg of ice.  This is a huge number.

Another way of expressing this is that each pound of carbon released to the atmosphere as CO2 is likely to end up melting more than a ton of glacial ice.

Often, people like to think in units of tons or kg of CO2 instead of tons or kg of carbon. In these units, each kg of CO2 ultimately melts about 650 kg of glacial ice.


Each American emits on average about 16 tons of CO2 to the atmosphere each year, primarily from the burning of coal, oil and gas, and atmospheric release of the resulting waste CO2.

This works out to about 1.8 kg (about 4 pounds) of CO2 per hour per American. This is more than twice the per capita emission rate of Europe and about twenty times the per capita emission rate for sub-Saharan Africa.

If I am an average American, the CO2 emissions that I produce each year (by participating in the broader economy) will be responsible for melting about 10,000 tons of Antarctic ice, adding about 10,000 cubic meters of fresh water to the volume of the oceans.

That works out to about more than a ton of Antarctic ice loss for each hour of CO2 emissions from an average American. Every minute, we emit enough CO2 to add another five gallons of water to the oceans through glacial ice melt.

If you do the units conversion, this means that each American on average emits enough CO2 every 3 seconds to ultimately add about another liter of water to the oceans. The Europeans emit enough CO2 to add another liter to sea-level rise every 8 seconds, and the sub-Saharan Africans add a liter of seawater’s worth of CO2 emissions every minute.

In my freezer, there is an ice cube tray with 16 smallish ice cubes. The ice cubes in this tray all together had a mass of 345 g, or about 1/3 of a kg. That means that I am responsible for, every second, emitting enough CO2 to melt about an ice-cube-tray’s worth of Antarctica.


Economists often like to think in terms of “carbon-intensity of our economy” meaning how much CO2 to we emit per dollar of value produced or consumed.  We can also think about the “ice-intensity of our economy”: How much ice is melted per dollar of value produced or consumed?

In the United States, per capita GDP is a little less than $60,000 per year.  If our CO2 emissions per capita will ultimately melt about 10,000 tons of ice, that means that, on average, for every $6 we spend in our economy, we are melting another ton of ice.

In the European Union, per capita GDP is a little over $32,000 per year. If you do the math, this works out to a ton of ice of ice ultimately melted for every $8 (7 euros) spent in their economy.

Sub-Saharan Africa has a per capita GDP a little over $1400 per year. Their per capita GDP is about 1/40th of per capita GDP in the US, but their per capita emissions are about 1/20th of ours. This means that on average, for every $3 spent in Sub-Saharan Africa, about one ton of ice will ultimately be melted.


Admittedly, by the time scales of our ordinary activities, ice sheets take a long time to melt. The melting caused by a CO2 emission today will extend out over thousands of years.

There are complex moral questions related to balance short-term and long-term interests. Not everyone thinks we should be taking the long-term melting of Antarctica into account.

However, if the ancient Romans had undergone an industrial revolution similar to ours and fueled a century or two of economic development using fossil-fuels with disposal of the waste CO2 in the atmosphere, sea level today would be rising about 3 cm each year (more than an inch a year) due to the long-term effects of their emissions on the great ice sheets.

If their scientists had told them of the long-term consequences, but they had nevertheless decided to neglect those consequences so that they could be a few percent richer in the short term, I imagine that we would take a fairly dim view of their moral standing.


Post updated 26 March 2018.

swatch-white_8

 

Looking for postdocs wanting to help facilitate a transition to a near-zero emission energy system

This is from an email sent today to colleagues in my department:

Folks,

Postdocs in my lab either have gotten or may be about to get more permanent employment, which puts me in the position of constantly trying to recruit great people.

If you know people who are really good and who are going to get their PhD degrees within a year or two (or have gotten their degree within the past year or two), please forward this email to them.

I really don’t care about people’s domain knowledge. I look to see that they are smart, productive, creative, able to complete projects, can write, can speak, can do math, etc.  Smart people can learn the relevant facts quickly.

We are a good place for people who want to understand the big picture, and who will not get lost investigating interesting but ultimately unimportant detail.

Ability to demonstrate an interest in the challenges associated with a clean energy system transition is important, but experience addressing these challenges is not important.

Two postdocs in my group engaged in geophysical modeling may move on this year, so there is space for at least two people who want to understand limits on and opportunities for clean energy systems from a geophysical perspective.

I am trying to build up our idealized energy-system-modeling effort, so there is room to hire a few people there. There is also room for people who want to do idealized economic analysis related to development and decarbonization.

On a different topic, we have had two Nature papers now which represent the culmination of our ocean acidification-related work on coral reefs in Australia (Albright et al, 2016, 2018). While I am not actively recruiting in this area, if there was a postdoc candidate who has a great idea on how to carry this work forward, and who would want to lead the project, I can make room for such a person.

In short, I would appreciate it if you would use your networks to help me find good people who are interested in topics that my group is interested in. We are open to hiring non-traditional candidates who have interest, but lack experience, in these topic areas.

The job postings can be reached through this link: http://carnegieenergyinnovation.org/index.php/jobs/

Best,
Ken

shaner-fig1a
swatch-white_8

Geophysical constraints on the reliability of solar and wind power in the United States

We recently published a paper that does a very simple analysis of meeting electricity demand using solar and wind generation only, in addition to some form of energy storage. We looked at the relationships between fraction of electricity demand satisfied and the amounts of wind, solar, and electricity storage capacity deployed.

M.R. Shaner, S.J. Davis, N.S. Lewis and K. Caldeira. Geophysical constraints on the reliability of solar and wind power in the United States. Energy & Environmental Science, DOI: 10.1039/C7EE03029K (2018).  (Please email for a copy if you can’t get through the paywall.)

Our main conclusion is that geophysically-forced variability in wind and solar generation means that the amount of electricity demand satisfied using wind and solar resources is fairly linear up to about 80% of annually averaged electricity demand, but that beyond this level of penetration the amount of added wind and solar generation capacity or the amount of electricity storage needed would rise sharply.

Obviously, people have addressed this problem with more complete models. Notable examples are the NREL Renewable Electricity Futures Study and another is the NOAA study (McDonald, Clack et al., 2016). These studies have concluded that it would be possible to eliminate about 80% of emissions from the U.S. electric sector using grid-inter-connected wind and solar power. In contrast, other studies (e.g., Jacobson et al, 2015) have concluded that far deeper penetration of intermittent renewables was feasible.

What is the purpose of writing a paper that uses a toy model to analyze a highly simplified system?

shaner-fig1a

Fig 1b. from Shaner et al. (E&ES, 2018) illustrating variability in wind and solar resources, averaged over the entire contiguous United States based on 36 years of weather data. Also shown is electricity demand for a single year.

The purpose of our paper is to look at fundamental constraints that geophysics places on delivery of energy from intermittent renewable sources.  For some specified amount of demand and specified amount of wind and solar capacity, the gap between energy generation and electricity demand can be calculated. This gap would need to be made up by some combination of (1) other forms of dispatchable power such as natural gas, (2) electricity storage, for example as in batteries or pumped hydro storage, or (3) reducing electricity loads or shifting them in time. This simple geophysically-based calculation makes it clear how big a gap would need to be filled.

Our simulations corresponds to the situation in which their is an ideal and perfect continental scale electricity grid, so we are assuming perfect electricity transmission. We also assume that batteries are 100% efficient. We are considering a spherical cow.

Part of the issue with the more complicated studies is that the models are black boxes, and one has to essentially trust the authors that everything is OK inside of that black box, and that all assumptions have been adequately explained. [Note that Clack et al. (2015) do describe the model and assumptions used in McDonald, Clack et al. (2016) in detail, and that the NREL study also contains substantial methodological detail.]

In contrast, because we are using a toy model, we can include the entire source code for our toy model in the Supplemental Information to our paper. And all of our input data is from publicly available sources. So you don’t have to trust us. You can look at our code and see what we did. If you don’t like our assumptions, modify the assumptions in our code and explore for yourself. (If you want the time series data that we used, please feel free to request them from me.)

Our key results are summarized in our Fig. 3:

Shaner-Fig3

Figure 3 | Changes in the amount of demand met as a function of energy storage capacity (0-32 days) and generation.

The two columns of Fig. 3 show the same data: the left column is on linear scales; the right column has a log scale on the horizontal axis. [In a wind/solar/storage-only system, meeting 99.9% of demand is equivalent to about 8.76 hours of blackout per year, and 99.99% is equivalent to about 53 minutes of blackout per year.]

The left column of Fig. 3 shows, for various mixes of wind and solar, that the fraction of electricity demand that is met by introducing intermittent renewables at first goes up linearly — if you increase the amount of solar and/or wind power by 10%, the amount of generation goes up by about 10%, and is relatively insensitive to assumptions about electricity storage.

From the right column of Fig. 3, it can be seen that that as the fraction of electricity demand satisfied by solar and/or wind exceeds about 80%, then the the amount of generation  and/or the amount of electricity storage required increases sharply. It should be noted that even in the cases in which 80% of electricity is supplied by intermittent renewables on the annual average, there are still times when wind and solar is providing very little power, and if blackouts are to be avoided, the gap-filling dispatchable electricity service must be sized nearly as large as the entire electricity system.

This ‘consider a spherical cow’ approach shows that satisfying nearly all electricity demand with wind and solar (and electricity storage) will be extremely difficult given the variability and intermittency in wind and solar resources.

On the other hand, if we could get enough energy storage (or its equivalent in load shifting) to satisfy several weeks of total U.S. electricity demand, then mixes of wind and solar might do a great job of meeting all U.S. electricity demand. [Look at the dark green lines in the three middle panels in the right column of Fig. 3.] This is more-or-less the solution that  Jacobson et al. (2015) got for the electric sector in that work.

Our study, using very simple models and a very transparent approach, is broadly consistent the findings of  the NREL, NOAA, and  Jacobson et al. (2015) studies, which were done using much more comprehensive, but less transparent, models. Our results also suggest that a main difference in conclusions between the NREL and NOAA studies and the Jacobson et al. (2015) study is that Jacobson et al. (2015) assume the availability of large amounts of energy storage, and that this is a primary factor differentiating these works. (The NOAA study showed that one could reduce emissions from the electric sector by 80% with wind and solar and without storage if sufficient back-up power was available from natural gas or some other dispatchable electricity generator.)

All of these studies share common ground. They all indicate that lots more wind and solar power could be deployed today and this would reduce greenhouse gas emissions. Controversies about how to handle the end game should not overly influence our opening moves.

There are still questions regarding whether future near-zero emission energy systems will be based on centralized dispatchable (e.g., nuclear and fossil with CCS) or distributed intermittent (e.g., wind and solar) electricity generation. Nevertheless, the climate problem is serious enough that for now we might want to consider an ‘all of the above’ strategy, and deploy as fast as we can the most economically efficient and environmentally acceptable energy generation technologies that are available today.

swatch-white_8

MZJ Hydro Explainer

If energy storage is abundant, then that storage can fill the gap between intermittent electricity generation (wind and solar) and variable electricity demand. Jacobson et al. (PNAS, 2015) filled this gap, in part, by assuming that huge amounts of hydropower would be available.

The realism of these energy storage assumption was questioned by Clack et al. (PNAS, 2017), but Clack et al. (PNAS, 2017) went further and asserted that Jacobson et al. (PNAS, 2015) contained modeling errors. A key issue centers on the capacity of hydroelectric plants. The huge amount of hydro capacity used by Jacobson et al. (PNAS, 2015) is necessary to achieve their result, yet seems inconsistent with the information provided in their tables.

Clack et al. (PNAS, 2017) in their Fig. 1, reproduced Fig. 4b from Jacobson et al. (2015), over a caption containing the following text:

F1.large

This figure (figure 4B from ref. 11) shows hydropower supply rates peaking at nearly 1,300 GW, despite the fact that the proposal calls for less than 150 GW hydropower capacity. This discrepancy indicates a major error in their analysis. 

(A dispatch of 1 TWh/hr is equivalent to dispatch at the rate of 1000 GW.)

Since the publication of Clack et al. (PNAS, 2017), Jacobson has asserted the apparent inconsistency between what is shown in Fig. 4b of Jacobson et al. (PNAS, 2015) and the numbers appearing in their text and tables was in fact intentional, and thus no error was made. Mark Z. Jacobson went so far as to claim that the statement that there was a major error in the analysis constituted an act of defamation that should be adjudicated in a court of law.

The litigious activities of Mark Z. Jacobson (hereafter, MZJ) have made people wary of openly criticizing his work.

I was sent a Powerpoint presentation looking into the claims of Jacobson et al. (PNAS, 2015) with respect to this hydropower question, but the sender was fearful of retribution should this be published with full attribution. I said I would  take the work and edit it to my liking and publish it here as a blog post, if the primary author would agree. The primary author wishes to remain anonymous.

I would like to stress here that this hydro question is not a nit-picking side-point. In the Jacobson et al. (PNAS, 2015) work, they needed the huge amount of dispatchable power represented by this dramatic expansion of hydro capacity to fill the gap between intermittent renewable electricity generation and variable electricity demand.


In the text below, Jacobson et al. (E&ES, 2015) refers to:

Jacobson MZ, et al. (2015) 100% clean and renewable wind, water, and sunlight (WWS) all-sector energy roadmaps for the 50 United States. Energy Environ Sci 8:2093–2117.

Jacobson et al (PNAS, 2015) refers to:

Jacobson MZ, Delucchi MA, Cameron MA, Frew BA (2015) Low-cost solution to the grid reliability problem with 100% penetration of intermittent wind, water, and solar for all purposes. Proc Natl Acad Sci USA 112:15060–15065.

and Clack et al (PNAS, 2017) refers to:

Clack, C.T. M, Qvist, S. A., Apt, J., Bazilian, M., Brandt, A. R., Caldeira, K., Davis, S. J., Diakov, V., Handschy, M. A., Hines, P. D. H., Jaramillo, P., Kammen, D. M., Long, J. C. S., Morgan, M. G., Reed, A., Sivaram, V., Sweeney, J., Tynan, G. R., Victor, D. G., Weyant, J. P., Whitacre, J. F. Evaluation of a proposal for reliable low-cost grid power with 100% wind, water, and solar. Proc Natl Acad Sci USA  DOI: 10.1073/pnas.1610381114.



Jacobson et al. (E&ES, 2015) serves as the primary basis of the capacity numbers in Jacobson et al. (PNAS, 2015)

May 25, 2015: Mark Z. Jacobson et al. publish paper in Energy & Environmental Science (hereafter E&ES), providing a “roadmap” for the United States to achieve 100% of energy supply from “wind, water, and sunlight (WWS).”

Picture1

To demonstrate that the roadmaps in Jacobson et al. paper (E&ES, 2015) can reliably match energy supply and demand at all times, that study cites forthcoming study (Ref. 2) that uses “grid integration model” .

Picture6Picture7

Ref. 2 is the at-that-point-forthcoming PNAS paper, “A low-cost solution to the grid reliability problem with 100% penetration of intermittent wind, water, and solar for all purposes” “in review” at PNAS.

This establishes the link between the two papers:

(1) The E&ES paper provides the “roadmap” describing the mix of renewable energy resources needed to supply the US;
(2) The PNAS paper then attempts to demonstrate the operational reliability of this mix of resources.




Jacobson et al. (E&ES, 2015) makes it clear that ‘capacity’ refers to ‘name-plate capacity’

Table 2 of the E&ES paper explicitly describes the “rated power” and “name-plate capacity” of all renewable energy and energy storage devices installed in the 100% WWS roadmap for the United States. Both of these terms refer to the maximum instantaneous power that a power plant can produce at any given moment. These are not descriptions of average output, and nowhere in the table’s lengthy description does Jacobson et al. (E&ES, 2015) claim that hydroelectric power is described differently in this table than the other resources.

The table states that the total nameplate capacity or maximum rated power output of hydroelectric generators in Jacobson et al. (E&ES, 2015) is 91,650 megawatts (MW). In addition, column 5 states that 95.87% of this final installed capacity is already installed in 2013. Only 3 additional new hydroelectric plants at a size of 1,300 MW each, for a total addition of 3,900 MW over existing hydroelectric capacity are included in Jacobson et al. (E&ES, 2015) .

Picture2

Picture3




Jacobson et al. (E&ES, 2015) describes hydro capacity assumptions in some detail

Section 5.4 of the E&ES paper provides additional textual description of the WWS roadmap’s assumptions regarding hydroelectric capacity.

The text states that the total existing hydroelectric power capacity assumed in the WWS roadmap is 87.86 gigawatts (GW; note 1 GW = 1,000 MW).

It further states that only three new dams in Alaska with a total capacity of 3.8 GW are included in the final hydroelectric capacity in the WWS roadmap.

Note that throughout this text, Jacobson et al. (E&ES, 2015) distinguish between “delivered power,” a measure of average annual power generation, and “total capacity,” a measure of maximum instantaneous power production capability. It is this later “total capacity” figure that matches the “nameplate capacity” in Table 2 of 87.86 GW in the 100% WWS Roadmap for 2050.

The text explicitly states that the average delivered power from hydroelectric generators is 47.84 GW on average in 2050.

In Jacobson et al. (E&ES, 2015), the authors state both the maximum power production capability from hydroelectric power assumed in the WWS roadmap and distinguish this from the separately reported average delivered power from these facilities over the course of a year.

Picture4Picture5

Most of the capacity numbers appearing in Jacobson et al. (2015) come from the US Energy Administration. They define what is meant by capacity as represented by their numbers:

Generator nameplate capacity (installed):  The maximum rated output of a generator, prime mover, or other electric power production equipment under specific conditions designated by the manufacturer. Installed generator nameplate capacity is commonly expressed in megawatts (MW) and is usually indicated on a nameplate physically attached to the generator.

Generator capacity: The maximum output, commonly expressed in megawatts (MW), that generating equipment can supply to system load, adjusted for ambient conditions.

The remainder of Section 5.4. discusses several possible ways in which additional hydroelectric power capacity could be added in the United States without additional environmental impact, if it is not possible to increase the average power production from existing hydroelectric dams as Jacobson et al. (E&ES, 2015) assume is possible.

Picture8Picture9

This text describes the potential to add power generation turbines to existing unpowered dams and cites a reference estimating a maximum of 12 GW of additional such capacity possible in the continguous 48 states.

The text also describes the potential for new low-power and small hydroelectric dams, citing a reference that estimates that 30-100 GW of average delivered power—or roughly 60-200 GW of total maximum power capacity at Jacobson et al.‘s (E&ES, 2015) assumed average production of 52.5% of maximum power for each hydroelectric generator.

Nowhere in this lengthy discussion of the total hydroelectric capacity assumed in the WWS roadmap and additional possible sources of hydroelectric capacity does Jacobson et al. (E&ES, 2015) mention the possibility of adding over 1,000 GW of additional generating capacity to existing dams by adding new turbines.

The May 2015 E&ES paper by MZJ et al. explicitly states that the maximum possible instantaneous power production capacity of hydroelectric generators in the 100% WWS roadmap for the 50 U.S. states is 91.65 GW.

Jacobson et al. (E&ES, 2015) also explicitly distinguishes maximum power capacity from average delivered power in several instances. The later is reported as 47.84 GW on average in 2050 for the 50 U.S. states.

Additionally, the authors explicitly state that 3.8 GW of the total hydro capacity in the 50 state WWS roadmap comes from new dams in Alaska. This is in addition to 0.438 GW of existing hydro capacity in Alaska and Hawaii as reported in the paper’s Fig. 4. This is important to note, because Alaska and Hawaii are excluded from the simulations in Jacobson et al. (PNAS, 2015).

The E&ES companion paper to the Jacobson et al. (PNAS, 2015) therefore explicitly establishes that the maximum possible power capacity that could be included in the PNAS paper in the contiguous 48 U.S. states is 87.412 GW (e.g. 91.65 GW in the 100% WWS roadmap for the 50 US states, less 3.8 GW of new hydropower dams in Alaska and 0.438 GW of existing hydro capacity in Alaska & Hawaii).



Summary of key relevant facts about Jacobson et al. (E&ES, 2015)

In summary, the May 2015 Jacobson et al. (E&ES, 2015) paper establishes several facts:

  1. The E&ES paper explicitly states that the maximum possible instantaneous power production capacity of hydroelectric generators in the 100% WWS roadmap for the 50 U.S. states is 91.65 GW (inclusive of imported hydroelectric power from Canada).
  2. The E&ES paper also explicitly distinguishes maximum power capacity from average delivered power. The later is reported as 47.84 GW on average in 2050 for the 50 U.S. states.
  3. The E&ES paper explicitly states that 3.8 GW of the total hydropower capacity in the 50 state WWS roadmap comes from new dams in Alaska and reports that existing capacity in Alaska and Hawaii totals 0.438 GW. This is relevant, because Alaska and Hawaii are excluded from the simulations in the Jacobson et al. (PNAS, 2015) which focuses on the contiguous 48 U.S. states.
  4. The E&ES companion paper to Jacobson et al. (PNAS, 2015) therefore explicitly establishes that the maximum possible power capacity that could be included in the PNAS paper in the contiguous 48 U.S. states is no more than 87.412 GW.
  5. No where in Jacobson et al. (E&ES, 2015) do the authors discuss or contemplate adding more than 1,000 GW of generating capacity to existing hydropower facilities by adding new turbines and penstocks. In contrast, the paper explicitly discusses several other possible ways to add a much more modest capacity of no more than 200 GW of generating capacity by constructing new low-power and small hydroelectric dams.
  6. Jacobson et al. (E&ES, 2015) establishes that Jacobson et al. (PNAS, 2015) is a companion to this E&ES paper and that the purpose of the PNAS paper is to confirm that the total installed capacity of renewable energy generators and energy storage devices described in the 100% WWS roadmap contained in the E&ES paper can reliably match total energy production and total energy demand at all times. The total installed capacities for each resource, including hydroelectric generation, described in the E&ES paper, therefore form the basis for the assumed maximum generating capacities in the PNAS paper.


Jacobson et al. (PNAS, 2015) relies on hydro capacity numbers from Jacobson et al. (E&ES, 2015)

December 8, 2015: The paper “Low-cost solution to the grid reliability problem with 100% penetration of intermittent wind, water, and solar for all purposes” by Jacobson et al. (PNAS, 2015) is published in PNAS as the companion to the May 2015 Jacobson et al. (E&ES, 2015) paper.

Picture10
Picture11

Jacobson et al. (PNAS, 2015) describes existing (year-2010) hydro capacity to be 87.86 GW.
Picture4

The text further establishes that the installed capacities for each generator type for the continental United States (abbreviated “CONUS” in the text) are based on ref. 22, which is Jacobson et al. (E&ES, 2015).

Picture13

“Installed capacity” is a term of art referring to maximum possible power production, not average generation. The paper’s description of “Materials and Methods” states the the “installed capacities” of each renewable generator type are described in the Supplemental Information Table S2 of Jacobson et al. (PNAS, 2015).

Table S2 of the Supplemental Information for the PNAS paper explicitly states the “installed capacity” or maximum possible power generation of each resource type in the Continental United States used in the study.

The explanatory text for this paper again establishes that all installed capacities for all resources except solar thermal and concentrating solar power (abbreviated “CSP” in the text) are taken from Jacobson et al. (E&ES, 2015), adjusted to exclude Hawaii and Alaska. Jacobson et al. (E&ES, 2015) is ref. 4 in the Supplemental Information for Jacobson et al. (PNAS, 2015).

Picture14

Reference 4  (E&ES, 2015).

Picture15

Total installed hydroelectric capacity in Table S2 of Jacobson et al. (PNAS, 2015) is stated as 87.48 GW. This is close to the 87.412 GW of total nameplate power capacity of hydroelectric generators in the 50 U.S. states roadmap, less the new hydro dams in Alaska and existing hydropower capacity in Alaska and Hawaii.

Footnote 4 notes that hydro is limited by ‘annual power supply’ but does not mention that instantaneous generation of electricity is also limited by hydro capacity:

Picture18

Additionally, columns 5 & 6 of Table S2 separately state the “rated capacity” per device and the total number of existing and new devices in 2050 for each resource.

“Rated capacity” is a term of art referring to the maximum possible instantaneous power production for a power plant.

The rated capacity for each hydroelectric device or facility is stated as 1,300 MW and the total number of hydroelectric devices is stated as 67.3. This yields exactly 87,480 MW or 87.48 GW, the installed capacity reported for hydroelectric power in column 3. This provides further corroboration that the 87.48 GW of installed capacity reported refers to maximum rated power generation capabilities of all hydroelectric generators in the simulation, not their average generating capacity as MZJ asserts.

Nowhere in this table, its explanatory text in the Supplemental Information, or the main text of the PNAS paper do the authors establish that they assume more than 1,000 GW of additional hydroelectric generating turbines to existing hydroelectric facilities, as MZJ will later assert.

In contrast, the table establishes that the authors assume that total installed hydroelectric capacity in the Continential United States is assumed to increase from 87.42 GW in 2013 to 87.48 GW in 2050, or an increase of only 0.06 GW or 60 MW.



The hydro power capacity represented in the Jacobson et al. (PNAS, 2015) tables is inconsistent with the amount of hydro capacity used in their simulations

Despite explicitly stating that the maximum rated capacity for all hydropower generators in the PNAS paper’s WWS system for the 48 continental United States is 87.48 GW, Fig. 4 of  Jacobson et al. (PNAS, 2015) shows hydropower facilities generating more than 1,000 GW of power output sustained over several hours on the depicted days.

Picture20

Examination of the detailed LOADMATCH simulation results (available from MZJ upon request) reveals that the maximum instantaneous power generation from hydropower facilities in the simulations performed for Jacobson et al. (PNAS, 2015) is 1,348 GW, or 1,260.5 GW more (about 15 times more) than the maximum rated capacity reported in Table S2.

It is therefore clear that the LOADMATCH model does not constrain maximum generation from hydropower facilities to the 87.48 GW of maximum rated power capacity stated in Table S2.

(Note that hydropower facilities also dispatch at 0 GW for many hours of the simulation. It therefore appears that the LOADMATCH model neither applies a maximum generation constraint of 87.48 GW or any kind of plausible minimum generation constraint for hydropower facilities.)



Summary of key facts related to hydro capacity in Jacobson et al. (PNAS, 2015)

In summary, the December 8, 2015 PNAS paper establishes the following facts:

  1. The installed capacity used in the simulations in Jacobson et al. (PNAS, 2015) is reported in Table S2 of the Supplemental Information for that paper. The total installed hydroelectric capacity or maximum possible power generation reported in Table S2 is stated as 87.48 GW.
  2. This maximum capacity figure is also separately corroborated by taking the rated power generating capacity per device and total number of devices reported in Table S2, which also yields a maximum rated power production from all hydroelectric generators of 87.48 GW.
  3. Table S2 states the the authors only assume 0.06 GW of additional hydroelectric power capacity is added between 2013 and 2050.
  4. Nowhere in the text of Jacobson et al. (PNAS, 2015), its Supplemental Information document, or the explanatory text for Table S2 do the authors state that the term “installed capacity” or “rated capacity per device” for each resource reported in the table is used in any other way than the standard terms of art indicating maximum power generation capability. Nor do the authors establish that total installed capacity of hydroelectric generation is described differently in this table than the other resources and refers instead to average annual delivered power as MZJ claims.
  5. Jacobson et al. (PNAS, 2015) also references and uses Jacobson et al. (E&ES, 2015) to establish the installed power generating capacity of each resource in the simulations performed in the PNAS paper, with the explicit exception of solar thermal and concentrating solar power. The maximum rated power from hydroelectric generation reported in Table S2 of 87.48 GW is consistent (within 68 MW) with the 87.412 GW of name-plate generating capacity reported in the E&ES paper for the 50 U.S. states less three new hydropower dams in Alaska and existing hydro capacity in Alaska and Hawaii reported in the E&ES paper.Recall also that the average delivered power from hydroelectric generators was explicitly and separately stated in the E&ES paper as 47.84 GW for the 50 U.S. states, and is therefore no more than 47 GW for the 48 Continential US states. The reported “installed capacities” for hydroelectric generation in PNAS Table S2 is therefore entirely consistent with the “name-plate capacity” reported in the E&ES paper and is not consistent with the average delivered power from hydroelectric generation reported in the E&ES paper.
  6. Despite establishing a maximum rated power capacity of 87.48 GW, the simulations performed for Jacobson et al. (PNAS, 2015) dispatch hydropower at as much as 1,348 GW, or 1,260.5 GW more than the maximum rated capacity reported in Table S2.



Conclusions

Given available information in the published papers, a reasonable reader should interpret the “installed capacity” or “rated capacity” figures explicitly reported in Table S2 of the Jacobson et al. (2015) paper as referring to maximum generating capacity, because that is the definition used by the studies reported on in the table.

This assertion that the 1,348 GW of maximum hydro generation used in the LOADMATCH simulations for the PNAS paper constitutes an intentional but entirely unstated assumption rather than a modeling error (e.g. a failure to impose a suitable capacity constraint on maximum hydro generation in each time period) is, as we understand it, the primary basis for MZJ’s lawsuit alleging that Christopher Clack and the National Academies of Sciences (publishers of PNAS) intentionally misrepresented his work and thus defamed his person.

A reading of the E&ES and PNAS papers establishes that the MZJ et al. did not omit explicit description of the total rated power capacity of hydroelectric facilities. In point of fact, the authors establish in multiple ways that the maximum power capacity for hydroelectric facilities in the PNAS WWS study for the 48 continental United States is 87.48 GW, not the 1,348 GW actually dispatched by the LOADMATCH model.

Thus, information in the E&ES and PNAS papers do not appear to be consistent with MZJ’s assertions that he and his coauthors had intentionally meant to add more than 1,000 GW of generating capacity to existing hydropower facilities in their model.  (It is outside the scope of this analysis to discuss the plausibility of adding more than 1,000 GW of hydro capacity to existing dams.) Nor does the available evidence indicate that they intentionally assumed more than 1,000 GW of additional hydro capacity and then simply failed to disclose this assumption at any point in either of the two papers.  Such failure to explicitly describe such a large and substantively important assumption to readers and peer reviewers might itself constitute a breach of academic standards.

The operation of the LOADMATCH model is inconsistent with the maximum power generating capacity of hydropower facilities explicitly stated in Jacobson et al. (PNAS, 2015) and in the companion paper, Jacobson et al. (E&ES, 2015) upon which the generating capacities are based. Whether you call failure to impose a suitable capacity constraint on maximum hydro generation in each time period a “modeling error” is up to you, but that would seem to be an entirely reasonable interpretation based on the available facts.

swatch-white_8