Guest post by Dr. Tim Ball
In a comment on the WUWT article about the abject failure of UKMO weather forecasts, “Slingo Pretends She Knows Why It’s Been So Wet!”, Doug Huffman wrote, “Each forecast must be accompanied by the appropriate retro-cast record of previous casts” (January 6, 2013 at 7:06 am). I pointed out years ago that Environment Canada (EC) publishes such information. They expose a similar horrendous story of absolute failure. This likely indicates why it is not done by others, but provides adequate justification for significantly reducing the role of the agency.
Both EC and UKMO predictions fail. The failure parallels Richard Feynman’s comment.
It doesn’t matter how beautiful your theory is, it doesn’t matter how smart you are. If it doesn’t agree with experiment, it’s wrong.
If your prediction (forecast) is wrong; your science is wrong. Unlike the IPCC, they cannot avoid the problem by calling them projections, not predictions. They can and do avoid accountability.
Initially I thought EC was admirable for publishing results. Now I realize it only shows arrogance and sense of unaccountability: we fail, but you must listen, act, and keep paying. It underscores the hypocrisy of what they do. More important, it shows why they and all national weather agencies must be proscribed. It is time to reduce all national weather offices to data collection agencies. When bureaucrats do research it is political by default. The objective rapidly becomes job preservation; perpetuate and expand rather than solve the problem.
EC is a prime example of why Maurice Strong set up the IPCC through the World Meteorological Organization (WMO) and member national weather agencies. EC participated and actively promoted the failed work of the Intergovernmental Panel on Climate Change (IPCC) from the start. An Assistant Deputy Minister (ADM) of EC chaired the founding meeting of the IPCC in Villach Austria in 1985. It continues, as they sent a large delegation to the recent Doha conference on climate change. Their web site promotes IPCC work as the basis for all policy on energy and environment. They brag about their role as a world class regulator. All this despite the fact their own evidence shows the complete inadequacy of their work.
They display their failures on maps. Pick any map or period and it shows how a coin toss would achieve better or at least comparable results. Here is their caption for the maps.
” The upper panel shows the seasonal air temperature or precipitation anomaly forecasts. The forecast are presented in 3 categories: below normal, near normal and above normal. The lower panel illustrates the skill (percent correct) associated to the forecast.”
The maps are for temperature and precipitation for 12, 6 and 1-3 months.
Everyone knows that regional weather forecasts are notoriously unreliable, especially beyond 48 hours. This fact weakened the credibility of the IPCC predictions with the public from the start. Some supporters of the IPCC position tried to counteract the problem by saying that climate forecasts were different from weather forecasts. It is a false argument. Climate is the average of the weather, so if the weather science is wrong the climate science is wrong.
Some experts acknowledge that regional climate forecasts are no better than short term weather forecasts. New Scientist reports that Tim Palmer, a leading climate modeler at the European Centre for Medium – Range Weather Forecasts in Reading England saying, “I don’t want to undermine the IPCC, but the forecasts, especially for regional climate change, are immensely uncertain.” In an attempt to claim some benefit, we’re told, “…he does not doubt that the Intergovernmental Panel on Climate Change (IPCC) has done a good job alerting the world to the problem of global climate change. But he and his fellow climate scientists are acutely aware that the IPCC’s predictions of how the global change will affect local climates are little more than guesswork. The IPCC have deliberately misled the world about the nature, cause and threat of climate change and deceived about the accuracy of their predictions (projections), for a political agenda.
Some claim the failures are due to limited computer capacity. It makes no difference. The real problems are inadequate data, lack of understanding of most major mechanisms, incorrect assumptions, and a determination to prove instead of falsify the AGW hypothesis.
Einstein’s definition, “Insanity: doing the same thing over and over again and expecting different results” applies. However, EC do the same thing over and over with results that indicate failure yet fail to make adjustments as the scientific method requires. What is more amazing and unacceptable is they use public money, are essentially unaccountable yet demand the public and politicians change their energy and economic policies. On their web site, they state; “The Government of Canada supports an aggressive approach to climate change that achieves real environmental and economic benefits for all Canadians.” They could begin by reducing EC to data collection. Their failures are more than enough to justify termination in any other endeavour. Another is their involvement and political promotion of well documented IPCC corruption.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
” It is a false argument. Climate is the average of the weather, so if the weather science is wrong the climate science is wrong.”
What if I say that in six months daily temperatures at my place will be about 20 degree celsius higher than they are now?
I know nothing about what weather there will be, but I know enough about my climate to place this forecast with sufficient certainity.
Do you really insist that my 6-month forecast is completely wrong because I can’t run a weather model that long?
You don’t need to simulate motions of individual particles to stude thermodynamics.
No, the problem is not that we must be able to simulate weather with absolute precision to be able to make climate forecasts.
The problem is that we still don’t understand climate. And it may take hundreds years of observation and research to get to some basic understanding, similarly how to it took many decades to get to today’s level of reliability for weather forecasts.
Tie compensation for all employees of the agency to the accuracy of the predictions. Saves 50% right off the bat and more if they use cruddy science.
“The met office says it does not believe global warming will be as severe as it had previously predicted.”
They are sounding ‘The Retreat’ and ‘Reveille” at the same time.
Adrian Kerton says:
January 8, 2013 at 3:15 am
Today, 8th Jan at 08.00 on UK Radio 4 news
“The met office says it does not believe global warming will be as severe as it had previously predicted.”
Nothing on the BBC website though.
Of course they have it. See Met Office Decadal forecast. They don’t go as far as to say it, they are just showing it to you. Compared to their previous forecast, which is gone, but not quite yet. Tricky.
Even more tricky is that they have changed their old forecasts retrospectively to match facts (while line). That’s what no scientist ever does.
“Who controls the past, controls the future: who controls the present controls the past”
Kasuha says:
January 8, 2013 at 1:26 pm
“What if I say that in six months daily temperatures at my place will be about 20 degree celsius higher than they are now?
I know nothing about what weather there will be, but I know enough about my climate to place this forecast with sufficient certainity.
Do you really insist that my 6-month forecast is completely wrong because I can’t run a weather model that long?”
Kasuha, the Null hypothesis for the weather in your place is probably that summer is about 20 degrees warmer than winter, so in that regard, you’re predicting what the Null hypothesis says.
“You don’t need to simulate motions of individual particles to stude thermodynamics.”
Like you, climate modelers need to use a statistical approach for their models. The statistical approach breaks down when the number of process instances per grid box becomes small. That’s why their approach cannot correctly approximate the behaviour of large scale processes like large convective fronts.
Richard Telford will tell you that all is well nevertheless, never mind the details, it’s a “boundary condition problem” so you don’t need much accuracy anyway… Well then the science should be available cheaply, if any ole shoddy model can do it; let’s start with paying climate modelers half the minimum wage plus a few food stamps (bring your own X-box).
DirkH says:
January 8, 2013 at 1:14 pm
If that is so easy,
——————–
Who said it was easy?
—————–
Richard, then why did the IPCC’s model fail even at the relatively simple task of modeling the past 16 years of stagnant temperatures.
—————–
Because, obviously, predicting the next few year’s weather is an initial value problem. To do that you need to initiate the models with observational data. Until recently, the models used by the IPCC haven’t even tried this.
————————
David L says:
January 8, 2013 at 7:48 am
@richard telford 3:33am
I see climate as the average of the weather. From a statistical standpoint weather is akin to the prediction interval, climate is the confidence interval. Has nothing to do with initial values or “boundry” conditions. It’s
just how we describe the individual data and summary statistics.
We all know the climate next summer will be hot and muggy in Philly. How hot and how muggy and specifically which days will be the hottest is anyone’s guess.
————–
Exactly. If you treat it as an initial value problem, you have know idea whether the end of next week will be warm or cold, and so impossible to predict as far ahead as next summer. As an boundary problem, it is much easier to predict that summer will be hot.
———————–
Pat Frank says:
January 8, 2013 at 10:46 am
richard telford, a specious argument. There will be growth of uncertainty due to initial value errors, combined with propagation of those errors through stepwise calculations that include serially re-initialized intermediate states, all of which employ a model also subject to theory-bias. When the propagated uncertainty exceeds the bounds of the model, the mean expectation value no longer has any physical significance; no predictive value whatever.
Using your coin toss analogy, a climate model prediction of future temperature is like attempting to predict the persistence runs of heads and tails in a sequence of coin tosses. The claim to know specifics of climate after hundred years is like claiming to know that the last 100 tosses of 10,000 tosses will all be tails.
————————————————
That is exactly what it is not like.
That would be to try to predict the weather on the 8th January 2113. Nobody will be trying to do this for at least 99.9 years. What the IPCC is trying to do is predict what type of weather can be expected in 100 years. That is an entirely different and massively more tractable problem.
This is not to say that initial values are not important for the new decadal predictions, but that in the long term boundary condition uncertainty is much the bigger problem.
Well done, Dr. Ball. Thanks!
A scam based on insufficient science is still a scam.
The Gray Monk says:
January 8, 2013 at 3:32 am
Bureaucrats are, by definition, overpaid and generally overqualified filing clerks. They should not be allowed to do anything other than keep the files in order, under no circumstances should they be allowed to “manage” anything and certainly not anything scientific.
While government scientists are often referred to as bureaucrats, most of the field-science staff are unable to influence opinions in the “boss” level mind. They are also temporary hires with little or no benefits, low pay and little chance of improvement. Most temporary “seasonal” staff are hired at GS-5 or GS-7 rates, which range from about 33 to 41 thousand per year. Because the hire is seasonal, the actual take-home can be roughly a third of that, or in other terms close to the poverty level. Been there, done it, left ASAP. By and large the US Gov’s treatment of its lower echelon staff makes Walmart look like a kindly, charitable, caring company where employees are concerned.
One of the more gruesome traits of government bureaus is that they could be more efficient, IF they actually hired staff, rather than contracted out work. The work isn’t cheaper because its done “outside,” and the results are generally neither better nor worse than government employees would produce. In fact, because of layers of overhead costs the contractor has to add on just to be sure slow business times are buffered, the same job done under contract costs more per hour and frequently is not as thoroughly done, since the contractor low-balled the estimate to begin with. Unhappily, once the work is completed, the analysis carefully done, and the results handed upstream for review, not infrequently comments come back such as “I don’t agree.” “Say [this] instead.” “Use the official government spelling for ….” [No joke there.] Arguing can lead to the door being indicated. Essentially, what happens inside a government bureau often resembles the alimentary process. What goes in is useful. What often comes out is only useful for spreading on gardens after a lengthy composting process.
Essentially Joel Shore claims the only thing you need to predict climate is the TOA forcing. So Joel, why do we need GCMs, why do we need massive computers? The fact is one wouldn’t even need a calculator. You could predict the climate with a slide rule.
Isn’t it nice to know we can fire all those people wasting their time and our money producing what Joel could do all by himself with a slide rule. When do we start?
LazyTeenager says:
January 8, 2013 at 4:37 am
“Medium term weather forecasts are not the same as climate forecasts.
Regional forecasts are not the same as global forecasts.”
Weather forecasts aren’t quite the same as climate forecasts because the latter are much worse.
Weather forecast use models with the forecast verified each day, the global forecast hasn’t even been verified once yet.
Since the 1970’s verified forecasts:- (weather v climate)
15349 v 0 (zero)
Climate is weather over a 30 year period and It is clear there is massive ignorance with how weather should behave in future. Weather patterns greatly depend on the jet stream, so this failure to predict includes failure for climate too.
Joel Shore says: January 8, 2013 at 11:38
“This is clearly seen in climate models…If you perturb the initial conditions, some things are sensitive to this and some things are not. Let’s make a list of what is sensitive and not sensitive to initial conditions —”
===================================
No, no Joel , you have it backwards. AGW types always get confused at this point.
Your observations are supposed to come from nature, and you should test climate models against these observations. If the model fails this test, i.e., by not replicating nature, then you are supposed to modify the climate model. Think about this, it’s an important principle. Let me know if you need any more help in understanding this crucial aspect of scientific prcedure.
There are two “redefinition lies” being accepted / used here. One is the assertion that “climate is the long term average of weather” or “the 30 year average of weather”. It is not.
Yes, the “climate science” folks have tried to hijack the term and turn it into that (for their own ends). It’s still not true.
Climate is a geography effect. Latitude. Altitude. Distance from water. Land form. Those geologic / geographic effects determine your climate. The Mojave Desert was still a desert when it was wetter and dryer, hotter and colder of the years. The Mediterranean Climate zone was still a Mediterranean Climate zone during the Roman Warm Period and during the Dark Ages / Migration era cold period. (And the Greek cold period and the Modern Warm Period and…)
It is this basic confounding of what is really climate with the “30 year average of weather” that is the heart of the basic lie of “Climate Change”. As there are known 60 year cycles in weather (PDO / AMO etc.) a “30 year average of weather” will be ramping up for 30 years, then down for 30 years. Wash and repeat. That’s “weather change” folks, not “climate change”.
Don’t drink the coolaid, don’t swallow the lie.
(One can make a case for adding the geologic change of precession, obliquity, etc. that causes the Ice Age cycle, which does shift climate zone enough to matter. I’ll worry about that fine point in about 2000 years… for all practical purposes it can be ignored during the lifetime of any one civilization.)
Once you realize that, and that the fractal nature of weather makes it ‘self similar’ at different scales, you see why shifting the ‘time scale’ of your weather prediction from 1 month to 1 decade doesn’t improve the accuracy.
The other “polite lie” is the idea that the TOA is a radiative boundary. This is enforced by the notion of a ‘pause’ in the world “tropopause” as though things are quiet and only a dead air radiative transfer is possible. It isn’t.
That space is running a Cat 2 Hurricane force wind sideways between the troposphere and stratosphere. (It’s about 1/2 that just a few thousand feet each side). There is also a large descending air mass at the cold pole (from the stratosphere level of altitude) in the polar vortex. That air had to get “up there” somewhere. Mass flow across the “pause” happens. A lot of it.
There are also intermediate zone descending air masses where various weather bands have rise / fall interfaces (Hadley Cells et. al).
There is no PAUSE at the tropopause, there is strong wind, mass transfer, and mixing. It is NOT a radiative only boundary. Arguing about how much of which light spectrum crosses the radiative boundary is a “MU!” question. “The question is ill formed!”
The first step to scientific accuracy (and freedom) is to recognize when the definitions have been cooked and get the more correct ones.
Look at the ‘wind speed by altitude’ graphs here:
http://chiefio.wordpress.com/2012/12/12/tropopause-rules/wind-speed-alt-1090/
Original source cited and discussion here:
http://chiefio.wordpress.com/2012/12/12/tropopause-rules/
The big peak wind speed is AT the tropo’pause’…
Once you buy into the ‘it is all radiative at that quiet stationary tropopause’ you end up sucked into endless useless dead end bickering over how that non-real place does that non-real thing and no one can ever be right about it. It’s an ‘angels and pins’ world then.
Once you realize it is a very very fast wind zone mass transport mixing zone, it becomes much easier to see why “IR transfer” doesn’t matter at that point.
So stick to Kopen climate zones for the definition of real climate and avoid the ‘pause’ trap. The world is then much easier to “get right”.
Joel Shore says:
January 8, 2013 at 11:38 am
Unraveling Joel’s response:
(1) The boundary conditions that are prescribed are not “future surface conditions”. They are the basic top-of-the-atmosphere energy balance conditions involving radiation emitted and absorbed by the Earth system.
NO. NO. NO. In addition to TOA, you must have surface boundary conditions for ocean and land boundaries (e.g. temperature or heat flux, velocities, mass/species fluxes etc.). How fast or slow the atmosphere/ocean heats up or cools down is critically dependent on this. As the numerical solutions are time marched, how are you setting these boundary conditions? Are they fixed? Do they change? If they change, what assumptions are being made? Do you know what they are decades out?
(2) The point is simply this: Although the detailed evolution of the weather…and even of the climate (in the sense of whether it will be warmer or colder or wetter or drier than average) over, say, monthly to yearly timescales…is sensitive to the initial conditions, the future climatic conditions in response to increases in greenhouse gases is not.
You are assuming that the differential equations and initial/boundary conditions that supposedly model the climate system are accurate representations of the system. The only evidence I’ve of any accuracy at all is seen in tuned hindcasts. Of course, some research organizations such as NASA/GISS don’t even document for us (1) what differential equations they are solving and (2) what numerical methods are being applied and their specific formulations (their code is pretty bad too). GFDL and NCAR are much better in this respect. I’m sure the European models are better too, although they won’t release their source code as far as I know.
(3) This is clearly seen in climate models…If you perturb the initial conditions, some things are sensitive to this and some things are not. Let’s make a list of what is sensitive and not sensitive to initial conditions —
Please cite references…do you have comparison of long term predictions for the temperatures at a specific place like Rochester, NY versus data? That would be interesting.
(4) It is thus not particularly scientifically literate to make the claim that because we can’t predict the weather a few weeks in advance or because we can’t predict whether this spring will be unusually warm or cold or wet or dry in some region, it therefore follows that we can’t predict the seasonal cycle or the response of the climate system to a significant increase in greenhouse gas concentrations.
Actually, even the Farmer’s Almanac can predict the seasonal cycle. But you can’t predict with any certainty the magnitudes of temperature or precipitation changes 1, 2, 3 years out, or for that matter decades out, with computer – they simply have NOT demonstrated any skill.
In any case, my thesis is still valid, namely that the climate problem as formulated and solved is an initial value problem, and it was pretty dumb of Richard Telford to say otherwise.
Richard Telford:
At January 8, 2013 at 3:33 am you asserted that climate modelling is relatively easy because you said
At January 8, 2013 at 4:32 am I rebutted that and explained
I expanded on that by using the ‘coin toss’ analogy you had used in your post.
At January 8, 2013 at 5:49 am you mentioned but evaded my rebuttal of your fallacious assertion.
Subsequently, at January 8, 2013 at 2:48 pm at you have repeated your assertion saying.
As my post – which you evaded – explained, it is not relevant whether or not “long term boundary condition uncertainty is much the bigger problem”. This is because
the boundary conditions cannot be defined and, therefore, their change cannot be determined.
Simply, climate modelling is based on the false premise – which you promote – that the boundary conditions of an inadequately defined system are known. They cannot be known when the system is little understood.
Richard
mpainter says:
How exactly do you think that Edward Lorenz originally discovered the phenomenon of chaos and the hypothesis that Earth’s atmosphere shows chaotic behavior? ( http://en.wikipedia.org/wiki/Lorenz_system ) Are you under the impression that Lorenz ran an experiment where he made an exact duplicate of the Earth, perturbed the initial conditions and then compared the weather on that Earth II to the weather on our Earth?
That is not in fact what he did. He wrote down some equations that, by modern standards, could at best be described as a cartoon of atmospheric dynamics (much, much, much cruder than modern climate models), and he found that when he perturbed the initial conditions in this model and ran the program for a while, he got very different results than with the original initial conditions.
And, since what we are interested in is whether perturbation of initial conditions in our models affects the result for the change in climate due to a rise in greenhouse gases, the best way to see if the results change is to run the models with perturbed initial conditions and see what happens. (We can also use our knowledge, gained through scientific study, of what sort of physical problems tend to show extreme sensitivity to initial conditions and what don’t.)
[I suppose one could hypothesize that in the real world, the final climate in response to a rise in greenhouse gases would depend very sensitively on the initial conditions even though it doesn’t in the models, but I know of no evidence to support that hypothesis…and the basic understanding that we have of under what situations chaotic behavior occurs leads us to believe otherwise.]
Frank K. says:
We know in principle what sorts of relations have to be satisfied at the boundaries, e.g., because of conservation of mass, conservation of energy, etc. (And, of course, if we have a coupled atmosphere-ocean model then the atmosphere-ocean boundary is not an external interface anyway.) And, even if there are uncertainties associated with the boundary conditions or various processes, that does not mean that the problem in principle has the sort of sensitivity to initial conditions that is the characteristic of chaotic systems. My whole point is that there are two different questions:
(1) Are there uncertainties in regards to modeling the future of the climate system under an increase in greenhouse gases?
(2) Is the problem intrinsically intractable because of the chaotic nature of the system?
The answer to the first question is Yes. But, the answer to the second question is No.
First of all, that is not a binary question. The question is always one of degree, i.e., how accurate the representation is. Making it a binary question assures us the answer is always NO, as it is for any model of any system.
Second of all, we know from the success of numerical weather prediction and other scientific evidence that the basic equations are accurate representations of the system to a good degree.
Third of all, the model does not have to be completely accurate in all respects in order to answer some basic questions about the system like which sort of problems will show extreme sensitivity to initial conditions and which sorts of problems will not.
The point is simply this: The seasonal cycle is an example of something that can be predicted that does not show extreme sensitivity to initial conditions. A problem like how the climate will change in response to increases in greenhouse gases (over periods of time long enough that the effect of this perturbation dominate over other fluctuations) is also of this type. And, indeed, when the modelers perform an ensemble of different runs with different initial conditions, they find that although the initial conditions affect the fluctuations up and down in the global temperature (and hence the climate over the next few years), they do not significantly affect the resulting change in global temperature over time periods long enough that the change is dominated by the radiative forcing due to the increase in greenhouse gases.
Do you think that the problem of determining the seasonal cycle is also an initial value problem? What evidence do you have to support your view in light of the evidence to the contrary, i.e., the evidence I have talked about above that the predictions that the models make for future warming are not sensitive to the initial conditions (whereas the predictions of the weather, say, a few weeks out or of the climate fluctuations over the next few years are seen to be sensitive to these initial conditions)?
I am really puzzled why these simple notions are so controversial to you guys. It is not as if saying that the predictions are not extremely sensitive to initial conditions means that the predictions are necessarily accurate. It just means that the response is not inherently unpredictable in the way that things that depend sensitively on the initial conditions are.
For those interested in an executive level description of how climate models work, this is a great resource:
Climate Models: A Primer
William O’Keefe
Jeff Kueter
2004
Andrew Weaver, Canada Research Chair in climate modelling at the University of Victoria says:
“Decadal predictability today is kind of what seasonal predictability was 15 years ago. Now we routinely look at El Nino forecasts, we routinely look at seasonal forecasts, and they’re very good…”
Dr. Tim Ball has shown that Prof. Weaver doesn’t know what he’s talking about.
http://news.nationalpost.com/2013/01/08/global-warming-hasnt-stopped-but-it-has-stalled-says-new-prediction-from-british-national-weather-service/
I dabble in applied maths for a hobby and liked Richard Telford’s suggested solution to the conundrum. This is a very basic concept in the calculus.
http://en.wikipedia.org/wiki/Boundary_value_problem
I also liked Sceptical’s post. (Mime casting a fishing line, getting a strike and reeling in.) Some people ain’t got no sensayuma.
joeldshore says: January 9, 2013 at 6:32 am:
===============================
I, of course, am not a climate modeler. My scruples would not allow it. This Lorentz perhaps was fundamental to present day climate modeling, which has built on and improved what Lorentz presented forty years ago, presumably. Well, forty years and umpteen $ billion later, I think that we are in a position to draw conclusions about climate models. Time and again, one sees verification of the climate models put in terms of… what the models show, such as in your last comment to rebut me. Never before had I imagined that scientists were capable of such fatuity. The trouble is that you only talk to each other and thus the illusion is thereby maintained. You will doubtless go back to to solicit more support from your modeling confraternity and return to add yet another card to the ediface. Look closely Joel, that house of cards is leaning and on the verge of collapse.
E.M.Smith says:
January 8, 2013 at 8:30 pm
Despite my insertions that climate is weather during a 30 year period, this was just the general idea and not necessarily the correct one.
I do agree that because 30 years is only half of the very noticeable 60 year cycle, it can only be weather not climate. Roughly 30 year alternate cooling and warming periods are observed during the temperature records, so when one half is warming or cooling this is half of a natural cycle and shouldn’t be called climate. A minimum 60 year period therefore should only be called climate and cover both warming and cooling parts of the shortest natural observed cycle.
Error in previous post, not sure how exertion become insertions.
For some reason my link got mangled. Let’s try this…
Climate Models: A Primer
Frank K. says:
We know in principle what sorts of relations have to be satisfied at the boundaries, e.g., because of conservation of mass, conservation of energy, etc. (And, of course, if we have a coupled atmosphere-ocean model then the atmosphere-ocean boundary is not an external interface anyway.) And, even if there are uncertainties associated with the boundary conditions or various processes, that does not mean that the problem in principle has the sort of sensitivity to initial conditions that is the characteristic of chaotic systems. My whole point is that there are two different questions:
(1) Are there uncertainties in regards to modeling the future of the climate system under an increase in greenhouse gases?
(2) Is the problem intrinsically intractable because of the chaotic nature of the system?
The answer to the first question is Yes. But, the answer to the second question is No.
First of all, that is not a binary question. The question is always one of degree, i.e., how accurate the representation is. Making it a binary question assures us the answer is always NO, as it is for any model of any system.
Second of all, we know from the success of numerical weather prediction and other scientific evidence that the basic equations are accurate representations of the system to a good degree.
Third of all, the model does not have to be completely accurate in all respects in order to answer some basic questions about the system like which sort of problems will show extreme sensitivity to initial conditions and which sorts of problems will not.
The point is simply this: The seasonal cycle is an example of something that can be predicted that does not show extreme sensitivity to initial conditions. A problem like how the climate will change in response to increases in greenhouse gases (over periods of time long enough that the effect of this perturbation dominate over other fluctuations) is also of this type. And, indeed, when the modelers perform an ensemble of different runs with different initial conditions, they find that although the initial conditions affect the fluctuations up and down in the global temperature (and hence the climate over the next few years), they do not significantly affect the resulting change in global temperature over time periods long enough that the change is dominated by the radiative forcing due to the increase in greenhouse gases.
Do you think that the problem of determining the seasonal cycle is also an initial value problem? What evidence do you have to support your view in light of the evidence to the contrary, i.e., the evidence I have talked about above that the predictions that the models make for future warming are not sensitive to the initial conditions (whereas the predictions of the weather, say, a few weeks out or of the climate fluctuations over the next few years are seen to be sensitive to these initial conditions)?
I am really puzzled why these simple notions are so controversial to you guys. It is not as if saying that the predictions are not extremely sensitive to initial conditions means that the predictions are necessarily accurate. It just means that the response is not inherently unpredictable in the way that things that depend sensitively on the initial conditions are.
I haven’t read that in detail and it may give some good information but readers should be aware that it is written by two people whose main credentials seem to be that they are the President and the Executive Director of an advocacy group, not scientists who actually work with climate models.
mpainter says:
Lorenz’s work was not only fundamental to numerical weather prediction and climate modeling. It was the first real identification of chaos, i.e., extreme sensitivity to initial conditions in any system…period. See here http://en.wikipedia.org/wiki/Edward_Norton_Lorenz for Wikipedia’s bio on him.
[And, by the way, I am not a climate modeler either. I am a physicist who has done modeling of a variety of physical systems.]
No…The models are not verified in terms of what they models show. However, making models of physical systems is fundamental to our understanding of these systems. And, in this particular case, one can test directly whether or not the predictions that the models make are or are not extremely sensitive to the initial conditions by simply running the models with perturbed initial conditions. And, the answer is that certain things do show such sensitivity to the initial conditions and other things (like the seasonal cycle or the basic global temperature response of the climate over long enough periods of time to a significant radiative forcing due to increasing greenhouse gases) do not show such sensitivity.
And, our understanding of the general situations in which one sees extreme sensitivity to initial conditions and when one does not, gleaned by both experimental and theoretical studies of chaos, lead us to understand why this is true and why we think it is true not only within the models but also in the real world.