Buy

Books
Click images for more details

Twitter
Support

 

Recent comments
Recent posts
Currently discussing
Links

A few sites I've stumbled across recently....

Powered by Squarespace

Discussion > Let's get real about climate models

Global warming, or climate change, is driven by the carbon dioxide concentration in our atmosphere, or so we are frequently told. Most of the people here are very doubtful about that for a myriad of reasons. The “science” underpinning this assertion depends on a number of factors including the greenhouse effect, the assumptions about the subsequent warming and the strength and direction of feedbacks.

The greenhouse effect is an hypothesis that has a basis in physics and the debate has been pretty well done to death so I don’t propose to go over that old ground.

These, and many other components, are brought together in the iterative calculations at the heart of climate models. The ultimate conclusion, or climate sensitivity, is an output of these models. Various papers claim a range of climate sensitivities depending on the assumptions selected as model inputs.

Another important output of these models is a forecast or prediction of future global climate. Again, there is a huge range of temperature curves, depending on the assumptions.

The famous Pause, various failed predictions, troposphere humidity and the missing hotspot cast doubt on the reliability of many of the models. This, for me, casts doubt on the credibility of the various climate sensitivities.

This is the background to the debate that I propose. Although downplayed by some, I suggest that the climate models are at the heart of climate science, future projections and policy making.

I propose that the models are unreliable. In seeking to be objective, I reproduce the following passage from Wikipedia.

A model is evaluated first and foremost by its consistency to empirical data; any model inconsistent with reproducible observations must be modified or rejected. One way to modify the model is by restricting the domain over which it is credited with having high validity. A case in point is Newtonian physics, which is highly useful except for the very small, the very fast, and the very massive phenomena of the universe. However, a fit to empirical data alone is not sufficient for a model to be accepted as valid. Other factors important in evaluating a model include:[citation needed]
• Ability to explain past observations
• Ability to predict future observations
• Cost of use, especially in combination with other models
• Refutability, enabling estimation of the degree of confidence in the model
• Simplicity, or even aesthetic appeal

Normal scientific practice is to reject models that fail to simulate observations. What is the justification for placing climate models in a category that is beyond such boundaries? Where are the boundaries? If, for example, the Pause is considered to be a background “Blip”, what constitutes a real pause, and what is the reasoning?

The same arguments apply to other failures of the models.

David Evans of Perth, Australia, claims that the basic models, for convenience a hundred years ago, equated long wave radiation blocking in the troposphere to increased solar heating of earth’s surface. This modelling device may be valid in terms of energy balance, but as a mechanism it is not valid when deciding feedbacks.

The warming of the atmosphere by blocking of IR radiation to space cannot be equated to the warming of the earth’s surface by increased visible radiation from the sun.

Warming of the atmosphere would result in more radiation to space by water vapour, short circuiting the energy loop, compensating for reduced CO2 IR emission and avoiding more heating of the surface. According to Dr Evans, the physics assumed in the models is correct, which is why the establishment gives the models their blessing. However, the architecture of the models is fundamentally flawed.

Decisions about the basic model become imbedded in history and the detail becomes buried from view as generations of developers build on the foundations. Most of the people here know that the models are flawed, but they do not know details of model construction. David Evans has made a study of that.

In conclusion, I propose that the models are flawed and ask why they are not judged according to the standards of normal science. Why are flawed models used as the basis for government policy? Why do academics and government scientists endorse models that do not agree with observation?

Why is this situation perpetuated year after year?

Jan 7, 2016 at 8:20 PM | Unregistered CommenterSchrodinger's Cat

I should add that if these questions are treated as rhetorical, then I know the answers already. I'm hoping that the believers have reasons worth hearing and that the doubters can help to refine the accusations.

It seems to me that the subject requires rigorous challenge and proper answers.

Jan 7, 2016 at 8:46 PM | Unregistered CommenterSchrodinger's Cat

If there were no models of climate, would you suggest:

a. models should be written to see whether anything can be learned;
b. no models should be written.

Jan 8, 2016 at 2:50 AM | Unregistered CommenterRaff

It seems to me that the reason Climate Models are given the special status they are is because, for some reason I don't really understand, climate change is regarded as the most likely cause of human demise. Couple with the fact that modelling the due date of a catastrophic meteor impact, a Deccan Traps event or Yellowstone megavolcano eruption.for instance is not that easy then the human need for there to be an impending disaster, we ll have a little bit of Private Fraser in us, is met by Climate Change.

The models were for the most part developed from weather forecasting techniques. Weather forecasts up to 12 months in advance can be useful if proved accurate. Apart from Piers Corbyn* I don't think model does better than a nature watcher would.

*I have no evidence other than what I read on the internet that his forecasts are accurate, although they only need to be measurably better than the MO for him to make a decent living.

Having committed a large amount of money, time and effort into the whole thing then no-one is prepared to say we're wasting our time here. Like many things it'll probably whither on the vine and a new End of the World Prophecy will take over, perhaps someone will develope a Megavolcano hockey stick model dependent of human mining and drilling activities? The earliest End of the World Prophecy dates to about 60 CE so they have a long history.

Finally I suspect a decently accurate global weather (should that be climate?) forecast for two years out is quite a way off. The same applies to predicting hurricane tracks and landfall, winters of 1947, storms of 1953 and that type of one off events

Jan 8, 2016 at 9:17 AM | Unregistered CommenterSandyS

In conclusion, I propose that the models are flawed and ask why they are not judged according to the standards of normal science. (...)
Jan 7, 2016 at 8:20 PM | Unregistered CommenterSchrodinger's Cat

".. why they are not judged according to the standards of normal science". Because "climate science" is not normal science.

New climate models are judged according to whether they agree with existing climate models. (No kidding - read the papers describing them.) So not surprising that their outputs are (i) all wrong (ii) all in agreement.

Jan 8, 2016 at 9:22 AM | Registered CommenterMartin A

models should be written to see whether anything can be learned
At last – something reasonably sensible from Raff! That is precisely why models should be written; that is the purpose of models, to help the understanding of a system or process, and view possible outcomes. They should then be tested against reality, to further the learning that they offer. What they should NOT be written for is to determine global political dogma policy.

Jan 8, 2016 at 11:16 AM | Registered CommenterRadical Rodent

i agree with Radical Rodent, models are very useful learning tools. I also believe that modelling is probably the best way to understand the climate. However, it is completely unacceptable to give any credence at all to failed models.

Climate models fail because they do not predict observations correctly. They are by definition useless. It is completely unacceptable that their output is used to shape policy of any kind. It is dishonest, non-scientific and fraudulent.

It is not acceptable to pretend that a model is fit for purpose just because elements of the output agree with expectations. That seems to be what happens in climate science.

Jan 8, 2016 at 11:41 AM | Unregistered CommenterSchrodinger's Cat

Interesting article by some non-disbelievers but far short of what really should have been said.

The United Kingdom Climate Impacts Program’s UKCP09 project makes highresolution forecasts of climate during the 21st century using state of the art global climate models. The aim of this paper is to introduce and analyze the methodology used and then urge some caution. Given the acknowledged systematic errors in all current climate models, treating model outputs as decision relevant probabilistic forecasts can be seriously misleading. This casts doubt on our ability, today, to make trustworthy, high-resolution predictions out to the end of this century.

The Myopia of Imperfect Climate Models: The Case of UKCP09

Jan 8, 2016 at 12:38 PM | Registered CommenterMartin A

So we have agreement, models should be written to see what lessons can be learned. But someone says they should not be used to determine policy - we can 'learn' from them but we can't use that knowledge. So what have we really learned?

Jan 8, 2016 at 12:50 PM | Unregistered CommenterRaff

Once more you leap to strange conclusions based on very little information, Raff. There is no reason why validated models cannot be used – but therein lies the rub: validated models. As yet, none of the climate models have been satisfactorily validated; why on Earth would you insist that they be used to influence policy, as so many seem to do?

Jan 8, 2016 at 1:07 PM | Registered CommenterRadical Rodent

I analyzed outputs of 42 CMIP5 models and found the Russian one is best of the bunch, though still flawed.

https://rclutz.wordpress.com/2015/03/24/temperatures-according-to-climate-models/

Jan 8, 2016 at 1:18 PM | Unregistered CommenterRon C.

Okay, so we can write a model to learn from it, but we can't use what we learned (and hence we haven't learned anything) until we have 'validated' the model. And Martin will doubtless tell us that we cannot validate the model. So tell me again, why would we write the model?

Jan 8, 2016 at 1:31 PM | Unregistered CommenterRaff

Also, Raff, you really should read posts more closely: while you do not mention me specifically, I am the only one saying anything like not using them to determine policy – however, what I actually wrote was “What they should NOT be written for is to determine … policy” What is meant by that is not that what has been learned should not be used, but that the model should not be written to influence policy. Quite why you cannot see what most others probably had no trouble seeing (verification, anyone?) is something that you should be asking yourself; no-one here will be able to help with that.

Jan 8, 2016 at 1:35 PM | Registered CommenterRadical Rodent

I can only imagine that the climate community regard the large deviation between model output and observation as some sort of temporary blip and the climate will soon behave itself and follow the temperature curve predicted by the models.

There is so much wrong with this. The models are invalidated by not predicting the pause. They cannot explain the pause.and they cannot predict when the pause will end. As the length of the pause approaches two decades, it cannot be dismissed as an unimportant blip. How long before the modellers admit that their models are unfit to simulate reality?

When will scientists of other disciplines point out that models that are not validated have little credibility and their output should be taken with a very large pinch of salt?

Jan 8, 2016 at 2:34 PM | Unregistered CommenterSchrodinger's Cat

Okay, so we can write a model to learn from it, but we can't use what we learned (and hence we haven't learned anything) until we have 'validated' the model. And Martin will doubtless tell us that we cannot validate the model. So tell me again, why would we write the model?
Jan 8, 2016 at 1:31 PM | Unregistered CommenterRaff

I'd prefer to use the term 'formulate'. 'Write' implies the process of translating what has already been formulated into code. Models are not necessarily executable code.

An unvalidated model is a hypothesis - in fact much of physics research has consisted of coming up with models that explain what is known. If they can't explain even what is already known they should be ditched at once. Once the ability of a model to explain (reproduce) what is known has been verified, it can then be tested by using it to make predictions and then either conducting experiments or making new observations that test the predictions. Tests that confirm the predictions provide (within a limited domain) validation of the model. With such confirmations, you have learned something.

If you know in advance that there is no possibility of validating the model, or in other words, testing the hypothesis, then its construction is merely what is technically termed, in real science, "a wank". Why would you write it? Yes indeed, why?

Jan 8, 2016 at 2:51 PM | Registered CommenterMartin A

Among climate modellers, some think it best to start over from scratch. Discussions are quiet, to limit any comfort to skeptics. The issues are summarized here:

https://rclutz.wordpress.com/2015/06/11/climate-models-explained/

Jan 8, 2016 at 3:10 PM | Unregistered CommenterRon C.

The process of optimising a model is the learning curve. Once the model is believed to be successfully completed, validation is necessary. This involves using different initialisation parameters, a range of conditions, forecasting and hindcasting until under all circumstances the model accurately simulates reality. Only then can the model output be used with any confidence.

Modelling is used in all branches of science and engineering. When designing a bridge or an aircraft, the use of a non validated model would probably lead to imprisonment.

The climate models are light years away from being validated and many people claim that the climate can never be adequately modelled.with confidence. They are clearly not reliable enough to be the basis of world policy.

Jan 8, 2016 at 3:45 PM | Unregistered CommenterSchrodinger's Cat

Ron C - Thanks for the RGB link. I have a great deal of respect for his views and will study this later.

A re-write of the basic model is covered here:
http://sciencespeak.com/climate-basic.html

Jan 8, 2016 at 4:03 PM | Unregistered CommenterSchrodinger's Cat

Some are thinking outside of the box, especially Lucarini, Herbert and Ozawa.

https://rclutz.wordpress.com/2015/03/25/climate-thinking-out-of-the-box/

Jan 8, 2016 at 4:50 PM | Unregistered CommenterRon C

Ratty, so if someone writes a new model then as long as they say it is not for policy purposes it can be used to develop policy, but if they say it is to be used for policy purposes it can't. Got it! I imagine most existing models were developed for research purposes and not for policy, so they are okay for policy. It is good to have that sorted out! I think I'm getting the hang of "skeptical" thinking.

Martin, that is somewhat different from your earlier position (or what I understood it to be) that climate models cannot be validated. What you now say is that we can validate a model by ensuring that it explains what has gone before and that it can predict what is to come. Obviously a model cannot predict human behaviour (changes in aerosol levels, rates of increase in CO2) or volcanoes. And maybe El Niño/La Niña, the AMO, the PDO or whatever oscillation cannot be predicted, but a model might still be able to explain temperature variations if given those cycles as inputs. This starts to look very much like existing modelling, so it is hard to see anything fundamental one might want to change.

Jan 8, 2016 at 6:00 PM | Unregistered CommenterRaff

Raff - The models can't hindcast successfully either without a lot of tweaking based on hindsight, Tweeaking, aka fudging can make the forecasting worse. The problem with using models when you don't understand the system and relationships is that you may get the right answer for the wrong reasons.

Part of the problem today is that 40 years ago the temperature increase,and theories were in perfect harmony so the relationships appeared to be validated. Then, 20 years ago it all went terribly wrong, I suspect that far too much of the earlier warming was attributed to CO2 so the models are now over-heating like mad and are clearly wrong.

Jan 8, 2016 at 6:12 PM | Unregistered CommenterSchrodinger's Cat

**sigh**

Raff: do you have to incessantly batter us with proof that you either do not read the posts that are written, or that you cannot understand (reasonably) plain English? Are you being deliberately obtuse, or are you genuinely thick?

I’ll type this very, very slowly, to give you a chance to absorb its meaning…

To repeat what I wrote: “What they should NOT be written for is to determine … policyAny model may be used for policy; though, to give the policy greater credence, let alone that it prove to be a sincere, workable policy, using a validated model would be wise – that many climate models do appear to have been formulated to influence policy, ignoring how valid they may or may not be, is what is irritating many sceptics. To formulate a model with the principle aim of influencing policy is stupid, if not rather sinister – it has been done at various times in history, and none of them have proven to have been honest, justifiable or ethical.

Jan 8, 2016 at 6:50 PM | Registered CommenterRadical Rodent

Cat, do you know that to be true from academic sources or just from the "skeptic" grapevine. An updated comparison between models and reality shows good agreement. Updated in this context means that the actual forcings have been used in place of estimated forcings. This comes back to what I was saying before:

Obviously a model cannot predict human behaviour (changes in aerosol levels, rates of increase in CO2) or volcanoes. And maybe El Niño/La Niña, the AMO, the PDO or whatever oscillation cannot be predicted, but a model might still be able to explain temperature variations if given those cycles as inputs.
The models, when given actual forcings up to 2005 and left to get on with it, show more warming between 2005 and 2015 than really occurred. When given the forcings that actually happened after 2005, the real outcome remains well within the predicted range.

Ratty, don't worry, I get it. When writing a new model, don't mention policy or you'll be banished to the refuse heap to be picked over by stray dogs and rats.

Jan 8, 2016 at 7:19 PM | Unregistered CommenterRaff

Martin, that is somewhat different from your earlier position (or what I understood it to be) that climate models cannot be validated. What you now say is that we can validate a model by ensuring that it explains what has gone before and that it can predict what is to come. Obviously a model cannot predict human behaviour (changes in aerosol levels, rates of increase in CO2) or volcanoes. And maybe El Niño/La Niña, the AMO, the PDO or whatever oscillation cannot be predicted, but a model might still be able to explain temperature variations if given those cycles as inputs. This starts to look very much like existing modelling, so it is hard to see anything fundamental one might want to change.
Jan 8, 2016 at 6:00 PM | Unregistered CommenterRaff

No. What has gone before is not a test or validation. It is merely a sanity check. If a model can't reproduce past data then it has failed before starting. Plus, past data was inevitably used in constructing the model. It is sometimes called "testing on the training data", and leads to hugely optimistic assessments of a model's capabability. My lotus 123 spreadsheet table can reproduce past climate data with great accuracy but that says nothing about whether it can predict future climate.

In principle, ability to predict future climate would provide a genuine test. The only problem is that, over any relevant timescale, you would have far too little data to do proper tests. Basic statistics says that even if (for example) twenty years of prediction matched almost exactly what happened, that would give you very little confidence that it was not simply due to luck.

You are of course right that climate is affected by things that cannot be predicted (volcanoes, human CO2,...). But there are firm reasons to think that, even without such factors, climate is something that cannot be predicted very far into the future.

[analogies deleted]

Jan 8, 2016 at 7:39 PM | Registered CommenterMartin A

Barf: no, you don’t “get it” (unless you really are being deliberately obtuse).

Jan 8, 2016 at 8:03 PM | Registered CommenterRadical Rodent