Bishop Hill GCMs and public policy
Aug 24, 2014 Climate: Models Climate: Parliament In the thread beneath the posting about the Chen and Tung paper, Richard Betts left a comment that I thought was interesting and worthy of further thought.
Bish, as always I am slightly bemused over why you think GCMs are so central to climate policy.
Everyone* agrees that the greenhouse effect is real, and that CO2 is a greenhouse gas.
Everyone* agrees that CO2 rise is anthropogenic
Everyone** agrees that we can't predict the long-term response of the climate to ongoing CO2 rise with great accuracy. It could be large, it could be small. We don't know. The old-style energy balance models got us this far. We can't be certain of large changes in future, but can't rule them out either.
So climate mitigation policy is a political judgement based on what policymakers think carries the greater risk in the future - decarbonising or not decarbonising.A primary aim of developing GCMs these days is to improve forecasts of regional climate on nearer-term timescales (seasons, year and a couple of decades) in order to inform contingency planning and adaptation (and also simply to increase understanding of the climate system by seeing how well forecasts based on current understanding stack up against observations, and then futher refining the models). Clearly, contingency planning and adaptation need to be done in the face of large uncertainty.
*OK so not quite everyone, but everyone who has thought about it to any reasonable extent
**Apart from a few who think that observations of a decade or three of small forcing can be extrapolated to indicate the response to long-term larger forcing with confidence.
So, let me try to explain why I think GCMs are so important to the policy debate.
Let us start by considering climate sensitivity. As readers here know, the official IPCC position on climate sensitivity is largely based on the GCMs. This time round we have had some minor concessions to observational estimates, but a significant proportion of the probability density of the observational studies remains outwith the IPCC's likely range of 1.5-4.5°C. Proponents of GCMs might counter that the upper end of the GCMs are ignored too, but I would suggest that one should conclude that an ECS of 5-6°C in the light of temperature history.
Estimates of climate sensitivity - and therefore in practice GCM estimates of climate sensitivity - directly inform estimates of the social cost of carbon. So when people like Chris Hope are arguing for a carbon tax of $100/tCO2, this is a function of GCMs. I recall, I hope correctly, that Chris suggested a figure of $18/tCO2 if one used an ECS of 1.6, in line with observational estimates. This matters of course, because the policy response, if any, to an $18 problem is significantly different to that for a $100 problem.
Wherever we look in the interactions between scientists and politicians on climate questions, we see an emphasis on catastrophe. We see no confessions of ignorance, but only occasional reference to uncertainties. Here's some notes of Tim Palmer addressing the All-Party Climate Change Group:
With the amount of carbon dioxide already in the atmosphere, future emissions will need to be reduced by half to that of historical emissions to limit global average temperature rise to 2°C. However, if emissions are not curbed (under the business as usual scenario), the amount of carbon dioxide in the atmosphere will be three times the historical emissions and the temperatures might rise up to 4°C.
And on the other hand they might not. This idea does not, however, seem to have been put forward for consideration.
Readers might also wonder what explanations were given to our political masters on the credibility of the GCMs. Here's what Palmer said:
Climate models are only flawed only if the basic principles of physics are, but they can be improved. Many components of the climate system could be better quantified and therefore allow for greater parameterisation in the models to make the models more accurate. Additionally increasing the resolution of models to allow them to model processes at a finer scale, again increasing the accuracy of the results. However, advances in computing technologies would be needed to perform all the necessary calculations. However, although the accuracy of predictions could be improved, the underlying processes of the models are accurate.
Apart from the transport of heat to the deep ocean, if Friday's paper from Chen and Tung is to be believed.
You can see that policymakers are getting a thoroughly biased picture of what GCMs can do and whether they are reliable or not. They are also getting a thoroughly biased picture of the cost of climate change based on the output of those GCMs. They are simply not being asked to consider the possibility that warming might be negligible or non-existent or that the models could be complete and utter junk. They are not told about the aerosol fudging or the GCMs' ongoing failures.
And this is just scratching the surface.
[BTW: Could commenters who like to amuse themselves by baiting Richard please refrain from so doing!]
Reader Comments (306)
Richard,
Oh, I didn’t think of that. Let me give it some thought ;-)
aTTP
What is your response to the post by HaroldW at 2.48pm?
What's HaroldW is suggesting - I believe - that it is technically possible to construct a dataset in which the last 15 years are definitely cooling but in which the 1880 - 1999 trend is smaller than the 1880 - 2014 trend. Sure, I can see that as being possible. My comment would be that in the real temperature dataset, the uncertainties on the trends are large for short time intervals, so saying anything definitive is not easy. In a similar sense, in my linear analysis the 1880-1990 trend was 0.052 +- 0.009 degrees per decade, and the 1880-2014 trend was 0.063+-0.008 degrees per decade. So, one could then argue (using the uncertainties around each trend) that there was a 95% chance that the warming over the period 1880-1999 was between 0.528 and 0.744 degrees. For the period 1880-2014 it becomes 0.737 to 1.0224. Therefore, that does indicate that there is a non-negligible chance that there was cooling during the period 1999-2014, but a much bigger chance that it warmed. Of course, if we are undergoing something like what HaroldW suggests, we should continue to see a bigger overlap until it's clear that we are cooling. If we continue to increase our emissions, that would seem rather unlikely.
Averages of chaotic response are chaotic. You can demonstrate this for yourself by use of the algebraic logistics map or the original Lorenz system from 1963, or your favorite system that exhibits chaotic response.
Calculate a very long series response for your favorite chaotic demonstration. Calculate an average of the response. Using any one of the available methods, calculate the Lyapunov exponents for the series of averages. There will be a positive exponent.
The Google, standard or Scholar, will lead you to canned, off-the-shelf software for calculating Lyapunov exponents for long series of numbers, including con-current calculations for systems of ODEs.
Weather is said to be chaotic. The weather, phenomena and processes internal to Earth's climate system, affects the energy balance at the top of the atmosphere, as do other physical phenomena and processes of longer time scales than weather. The oceans have an extremely long time constant relative to responses to changes both internal and external to Earth's climate systems. The swamp-oceans vs. coupled-oceans approaches clearly reflect the importance of high-fidelity modeling and the response time of the oceans. So long as responses internal to the systems are reflected in the ToA energy balance, the problem is not a boundary-value problem.
Because the long-term climate trends are not chaotic.
If the statement means, say, averages over different time periods (a) indicate that a trend is present, and (b) the averages do not exhibit chaotic behavior, then see first paragraph of this comment. Temporal chaotic response for autonomous ODEs cannot display trends; it's impossible. The region of phase space on the attractor that will be occupied at any time cannot be predicted. And this refers to the relatively simple case of low-dimensional temporal chaos. The real-world case of infinite-dimensional temporal-spatial chaos is more difficult.
If I put a double pendulum on the floor and set it in motion, it's behaviour will be chaotic. If I kick it, will it's motion through the air be chaotic, or will it's overall trajectory be parabolic?
Kindly map the elements of this analogy to the case of Earth's climate system response. It seems to me that you have applied a dominant external forcing that is completely unrelated to the inherent chaotic response of the double pendulum. Its motion through the air is likely somewhat parabolic, but we're interested in what happens to the inherent chaotic nature of Earth's energy content.
. . . but you can't predict what the temperature will be on any given day in 7 years time.
But isn't this exactly what the GCMs are used to project; the difference in temperature between December 2000 and December 2100 ( December and all other months )? I think getting the delta for a specific month correct is a more difficult problem than getting the difference between July and December correct.
ATTP, your statements about chaotic systems are simply wrong and the exact reverse of what anyone *who actually understands such systems* would claim. Your simple assertion that you do understand is just laughable given your responses.
You are trying to determine statistical properties of weather in 20 years time. Think in terms of sample and population statistics at different timescales.
Answer me one simple question. What is the relationship between the sample mean and the population mean for a fractal system at one month time scale, in comparison to the relationship between the sample mean and population mean for the same system at a 30 year time scale? And further, the relationship between sample mean and population mean for a thousand year time scale? You can assume a Hurst exponent of unity.
This is a really easy question in the context of fractal dynamics (and, by definition, climate dynamics). If you really understood the answer to this, you would realise why your response about "predicting the weather in 20 years time" is so ignorant (and absolutely not an answer to my original concerns, which you still have yet to grasp).
"Well, I'm pretty confident that as we move into the latter part of the year we will start to get colder ..."
What's that got to do with anything?
"I do remember reading some of that paper, but have forgotten what it said."
Evidently!
Let me remind you:
"Here's a question for you. If I put a double pendulum on the floor and set it in motion, it's behaviour will be chaotic. If I kick it, will it's motion through the air be chaotic, or will it's overall trajectory be parabolic?"
If you want to get picky about it, yes it will be chaotic, and no, it will not be parabolic. The component masses will vary in distance from the body of the Earth, and therefore vary slightly in magnitude. That is to say, there are tidal forces. And of course there's air resistance.
If we're speaking approximately, and we consider in both cases an isolated system with no external forces, then the motion is chaotic in exactly the same sense it was before. The motion of the masses relative to one another is chaotic, the motion of the centre of mass is not.
Richard Betts says, "Everyone* agrees that CO2 rise is anthropogenic"
Sorry, those who have thought about - and done research know this is not true. Only 4% of the observed rise is anthropogenic (a relatively simple sum to calculate from known consumption of fossil fuel), the rest is natural. The fall in C13 ratio is claimed as evidence that it must be from fossil fuels which are very low in C13, being organic in origin. Oh dear! As most of the natural rise in CO2 is ALSO FROM ORGANIC MATERIAL, it too will be low in C13, so tat puts that claim back in its basket!
"Only 4% of the observed rise is anthropogenic (a relatively simple sum to calculate from known consumption of fossil fuel), the rest is natural."
Don't confuse the proportion of the extra CO2 molecules that are anthropogenic, and the proportion of the extra CO2 amount that is anthropogenic. The two concepts are not the same.
The facts show that the agreement between models and observations is tenuous and steadily eroding and will be statistically unacceptable in about a decade. And yet the IPCC tells us with “very high confidence” that models agree with observations and therefore are a reliable indicator of future climate changes.
http://www.cato.org/blog/clear-example-ipcc-ideology-trumping-fact
Nullius,
Exactly. The boundary conditions of our climate constrain the range of chaotic motion. Everyone here seems to be focusing on the nature of the chaotic aspects, while ignoring that there are boundary conditions that constrain it.
And the relevance is what? I don't know the answer to this and I'm not sure that I'm even going to work it out. What are you really suggesting? That we could, by chance, have -20 degrees in London in July?
Everyone* agrees that CO2 rise is anthropogenic
Not true. Most scientists know that carbon dioxide solubility in the oceans is a function of temperature. The warming of the planet since the LIA will have increased atmospheric CO2.
Everyone** agrees that we can't predict the long-term response of the climate to ongoing CO2 rise with great accuracy. It could be large, it could be small. We don't know. The old-style energy balance models got us this far. We can't be certain of large changes in future, but can't rule them out either.
True, I couldn't agree more, but somehow policymakers have got the idea that you can predict the response and they get the impression that it is very large indeed.
I wonder who advises the government on such matters?
"Everyone here seems to be focusing on the nature of the chaotic aspects, while ignoring that there are boundary conditions that constrain it."
No. The boundaries for variability are enormous, compared to the quanta of change sought to be modelled by GCMs and the timescales under consideration. What you seek to do is stress the long-term and large-scale constraints and predictability as though these had any relevance to the question at hand.
In other words, predicting climate state 100 years from now is the same problem as predicting temperatures 'midday on 20 July 2015'.
Are the computer models reliable?
Computer models are an essential tool
in understanding how the climate will
respond to changes in greenhouse gas
concentrations, and other external effects,
such as solar output and volcanoes.
Computer models are the only reliable
way to predict changes in climate. Their
reliability is tested by seeing if they are able
to reproduce the past climate which gives
scientists confidence that they can also
predict the future.
Met Office pamphlet 2009. Warming Climate change – the facts
Models help to understand the potential impacts
of climate change at a regional level
Increasing temperatures and other
related impacts are likely to cause
changes in the winds in the atmosphere
and rainfall patterns. This could affect
weather at regional and local levels.
The increasing sophistication of climate
models and increases in computer power
both contribute towards scientists being
able to make more robust projections
about regional changes and the impacts
those changes may have on factors such
as food, water availability, and flooding.
Met Office Pamphlet 2013. The changing climate: past changes and future projections
Martin A - gotta love that stuff from the Met Office!
Re: their 2013 spiel - I wonder what they think of Bony and Stevens?
Whether a system is chaotic or not is going down the wrong track, IMO.
If you have a quantity like T(theta,phi,t) (thermodynamic temperature field at a latitude, longitude measured at time t), what we're interested for forecasts is not the exact value of a quantity, but the possible range of values of that quantity. Typically, that means the mean and other low order moments of that quantity (e.g., its variance).
If the range of possible values is very large, the forecast is not very useful.
If the range of possible values is very small, the forecast is useful, even if we can't mathematically predict the exact value of a quantity at a particular time.
For complex systems, we generally aren't even interested in the individual trajectories whether they are knowable or not, just the mean, variance and other low-order moments of the system.
Forced motion of a double pendulum is a good example of this:
Even though the error in the forecast of the motion of the pendants will grow exponentially over time, we can still write down relationships between the magnitude of the forcing and the mean amplitude and variance in amplitudes of the responses of the individual pendants.
Note by the way that you can get an exponential increase in the uncertainty in the phase of oscillating pendants even though the system remains bounded. The amplitude obviously does not arbitrarily increase when you increase the magnitude of the forcing on the pendants.
The same applies to climate: Just like the double pendant, the system is bounded and constrained by energy conservation properties. As long as you can predict the mean temperature (and its variance) in response to an change in forcing, this is all that is needed for making policy decisions. Whether or not you can accurately describe the temperature over the course of a year one century from now is not a relevant question.
ABSTRACT
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
'
Why is there such a fear of ignorance and uncertainty within the “climate science” groups? Do they not realise that, it is in the admission of ignorance and the admission of uncertainty that there is a hope for the continuous motion of human beings in some direction that doesn’t get confined, permanently blocked, as it has so many times before in various periods in the history of man?
...and Then There's Physics said
"Remember, that when we refer to something as being chaotic in this context, we mean that it is deterministic."
I'm not arguing that chaotic behaviour is non-deterministic, but it is not computable, ie not predictable. Just look at the simple 3-body problem ie 2 fixed suns and a planet and try an predict how a slight change in starting position affects long term clumping of orbits about the suns (analogous to climate or the jetstream). And as you can't get initial conditions exactly right, and because computers round on each iteration of their calculation one doesn't even know if the predicted system state at any point in the time evolution is even real.
http://faraday.physics.utoronto.ca/GeneralInterest/Harrison/Flash/Chaos/ThreeBody/ThreeBody.html
You certainly won't be able to skillfully predict jetstream year by year patterns and associated emergent (important word) climate for europe.
"Exactly. The boundary conditions of our climate constrain the range of chaotic motion. Everyone here seems to be focusing on the nature of the chaotic aspects, while ignoring that there are boundary conditions that constrain it."
Yes, they're constrained, but they're not so constrained that you can't get 2 C multi-decadal variations from the chaotic bit. Which is what you appear to be arguing with.
By saying there are boundary conditions that constrain the changes, that does nothing to show that they constrain the chaotic aspects people are talking about. It appears to be an irrelevant observation.
I can tell you the answer. They are the same.
The sample mean, as a method of estimating the population mean, of a climatic quantity (such as global temperature), has almost identical confidence intervals whether you measure it at one month scale, thirty year scale, or one thousand year scale.
The reason we can expect this is that it is an invariant property of an attractor that behaves as a fractal - and up to now that is about the only invariant statistical property of the attractor that has been characterised. Of course you can argue that this is not a property of the attractor, which is even worse as you do not even have this relationship to establish properties of the climate.
Any property - such as the behaviour of long term means - MUST be analysed in terms of invariant properties of the attractor. These cannot be *assumed*.
And no, I'm not arguing we cannot predict that London will be unlikely to reach -20 deg C in July. I am comparing our ability to accurately predict at different scales, and noting climate change is just as detectable in a single month as it is over 30 years, or even over 1000 years, due to the invariant properties of the non-linear dynamics of the system. This is fundamental as it shows the need to wait for 30 years is an irrelevant nonsense. And the argument I present is a rigorous one based on properties of the chaotic system - which absolutely constrains the limits of predictability of the system.
It is very tiring that you trot our completely irrelevant memes that are the standard set-response of climate activists, without showing the first sign of understanding the points being put in front of you.
Carrick: where we differ on this is that you cannot make assumptions about the invariant properties of the attractor, you must investigate the attractor for properties and determine physical relationships from these invariants. One of the classic assumptions - that there is some level of independence in climate at longer time scales which allows things to "average out" - is fundamentally flawed due to the invariant property of the attractor of fractional integration. This places bounds on the predictability of the system at different scales, and has surprising / unexpected results, for example a step change in the population mean is equally detectable at the one day scale as it is at the thousand year average. The most effective route to understanding this paradox is through understanding climate through seeking out invariant properties of the chaotic system - not through arbitrary assumptions.
aTTP (4:49 PM) -
I don't think you've quite got the gist of my earlier (2:48 PM) comment. Mathematically, one *can't* combine OLS trends in the fashion which you did. For example, one can't get the OLS trend for (say) 1900-1999 by averaging the trends for 1900-1949 & 1950-1999. Likewise, one can't impute the trend for 1999-2014 by comparing the trends for 1880-2014 with that for 1880-1999. [I say "one can't" in the sense that sometimes the answer may come out right, but it's rather like being set the division problem 16/64 and canceling the 6's to get the answer 1/4.]
The quadratic temperature series was the simplest one I could come up with to demonstrate the principle, which you seem to feel is artificial or unlikely. Very well, let's use the actual HadCRUT4 series instead. As you wrote, the OLS trend from 1880-1999 is 0.052 K/decade, and the 1880-2014 OLS trend is 0.063 K/decade. Let's construct a no-warming-for-15-years scenario by artificially setting the value for years 1999-2013 to the 1998 value, viz. 0.531 K. The 1880-2014 trend comes to 0.066 K/decade. Higher by about 20% than the 1880-1999 value, yet by construction there has been no warming since then. Alternatively, if one sets all years 1999-2013 to the coolest observed value (2000's 0.294 K), the 1880-2014 trend comes to 0.056 K/decade, still higher than 1880-1999.
The increase in the OLS trend from 1880 over the last 15 years was pretty much pre-destined by the pattern of the earlier values. Just as it's close to guaranteed that when the 2014 figures are final, the new 15-year OLS trend will be lower than the current one (which for HadCRUT4 is 0.075 K/decade).
Representing the OLS trend as an estimate of the linear warming rate (with an associated uncertainty), and interpreting this as a probability distribution of an actual parameter, isn't any better. The temperature series is not a linear trend plus white noise. As a result, the OLS trend, while well-defined mathematically, does not represent anything so clear-cut. The metric gives a good idea of the average rate of warming over an interval, but it should not be reified into something more than it is. You are hardly alone in making more of the OLS trend metric than it deserves, but it behooves one to remember its many limitations.
Why is there such a fear of ignorance and uncertainty within the “climate science” groups? Do they not realise that, it is in the admission of ignorance and the admission of uncertainty that there is a hope for the continuous motion of human beings in some direction that doesn’t get confined, permanently blocked, as it has so many times before in various periods in the history of man?
Aug 25, 2014 at 6:39 PM | Registered CommenterRadical Rodent
----------------------------------------------------
Excellent point, RR.
For example, even 50 years ago, who would have predicted that people from all over the world could be having this discussion, in real time, at a price that most people in the West can easily afford.
Human ingenuity and technological progress is the ultimate chaotic system. And, anyone who claims to be able to predict what people 50 years from now will be doing in terms of communications technology is just making stuff up.
That is why, even if the worst prognostications about climate in, say, 2100 are true, it is just plain foolish to impoverish ourselves now in the vain hope that we can do something about it. On the contrary, we need to do everything we can to support science, technology and wealth creation. That is the only way to maximise the chances of humanity surviving whatever Gaia throws at us.
Spence_UK, I actually don't think we're that far apart. In particular I fully endorse making progress "through seeking out invariant properties" of the system (regardless of whether it meets the definition for being chaotic).
Because numerical modeling costs so much to do effectively (to the point that it is sapping funding for ordinary meteorological forecasting), I think there is a certain incentive within the group of numerical modelers to "oversell" their products. And I think we are seeing a certain amount of pushback against people who try to develop simplified models, even though these simplified models have greater practical utility.
The over-reliance of the IPCC on the GCM outputs is more of an outgrowth I think of the political expedience in not ignoring results that cost so much to achieve, even when those results are of questionable actual value.
Thank you, Johanna, but I have to confess that it was troll-bait (hence the erroneous comma). Pity they are not biting tonight, it would have been interesting to see the response when I revealed that it is actually a quote from Richard Feynman.
As you rightly say, human progress is an archetypal chaotic system; the only direction it cannot go is backwards, though there are many who would try to force it to do so. As the world today is beyond the wildest dreams of anyone trying to imagine how it would be, as they waved their children off to the Great War, so will the world be as our grandchildren welcome in the next century
HaroldW,
I think I do get what you're getting at and I didn't actually do what you seem to be suggesting. I agree with you that there is a function for which one can have a larger trend for the longer time period even if there was cooling in the latter part of the time interval. The issue, though (as I suspect you understand) is that in the case of something like the instrumental temperature record, the variability is such that determining trends for short time intervals is difficult because of the large uncertainties. So, I'm certainly not arguing that we couldn't have been cooling for 10 - 15 years. I'm suggesting that we can't state that we've been cooling.
and Then There's Physics - Aug 25, 2014 at 1:58 PM:
and Then There's Physics - Aug 25, 2014 at 8:25 PM:
Why are you guys wasting your time with this joke? Seriously?
Laurie,
Fair point, what I said in the main part of the comment was too strong, but my parenthetic comment,
was intended to clarify the situation. As I responded to HaroldW, it is possible there's been no warming. My issue is with the claim that there has been cooling, not with the suggestion that there could have been cooling.
Spence_UK,
Then I'm confused by the point you're trying to make. Are you trying to suggest that the mean temperature of the planet (or some region of the planet) is determined by the sample mean of some chaotic process? If so, that would appear to be largely nonsensical. The mean temperature is set by how much energy we get from the Sun, our albedo, and the composition of out atmosphere. The variation about that mean can certainly be chaotic (i.e., the temperature we have at any particular instant in time) but the mean is not simply the mean of some chaotic process. That's essentially the fundamental point. Our climate is chaotic in the sense that we can't accurately predict something in advance (well more than a few days) but it's not chaotic in the mean (i.e., we have some idea of the typical climate of a particular region at particular time).
Or, what you are calling completely irrelevant memes actually do illustrate what you're saying, and you just haven't realised this yet.
Radical,
I'm not aware of the context - or missed something - but I find that when people quote Feynman so as to, supposedly, illustrate some deep and meaningful point, what they normally do is illustrate that they didn't really understand what Feynman was trying to illustrate. Just an observation. May not be the case in this instance.
Carrick, I suspect we are not that far apart, but you cannot put the non-linear dynamics to one side; any analysis must take the chaotic nature of the system into account (i.e. the analysis performed must be appropriate to the nature of the system).
Many counter-intuitive results appear as a consequence to this.
I broadly agree with your sentiment about GCMs though - which I discussed on the previous page, they have value but not in the way they are currently being used.
Of course when people who have some depth of understanding of chaotic systems are presented with frankly daft analogies and straw men about predicting London's temperature in July frustrations can set in. I don't really blame ATTP - s/he doesn't know any better, and is just regurgitating the nonsense one-liners fed to them by the climate activist community. But reasoned debate on the topic can be... challenging.
aTTP: I’m sorry, but I have to give up with you – if you are unable to see the context or point of the use of the quote, then there is no way that I will be able to explain it to you, though I will make an attempt. My understanding is that Feynman was pointing out that ignorance and uncertainty are nothing to be afraid of in science, but to refuse to admit to either is to deny any advancement – and “climate science” has long been denying both.
Spence_UK,
Good of you to be so condescending :-) My own feeling is that people who don't understand basic physics resort to complicated arguments about chaos. Maybe you don't know any better either :-)
I'd still like to know what you're getting at with your emphasis on the sample mean. Are you really suggesting that various climatic properties are simply set by the sample mean of a chaotic process and are not influenced at all by whatever boundary conditions exist?
FWIW, I agree broadly with what Carrick has said (although haven't quite digested it all). Even this may be a reasonable comment
Just to be clear, I'm certainly not defending all aspects of GCMs. They have limitations and it may well be that some are giving more credence to some GCMs results than is warranted. But aspects like climate sensitivity, global circulation, water cycle are likely to be much more reliable than other aspect. Having some understanding of what we can trust and what we should be more wary of is important and reasonable. Dismissing it all is medieval.
This much, we can agree on.
Ermmm... no. Do you understand the concept, and the difference, between a *sample* mean and a *population* mean? They are not the same thing. The *population* mean is the true mean of the underlying process. We cannot directly measure the population mean. The *sample* mean is the mean of observations we make of that process. We often use the sample mean as an *estimate* for the population mean. But it is an imperfect estimate.
In classical statistics, the sample mean converges onto the population mean with increasing numbers of samples, due to independence of samples. In fractal dynamics, this is not true; the sample mean does not converge on to the population mean, or at least converges so slowly that more samples than there is time for in the lifetime of the universe are required for convergence.
As Dan Hughes has already noted, it is trivial to show that if a process f(x) is chaotic, then an average (or locally integrated) f(x) is also chaotic. So your statement here is exactly wrong. By definition, the mean must be chaotic.
Well, it is always possible that I am wrong. However, for someone to convince me of this, they would have to (a) understand what a sample mean is, and (b) understand that the local average of a chaotic system is also a chaotic system, since these are fairly obvious when you are aware of the fundamentals of chaotic systems.
Chaotic systems are not completely intractable, although there are hard limits to their predictability. But when you make so many basic errors in your assertions, it is hard to take you even remotely seriously on this topic.
Bish, as always I am slightly bemused over why you think GCMs are so central to climate policy.
Yes. Just because it says Aspunguent on the tin doesn't mean it actually contains an asp - or that it is an unguent.
Spence_UK: one of the challenges is that aTTP is a champion cherry-picker, as well having proved to be unable to see evidence in front of his/her eyes (odd, for one who so desperately wants to be known as a physicist). It will be interesting to see what is cherry-picked from your post, and how it will be (mis?)interpreted.
Spence_UK,
It is distinctly possible that I don't understand the difference between a sample mean and a population mean, but this seems like the standard "talk about some kind of statistical thing and then use that to argue that we don't understand some physical process". You say,
What does this really mean? If we consider global temperatures, they really are set by the amount of energy we get from the Sun, our albedo, and the composition of our atmosphere. Of course, at any instant in time the actual temperature of a point on the surface of the planet will not be precisely set by these properties, but it will be constrained by these properties. If the albedo goes down, the average temperature goes up. If we increase atmospheric GHG concentrations, the average temperature goes up (please tell me if you disagree with this or not, because if you do disagree, then I truly am wasting my time). If the Sun gets brighter, average temperature goes up. Of course there are chaotic processes in our climate, but arguing that it is all fundamentally chaotic is ignoring basic physics.
So, I may not be able to determine precisely the temperature of some region of the planet at some point in the future, but I can estimate how much the average temperature may have changed compared to today because of our emissions. Similarly with other climatic processes. We can't predict in the future how much rain we will get on a particular day in 2035, but we can estimate how much or less rain a particular region may get.
It's also possible that I'm wrong, but until someone convinces me that they understand the underlying physics and the theory of chaos, I'm going to have to assume that they're using their understanding of chaos to make arguments that don't necessarily satisfy the basic laws of physics. Again, we're not trying to make weather predictions, we're trying to make projections about changes to our climate driven by increases in our emissions.
Pointman has an excellent, and funny, post up this week about how the zombies are invading sceptic blogs because nobody is coming by their own.
"How the fortunes of the once mighty troll armies of undead have turned; reduced to creeping around skeptic blogs while pathetically trying to cadge a click or two. Gowon, gowon, gissa click Guv’nor. I knew one day we’d pass them in the up elevator, as they were going in the other direction, but frankly I never thought it’d be this toe-curlingly embarrassing.
Have they no self-respect?"
http://thepointman.wordpress.com/2014/08/20/28-predictions-later-or-the-invasion-of-the-living-dead/
The concept of sample mean and population mean is a fundamental aspect of any observational science. Any experiment that involves observations implicitly relates sample and population means. What I am highlighting is the difference between classical statistics and how the application of the concept to complex, non-linear dynamics requires us to revisit the fundamentals, and can often result in what we accept without even thinking about it can turn out to be completely wrong. And the impact of fractal dynamics do undermine the properties of averaging due to non-independence of samples.
Actually, I would argue it is pretty much precisely set by these properties. But these properties - particularly albedo and atmospheric composition (not so much the sun) are a part of the state vector of climate, and inherently part of the chaotic system. The hydrological cycle in particular (water vapour and clouds) are part of the chaotic system themselves, and so the fractal dynamics apply to all of these. The mistake is to believe these average out - they do not average out, they are part of the attractor and follow the properties of the attractor. As for the sun, this is somewhat external to the chaotic behaviour of the system but the variation of this is small in comparison to the fluctuations imprinted on the dynamics of the system.
I dispute your ability to do that, on the grounds outlined above.
Precipitation predictions for GCMs are a farce - they predicted no change over the 20th century for global rainfall, which varied hugely (at a multi-decadal level) throughout the period. Given that precipitation doesn't even manage the 2-3 degrees of freedom in hindcast, your belief that predictions out to 2035 have skill is... quaint.
Principles of chaotic systems are really not that difficult. A chaotic system must meet three criteria; exponential divergence from arbitrarily small differences in initial conditions; bounded (i.e. does not go off to infinity); and topologically mixing. Which of those things do you think will change under local integration?
Spence -careful of the sophistry will you try to be please
Anders is not claiming the ability to compute how much temperatures will have changed by 2050, rather how much thry MAY have changed by then, due to CO2 emissions.
This is of course meaningless, but logically robust.
Spence_UK:
Again I agree with you… the fact the system is nonlinearity is one that many people want to place on the side and just treat it as a linear system. I think this is an exaSpence_UK:
Again I agree with you… that the system is nonlinearity is something many people want to place on the side and just treat it as a linear system. I think this is an example of "if what I have is a hammer, everything looks like a nail". That is, many people don't work with nonlinearity and so are somewhat naive about the consequences of nonlinearity on the phenomenology of the system.
I don't claim an applied mathematician's background on chaos, but many of the problems I work with involves modeling complex nonlinear systems, so I think I'm significantly less naive than the average scientist about the "gotchas" associated with nonlinearity in complex systems.
On another note, I think many people are naive about just how bad the GCMs are. It's not just an issue of getting the trend wrong, the bigger problem is the absolute temperature predicted by the various models is all over the place. This is a huge issue given the types of nonlinearity present in the model.
Again agreed...I see them as an essential tool in the toolbox, but not as they are typically being used and especially not with their over reliance in the IPCC reports. That said, some people, e.g., Isaac Held, I think do understand the limits of the full-blown models and do use them in a more proper manner as research tools, rather than forecast programs.
Some of my more technical research involves modeling from semi-analytic limit-cycle oscillator models to linearized bulk-parametrized 1d frequency-domain CFD models to full blown 3d nonlinear time-domain CFD models with as complete of a physical description as is practicable. I don't use the more full blown models to directly study the system or make direct predictions, but rather to study the accuracy and limits of applicability of the more simplified models.
With the simplified model, it is sometimes possible to understand the behavior of complex systems and make experimentally verifiable predictions. With full blown "kitchen sink" models and without the insight of simplified semi-analytic modeling, even if you can accurately forecast with them, there is a limit to how informative these models can ever be.mple of "if what I have is a hammer, everything looks like a nail". That is, many people don't work with nonlinearity and so are somewhat naive about the consequences of nonlinearity on the phenomenology of the system.
I don't claim an applied mathematician's background on chaos, but many of the problems I work with involves modeling nonlinear behavior, so I think I'm a bit less naive than the average scientist.
Again agreed...I see them as an essential tool in the toolbox, but not as they are typically being used and especially not with their over reliance in the IPCC reports. That said, some people, e.g., Isaac Held, I think do understand the limits of the full-blown models.
Some of my more technical research involves modeling from semi-analytic limit-cycle oscillator models to linearized bulk-parametrized 1d frequency-domain CFD models to full blown 3d nonlinear time-domain CFD models with as complete of a physical description as is practicable. I don't use the more full blown models to directly study the system or make direct predictions, but rather to study the accuracy and limits of applicability of the more simplified models.
With the simplified model, it is sometimes possible to understand the behavior of complex systems and make experimentally verifiable predictions. With full blown "kitchen sink" models, even if you can accurately forecast with them, there is a limit to how informative these models will ever be. But as they stand, they are not accurate at all and should not be used for forecasting.
I'll go back and read your earlier comments later this evening.
The first paragraph of mine was duplicated. Teaches me to not preview. Sorry about that.
Here is a redo of that comment:
Spence_UK:
Again I agree with you… that the system is nonlinearity is something many people want to place on the side and just treat it as a linear system. I think this is an example of "if what I have is a hammer, everything looks like a nail". That is, many people don't work with nonlinearity and so are somewhat naive about the consequences of nonlinearity on the phenomenology of the system.
I don't claim an applied mathematician's background on chaos, but many of the problems I work with involves modeling complex nonlinear systems, so I think I'm significantly less naive than the average scientist about the "gotchas" associated with nonlinearity in complex systems.
Again agreed...I see them as an essential tool in the toolbox, but not as they are typically being used and especially not with their over reliance in the IPCC reports. That said, some people, e.g., Isaac Held, I think do understand the limits of the full-blown models and do use them in a more proper manner as research tools, rather than forecast programs.
Some of my more technical research involves modeling from semi-analytic limit-cycle oscillator models to linearized bulk-parametrized 1d frequency-domain CFD models to full blown 3d nonlinear time-domain CFD models with as complete of a physical description as is practicable. I don't use the more full blown models to directly study the system or make direct predictions, but rather to study the accuracy and limits of applicability of the more simplified models.
With the simplified model, it is sometimes possible to understand the behavior of complex systems and make experimentally verifiable predictions. With full blown "kitchen sink" models and without the insight of simplified semi-analytic modeling, even if you can accurately forecast with them, there is a limit to how informative these models can ever be.
I liked this part:
ATTP says *we* can't state it is cooling but it may be.
ATTP says the climatocrats are not trying to predict weather in 20 years time.
So who can say it is cooling/paused? Where do they get the proper permits?
If they are not trying to tell us the weather is not going to be catastrophically bad, then what the effin' heck are we wasting the billion$ of dollars on?
Spence_UK,
I'll make two brief comments. This for example is something that many would dispute
There is little evidence to suggest that our albedo and the atmospheric composition are significantly influenced by variability. Yes, they can be to some small extent, but we have little evidence to suggest that variability in our climate can suddenly change the albedo or the composition of the atmosphere to the extent that it actually switches us into a different state. Our climate is forced (externally) rather than unforced. You really seem to be arguing that the entire system is chaotic, rather than it being chaotic variations about a largely externally determined mean state. I think you're wrong about this, but I would encourage you to publish it, because if you're right, you may have just completely revolutionised climate science.
However, this, makes me think you're not
This just seems like a complicated way to say that we can't really do something. Let's be clear. Over the coming decades, we don't expect solar flux to vary significantly, we don't expect our albedo to change substantially (or chaotically) and hence we can estimate how much we will warm based on our future emissions. This is clearly possible to do. Sure, we can't determine it precisely, but that's what uncertainty intervals are for.
I think I'm going to call this quits. Maybe you're right and I'm missing something here, but I doubt that I'm going to have an epiphany via blog comments. It really just seems that you're using a seemingly complex argument about chaos theory to argue that we can't understand a system. Many people who actually work in climate science disagree, so I would encourage you to publish your idea and see what feedback you get from the actual experts.
Many people who actually work in homeopathy disagree, so I would encourage you to publish your idea and see what feedback you get from the actual experts.
Spence_UK,
I was going to withdraw, but was interested in this comment you made,
Do you mean that there was a trend in global rainfall over the 20th century that GCMs have failed to predict, or do you mean that there was substantial variability that no single GCMs has been able to predict (in magnitude at least, if not in phase)?
Well, I'm thinking of the hydrological cycle, which includes water vapour and clouds. And yes these vary a lot. The mistake made is to believe these somehow average out over time; they do not, due to the fractal dynamics imparted by the attractor.
Note that clouds are often used as examples of fractals when explaining fractals to others. But the fractal nature does not end at the fluffy bits of the cloud edge. It continues throughout the cloud, throughout clusters of clouds, to regions, to global averages and temporal averages. And these things (cloud, water vapour) have a huge impact on temperature, radiative balance, etc., far bigger in fact than CO2.
Many problems with this part. When I refer to climate state, I am referring to the continuous variables defining the climate (whether it be temperature, pressure, atmospheric components etc). It is wrong to think of "switching to a different state", or to think of things "suddenly" changing. Remember fractal dynamics describes behaviour across timescales; what is "sudden" in one timescale is gradual in another. Far better to think of this as the trajectory of an attractor, with a state vector, rather than quantised states.
Actually - you are quite wrong, I consider the climate to have a population mean, as I've already explained, but the sample mean is a poor estimator of population mean, at all timescales. This is quite normal in fractal dynamics.
It has already been published. Some original analysis by Kolmogorov in the 1940s, first observations in nature by Hurst in the 1950s, then developed by Mandelbrot in the 1960s. With the exception of Klemes, relatively little development of these ideas happened in climate in the 70s/80s/90s, despite fractal concepts being developed in other areas (most notably computer networking). However, publications in climate started up again in the early 2000s, with notable contributions by Cohn, Lins, Koutsoyiannis. Prof Koutsoyiannis has probably published more than most, some reading material here:
http://www.itia.ntua.gr/en/documents/?title=&authors=koutsoyiannis
IPCC AR4 should have included a commentary on this - the first order draft had one paragraph arbitrarily dismissing it, on hand-waving grounds. This paragraph was deleted by the second order draft (presumably the authors realised it was better to say nothing than to say something absurd). IPCC AR5 is better; they acknowledge the body of work and simply state that fractal dynamics have not been accounted for in their analysis. That is a much more honest, scientific approach - to acknowledge work has been done but the IPCC authors do not know what the impact of this will be. Of course, it would be better if they had researched it, and did know, since the information is available to them.
Once again, you abuse the word "chaotically". It has a specific technical meaning. If weather is chaotic, and we define climate as the average of weather on some timescale, then by definition climate will indeed behave chaotically. Secondly, yes I expect the hydrological cycle to vary within the constraints of the fractal dynamics that govern it, and this will impact albedo and radiative imbalance through clouds and water vapour. Since the timescale (coming decades) is beyond the point at which exponential error growth swamps our ability to predict, we do not know the path these will take.
Well, I need to get back to the day job today (UK holiday yesterday!) so I'm unlikely to be commenting further; I have internet access from my lab but I prefer not to post from there. And I agree you are unlikely to have an epiphany, since it is clear you still do not understand the basic concepts I am trying to convey. That may be in part my fault for not conveying them clearly.
To me, the arguments are not complex, but simple. The system is complex. But I am not arguing that complexity prevents us from understanding a system; I am arguing that it is necessary to view the system in its correct form (complex, chaotic) in order to understand it. There are hard limits of predictability of the system which can only be determined by recognising the systems true form. The climate activist approach to this is best described as an ostrich with their head in the sand; because they do not recognise the true form of the system, the cannot recognise the limits of predictability of the system.
And just to reiterate, there are already publications on the subject, and in terms of feedback from "actual experts" (ahem), they are stilling thinking about it (!). This topic has been discussed often, I remember when commentator "bender" was active on climate audit, perhaps 6 or 7 years ago, he predicted that once as climate failed to follow the predictions of models, that climate scientists would eventually embrace fractal dynamics as they would need the increased error bars associated with them to keep their models alive; of course accepting this would be expensive for them because many of the totems from climate science (including very basic definitions - "weather" vs "climate", "forcing" vs. "feedback") would need to be completely revisited. Science doesn't move that quickly unfortunately - it is made of humans who historically are very slow to change their minds once they are set on something - although the falsification of models may well come much sooner than either bender or I might have anticipated.
Spence_UK,
Quite possibly, but I think you're overstating the role of chaos and completely ignoring the basics of energy conservation. Pedantically arguing about mathematical definitions of the term chaos doesn't change that - despite you claiming I'm wrong - that most evidence suggests that our climate is forced, not unforced.
Either you and Bender are amazing polymaths who are seeing things that thousands of experts are unable, or unwilling, to see. Possibly, but in my experience, those who profess to be amazing polymaths are the least likely to actually be ones.
Well it helps, if we're discussing chaos, that we use technical terms correctly - otherwise communication simply becomes impossible. Furthermore, the constraint of energy conservation does not preclude fractal dynamics in the climate system, so that is a red herring. Changes in (say) cloud cover lead to changes in albedo lead to changes in global temperature. No violation of energy conservation required.
I have never professed to be a polymath, much less an amazing one, so I'm not sure where this comment has come from. Furthermore, I'm hardly alone in my analysis - many others note that there are problems with GCMs, even on this thread, plus the scientists I have already linked you to above.
As to precipitation, is it really controversial that precipitation is less well understood than temperature as model output? I'm surprised there would be disagreement on that point.