Summary: A new book by a Japanese climate scientist explains why the predictions of climate models are useful but insufficiently reliable for large scale public policy action. None of this is new information. But none of it is widely known.
Excerpts from Confessions of a climate scientist
By Mototaka Nakamura.
From the English intro and summary in the Kindle version of his Japanese book.
This excerpt shows some of his findings and explanations.
Read the e-book to see his detailed and technical analysis, and his references.
Climate simulation models are fine tools to study the climate system, so long as the users are aware of the limitations of the models and exercise caution in designing experiments and interpreting their output. In this sense, experiments to study the response of simplified climate systems, such as those generated by the “state-of-the-art” climate simulation models, to major increases in atmospheric carbon dioxide or other greenhouse gases are also interesting and meaningful academic projects that are certainly worth pursuing. So long as the results of such projects are presented with disclaimers that unambiguously state the extent to which the results can be compared with the real world, I would not have any problem with such projects. The models just become useless pieces of junk or worse (worse, in a sense that they can produce gravely misleading output) only when they are used for climate forecasting.
All climate simulation models have many details that become fatal flaws when they are used as climate forecasting tools, especially for mid- to long-term (several years and longer) climate variations and changes. These models completely lack some of critically important climate processes and feedbacks, and represent some other critically important climate processes and feedbacks in grossly distorted manners to the extent that makes these models totally useless for any meaningful climate predictions. …
Measuring global temperature.
A quasi-global observation system has been operating only for 50 years or so, since the advent of artificial satellite observation. Temperature data before then were collected over extremely small (with respect to the earth’s entire surface area) areas and, thus, have severe spatial bias. We have an inadequate amount of data to calculate the global mean surface temperature trend for the pre-satellite period. …
The Sun.
It has been only several decades since we acquired an ability to accurately monitor the incoming solar energy. In these several decades only, it has varied by 1 to 2 Watts per square meters. Is it reasonable to assume that it will not vary any more than that in the next hundred years or longer for forecasting purposes?
Editor’s note: The Sun emits 1,361 watts per sq. meter. Over the 11-year solar cycle, it varies by roughly 0.1%. The IPCC’s AR5 WGI says that estimates of radiative forcing from 1750 to present have medium agreement, medium evidence, and medium confidence – although confidence is higher for the past 3 centuries. (8.4.1.2). As NASA says, “Even fluctuations at just a tenth of a percent can affect Earth.” There is as yet limited data about variation over longer time periods. From another NASA article …
“Over the past century, Earth’s average temperature has increased by approximately 0.6°C (1.1°F). Solar heating accounts for about 0.15°C, or 25%, of this change, according to computer modeling results published by NASA Goddard Institute for Space Studies researcher David Rind in 2004.”
{The} first of the two problematic details of climate simulation models mentioned earlier: erroneous representation of actions of oceanic motions …{I}t can be remedied only by increasing the resolution of climate simulation models from the typical 1˚ x 1˚ or lower, to 0.1˚ x 0.1˚ or higher in longitude and latitude. It is simply an issue of limited computer resources and is not an issue of our limited knowledge of the ocean dynamics and thermodynamics. …
Needless to say, it is absolutely vital for any meaningful climate prediction to be made with a reasonably accurate representation of the state and actions of the oceans. …
Water vapor.
Another major contributor to the predicted major global warming is water vapor, the most important greenhouse gas in the Earth’s atmosphere. …The enhanced warming effect of its changes predicted by the climate simulation models also dwarfs that of the projected carbon dioxide increase. So, predicting changes in the radiative forcing associated with the atmospheric water vapor accurately is essential for any meaningful prediction of climate changes. But the fact is this: all climate simulation models perform poorly in reproducing the atmospheric water vapor and its radiative forcing observed in the current climate. …
Clouds.
Clouds’ role in the global climate is extremely important and extremely complex, to say the least. Ad hoc representations of clouds in climate models may be the greatest source of uncertainty in climate prediction. A profound fact is that only a very small change, so small that it cannot be measured accurately with the currently available observational devices, in the global cloud characteristics can completely offset the warming effect of the doubled atmospheric carbon dioxide. …
Clouds are represented with parametric methods in climate models. Are those methods reasonably accurate? No. If one seriously studies properties of clouds and processes involved in cloud formation and dissipation, and compare them with the cloud treatment in climate models, one would most likely be flabbergasted by the perfunctory treatment of clouds in the models. The parametric representations of clouds are ad hoc and are tuned to produce the average cloud cover that somewhat resembles that seen in the current climate. Can we, or should we, expect them to simulate the cloud coverage and properties in the “doubled atmospheric carbon dioxide” scenario with reasonable accuracy? No. …
Parameterization.
Ed. note: See Wikipedia for a description.
I hate to say this, because I know well how much of serious efforts have been put into improving these parametric representations (I spent hundreds of hours in vain myself), but all of these parametric representations, even the best of them, are Mickey Mouse mockeries when compared with the reality. …That is, the selection of the parameter values is an engineering process to “make the model work” rather than a scientific process. The models are “tuned” by tinkering around with values of various parameters until the best compromise is obtained. I used to do it myself. It is a necessary and unavoidable procedure and is not a problem so long as the user is aware of its ramifications and is honest about it. But it is a serious and fatal flaw if it is used for climate forecasting/prediction purposes. …
Even if the best compromise so obtained from the tuning looks very close to the observation, the models’ behaviors are guaranteed to be grotesquely unrealistic, since the tuning requires other aspects of the models to be extremely distorted in order to counterbalance the distortion associated with the Mickey Mouse representations described above. …
The models use various parametric representations that estimate the water vapor profiles from the large-scale atmospheric state that can be calculated by the models. All but one of these parametric representations are ad hoc and rely on major simplifying assumptions that are not justifiable when scrutinized against the reality. They have only a few parameters that can be used to “tune” the performance of the models and utterly unrealistic. …
With values of parameters that are supposed to represent many complex processes being held constant, many nonlinear processes in the real climate system are absent or grossly distorted in the models. It is delusion to believe that simulation models that lack important nonlinear processes in the real climate system can predict at least the sense or direction of the climate change correctly. …
Conclusions.
The real or realistically-simulated climate system is far more complex than an absurdly simple system simulated by the toys that have been used for climate predictions to date …{N}one of the climate simulation models used for predictions can reproduce the current climate accurately despite the heavy tuning and engineering efforts by climate researchers.
————— End of excerpt. —————
Buy your own copy
It is only $0.99, well-written, and well worth your time.
Editor’s note
Sadly, as Nakamura notes, all of this has been discussed by climate scientists for a decade or more. In an amazing display of power, this has been hidden from the general public. I still see people saying that models are built from first principles of physics – and are therefore as reliable as F=MA.
Those who get their information from journalists or activists will see Nakamura’s analysis as astonishing or impossible to believe. It is, however, roughly consistent with the findings of the WGI in the IPCC’s Fifth Assessment Report. Climate models are a pyramid of assumptions and estimates. AR5 describes many of these as having low or medium confidence. It creates a rickety structure.
So how reliable are long-term forecasts by climate models? Nobody knows. Climate science institutions have focused their efforts on back-testing, quite futile since models are tuned to match past observations. Scientists debate the question among themselves, producing no evidence sufficiently robust for taking drastic public policy action.
Yet there is a large body of expert knowledge about the validation of quantitative models. Climate science just ignores it. A simple first step would be for climate scientists to restart the climate change debate by testing the models. That is, run past models (i.e., those used in past Assessment Reports) with actual data from model creation until present – and compare their forecasts with observed weather. The result would tell us much. I wonder why they do not do so. Congress would cheerfully fund such tests.
Until then, some climate scientists will attempt to bust the “science is settled” myth. And the policy deadlock might continue, preventing us from even preparing for the repeat of past extreme weather.
About the author
Mototaka Nakamura has a Sc.D. in Meteorology from MIT (1995). He has worked at the Georgia Institute of Technology and the Goddard Space Flight Center. Now he is a Visiting Associate Researcher at the U Hawaii’s International Pacific Research Center at the School of Ocean and Earth Science and Technology. See his papers here.
For More Information
Ideas! See my recommended books and films at Amazon.
Hat tip on this to Tony Thomas’ article at Quadrant: “A Climate Modeller Spills the Beans.”
Eminent climate scientist Judith Curry has written much about this. I strongly recommend these!
- “Climate Science and the Uncertainty Monster” in BAMS, December 2011.
- “Climate Models for the Layman” (2017).
If you liked this post, like us on Facebook and follow us on Twitter. For more information about this vital issue see the keys to understanding climate change, all posts about computer models, and these posts about the climate wars …
- We must rely on forecasts by computer models. Are they reliable? (Many citations.)
- Milton Friedman’s advice about climate models, & how to win the policy debate.
- We can end the climate policy wars: demand a test of the models.
- A climate science milestone: a successful 10-year forecast!
- About Hansen’s powerful demo that climate models work!
- Panicking about climate change? See the rest of the story. – By Judith Curry, about models.
- Important: Let’s prepare for the repeat of past extreme weather instead of bickering about predictions of climate change.
Activists don’t want you to read these books
Some unexpected good news about polar bears: The Polar Bear Catastrophe That Never Happened
To learn more about the state of climate change see The Rightful Place of Science: Disasters & Climate Change
