Golf Charlie, Steve Richards, supertroll

Did you ever play cricket?

Batting, you watched the ball and swung your bat to intersect it's expected path.

You used a mental model to project where the ball would be in the future.

If your model was correct, the ball went for four. If your model was wrong, the ball hit your stumps.

In all aspects of life you use models to project future events based on a mixture of available information and experience. The quality of the input determines the quality of the output.

There have been many models intended to project the future behaviour of a system. In the 1950s engineers calculated the design life of a Comet 1 airframe. They got it badly wrong because their model did not include metal fatigue.

Later they calculated the fatigue life of Vulcan wings. They got it wrong because they assumed that the aircraft would be flown mostly at high altitude. The RAF then changed tactics and flew them in the much harsher low altitude environment.

These two examples show two ways in which a projective model can be wrong. The first did not properly describe the physics because of an extra unknown process, the second was wrong because the future behaviour of the system changed in a way they could not predict.

In both cases the error is obvious in retrospect, but was not detectable when the projection was made.

What of climate models?

Most of the physics is old fashioned heat transfer. Effects such as cloud cover are still incompletely understood and are included as parameters.

Some variables are not known in advance. Natural changes such as cooling due to volcanoes ,ENSO etc are random. Artificial variables such as land use, population,pollution and fossil fuel burn, depend on future social and political decisions and are not predictable.

The solution to unpredictable variables is the ensemble.

Model runs use one of four fossil fuel burn options, RCPs.
RCP8.5 is "the business as usual" option. The other three assume different degrees of reduction.

Natural variables such as volcanoes and ENSO are randomised.

The result is a large number of model runs Those in which natural and artificial variables all favour warming indicate the probable upper warming limit the system is capapable of producing. Those in which the variables all favour minimum warming indicate the probable lower limit. The rest of the runs will come in somewhere between.

Note that all of these models will be wrong. The real world will produce CO2 at a rate which does not exactly match an RCP. The number and timing of volcanoes and ENSOs will not exactly match any of the models. The actual outcome will be somewhere in the range established by the ensemble, but exactly where is not predictable in advance.

How do you judge the performance of a model intended to make future projections?

The first way is to look into the past. Begin your model run from a date in the past, using the starting conditions for that date and run it up to the present using observed emissions and natural variables. Compare the model output with your records. This gives you a chance to tune the model. The physics and the events can't change, but you can adjust the values of parameters to get a closer match between runs and reality.

How do you judge the projections? Compare them with reality. The CMIP5 runs were calculated in 2005. Take the artificial and natural variables since then and look for a comparable model run. If the model and actual temperatures are similar, the model has worked. If they are not similar, the model has failed.

In reality the Earth followed RCP8.5 and went through a period of above average vulcanism and industrial pollution through the latter noughties. Then albedo decreased. This corresponded to a "cool" model run initially, with the last five years more like a mid-range run.

When you compare the reality with the ensemble results you get these graphs.

To my untutored eye they show that reality is following a path within the projected range ie the models and the reality agree. You see the observed data near the lower ensemble limit between 2005 and 2011, then moving towards the middle of the range.

Finger trouble. Graphs are here.

http://www.climate-lab-book.ac.uk/comparing-cmip5-observations/

Sorry EM. I don't agree.
"The quality of the input determines the quality of the output." Yes, but the output also depends upon the quality of the model itself. Place too much emphasis on one variable, mistake a forcing for an output, and you end up with outputs that do not match reality. Put in fudges to make your model hindcast correctly and then expect such a doctored beast to predict the future is very poor science.
Models should represent the very best understanding we have of the climate system. When we run them, departures from reality should be used for further understanding, not fudged yet again. The fact that we have several tens of models, each producing often very different outcomes should be something to be ashamed of. It indicates, more than anything else, just how poorly we understand the intricacies of climate. To assemble them all together in a spaghetti graph and proudly boast that reality lies within the variation is, to me, an admission of abject failure.

Steve Richards. I trust this spells out more accurately what I think of the current use of climate change models as long-term predictors.

EM, fine set of excuses and non-sequiturs. Nobody who understands fatigue life would accept the characterisation of the Vulcan wing design.. Nobody aware of any kind of scientific method would accept that climate models include every influence on climate. That's not to say they are useless, but they are only useful to refine climate modelling. Until it works, if indeed it ever can. However, the claim that the ensemble in some way gives a better quality result is totally spurious.

St, totally agree with you on your response.

Supertroll

" Place too much emphasis on one variable, mistake a forcing for an output, and you end up with outputs that do not match reality. "

Perhaps you could give examples. I keep seeing this on the propaganda sites, but never anything specific. I remember you once misclassifying the water vapour greenhouse effect as a forcing, when it is a feedback. It does not increase my confidence in your judgement, or theirs.

"To assemble them all together in a spaghetti graph and proudly boast that reality lies within the variation is, to me, an admission of abject failure."

I notice that you focus on the spaghetti graph, and not on the second graph which is much more informative. Perhaps you could explain clearly how that second graph shows that the models are wrong. Once again, I keep seeing that claim as propoganda, but without any scientific backup. Your inability to put proper evidence and analysis behind your opinion is unscientific.

Rhoda

"Nobody aware of any kind of scientific method would accept that climate models include every influence on climate. "

Nope, just the significant ones.

"Nobody who understands fatigue life would accept the characterisation of the Vulcan wing design."

The Vulcan was originally designed to drop free fall atomic weapons from high altitude against weakly defended targets. The fatigue life was calculated accordingly.

When the Soviet defences became capable of detecting and shooting them down, the V bombers switched to low-level penetration and trained to fly in under the radar coverage before popping up onto their targets. This put much more stress on the aircraft and shortened their fatigue lives accordingly.

The reason I mentioned it was to illustrate that the validity of a projection may change when the original assumptions change.

"However, the claim that the ensemble in some way gives a better quality result is totally spurious."

Really? The purpose of an ensemble is to produce a range of projections to allow for the potential range of future variables.

Why would this technique not give a better guality result?

Oct 27, 2017 at 1:30 PM | Entropic man

Sorry, but as a country bumpkin, I have not seen what Climate Science claims to have "observed". By the mid 1970s(?), I had been taught that the weather had been warmer and colder in the past, though I don't remember the terms MWP and LIA from then.

The Comet disaster, is a good example of a tragedy that could have been avoided, with a better understanding. If computer models had existed at the time, they would not have foreseen anything, as nobody knew about metal fatigue to programme it in. (Apart from author Nevil Shute? https://en.m.wikipedia.org/wiki/No_Highway )

"How do you judge the performance of a model intended to make future projections?" Well, Climate models still cannot predict or explain the MWP or LIA, so how reliable are they at hindcasting what has already happened?

With Climate Science modelling skills, it would be simple to prove that Climate Science has caused more death, misery, conflict and destruction than nuclear weapons, and cost more too.

If only Corporal Phil Jones had told presumptive Hockey Team Captain Mannwaring "Don't Panic!". Instead they listened to the grim undertaker, thinking of his potential income, saying, with relish, "We're doomed!"

Is Climate Science going to debunk Christine Lagarde for her 'We will be toasted, roasted and grilled' speech, now that Al Gore has fulfilled and outlived his uselessness? It does make the IMF and Climate Science look a bit stupid.

What I mean about fatigue lifing is that although fatigue lives are in hours, an hour at high-alt cruise doesn't count the same as an hour doing circuits or one doing terrain-following low altitude work. Each activity has a different value in fatigue hours per clock hour. If any mistakes were made it would be in that equivalence. Although I can't recall the Vulcan being particularly bad. Valiant was withdrawn form service, Victor relegated to tanking, B-47 had a massive wing spar bolt replacement programme and B-52 had a problem with losing the vertical tail. Not particularly relevant to climate models though.

Oct 27, 2017 at 4:28 PM | Entropic man

Why should anyone trust the spaghetti graphs (or anything else) produced by Climate Scientists, to support Climate Scientists, when the larger pasta shapes of the LIA and MWP have been liquidised by Climate Scientists, to support Climate Scientists?

On the usefulness of the ensemble, this was widely discussed in 2013 and I am convinced by RG Brown:

https://wattsupwiththat.com/2013/06/18/the-ensemble-of-models-is-completely-meaningless-statistically/

but of course that represents only one side. Wm Briggs has a different statistical view which seemed duff to me. I repeat, if the Canadian model runs too hot and has always run too hot, why can't anyone say that's obviously at variance with observations and chuck it out? That would be the sensible thing to do if a more correct result was a priority. But it isn't, a warmer result is desired and those Canucks deliver the goods! Plus it would be politically difficult to ditch them.

Rhoda, your link then leads here:

http://bishophill.squarespace.com/blog/2013/6/14/on-the-meaning-of-ensemble-means.html?currentPage=2#comments

and hence to these comments

JamesG, "So if you want to bitch-slap someone it has to be either Santer or Schmidt, neither of whom seem to care about correctness."
--------------

I'm keeping an open mind. The number could be more than two.
Jun 14, 2013 at 1:04 PM | michael hart

You'd need to bitch-slap a sufficiently large number and take the mean in order to be significant.
Jun 14, 2013 at 1:13 PM | rhoda

I am not sure that the science has evolved very much since, though the "sufficiently large number" needs to be increased, to match the expanding margins of error.

It's fun when you forget writing a joke comment then read it again as if fresh.

The concept of multiple model runs with different inputs and or different constant values fed into a fully verified model would be a sensible thing to do.

To run an unverified model other than for interest or development purposes is pointless.

To run multiple unverified models, multiple times is pointless squared.

Why would you do anything with an unverified model?

The purpose of a climate model is to be fixed, no more, no less.

Until it has been fixed (verified) it has no use, or purpose in life.

People who say that an unverified model gives clues to how the climate works are frauds, no less.

Can anyone here other than EM say what the use of an unverified model is?

I understand the fact that most people are not engineers or of a technical persuasion, but please, just tell us any useful purpose of an unverified model.

Steve Richards

How do you verify a model?

Can anyone here other than EM say what the use of an unverified model is?

Oct 27, 2017 at 9:08 PM | Steve Richards

In the absence of any proof or evidence, Climate Science has Mann's Hockey Stick, Cook's 97% consensus, and unverified models based on adjusted data.

The rest of Settled Climate Science is not as convincing.

'Can anyone here other than EM say what the use of an unverified model is?"

Steve, GC, you haven't been paying attention. If you've got enough of them the average must be right. I may just have over-paraphrased EM a bit there.

Oct 27, 2017 at 9:30 PM | Entropic man
Not for the first time, you (and others) are playing a dangerous game, and trying to shift the responsibility of proof away from those making unevidenced claims. Mann has already kicked off, and is now trying to play for time, indefinitely.

Why should Taxpayers pay for Climate Science? Why don't we just agree to test the Zero Expenditure Climate Science Model for the next 20 years, and see if anybody cares?

If Climate Science cannot prove or justify its own existence, whilst claiming the science is settled, it is not up to anybody outside of Climate Science to prove anything, or pay for it.

When will individual Climate Scientists be forced to repay their Taxpayer Funding, and compensate those who have found the errors that Climate Scientists and expert Peer Reviewers couldn't?

Rhoda,golf Charlie, Supertroll, Steve Richards

Yes. I get your point. My point was that a calculated life in clock hours for the Vulcan under one set of assumptions about its use then had to be reduced to fewer clock hours when the assumptions changed.

I chose two pre-computer and non-climate change models examples to make two points about sources of error in models. I know a bit about aircraft, so I tend to like aeronautical examples.

I even made my point explicitly.

"These two examples show two ways in which a projective model can be wrong. The first did not properly describe the physics because of an extra unknown process, the second was wrong because the future behaviour of the system changed in a way they could not predict."

. I take it that none of you actually read this paragraph, or you would not have missed my point so spectacularly.

My subtext was some thoughts about the nature of modelling and a point to golf Charlie that these are mathematical models. You do not need a computer. Calculations can be done on a slide rule, on paper, on your fingers or even in your head.

I even discussed how climate models are verified by comparing their projections with the past and with the early future years after the models were run. You missed that too.

Oct 27, 2017 at 10:47 PM | rhoda

You need to multiply by a variable fudge factor to arrive at the magical number that you had previously agreed to conclude, with a confidence level in proportion to the funding.

Rhoda

" I may just have over-paraphrased EM a bit there."

The projected temperature is not the average, though some might argue that the average is the most likely outcome.

What the ensemble does is set probable upper and lower limits for the temperature change between 2005 and 2050.

The projection is that when you look back from 2050 the observed temperature change will be within those limits.

In the first 12 years the observed pattern skirted the lower limit until 2011 and then jumped above the ensemble average. I see no reason why the remaining 33 years will be any smoother.

The shape of that graph will depend on El Ninos, La Ninas, volcanoes and how much fossil fuel we burn. None of these are predictable in advance. The final graph will probably not follow the average More likely it will bounce around between the limits as short term periods of faster or slowerr warming kick in.

Oct 27, 2017 at 10:58 PM | Entropic man

You keep missing the point that Climate Science has completely missed the MWP and LIA, since Mann invented his Hockey Stick. What are the Rules of Climate Science that ban you and others, from acknowledging the historical record? Do you lose financial support for being honest?

This abuse of Science was sanctioned by the IPCC, its cheerleaders and "Nobel Prizewinners". What level of greed, deceit, incompetence, or negligence is UNacceptable to the IPCC? It is all worthless, though phenomenally costly.

The Vulcan was designed for one purpose, and then required to do another. Climate Models were only designed for one purpose, to predict Global Warming. They are incapable of doing anything else, which is why they can't, and they were obsolete from inception.

There remains something of beauty about the Vulcan Bomber as a relic of the Cold War that people will pay to see for years. Climate Science on the other hand .........

As a Country Bumpkin, and knowing many other Country Bumpkins, I know I don't need a computer to measure Global Warming, or lack of it, Mother Nature tells me that.

Hubert Lamb did not need a computer to sketch out a record of temperature changes including the MWP and LIA. Mann DID need a computer to erase the MWP and LIA.

EM: How do you verify a model!

Oh dear, the 64,000 dollar question that should not need to be asked.

You test it. Always. You give it known inputs, run it and it gives you expect outputs.

The process works on all know modelling system .... except for climate models.

EM, how else did you think that you tested a model? Keep running it and hope it self corrects in some Star Trek way?

Come on, computer models or simulations as I know them as, have been around for as long a computing has.

We use them in most branches of engineering design.

If you have a model/simulator that does not give the expected outputs when stimulated with known inputs, it is faulty.

If you are actually trying to create a useful simulator that will do a job of work, then you analyse the failure and fix it.

Pretty normal engineering stuff really.

If however, you are spending taxpayers money, and you model/simulator does not actually need to work, then, whatever your model/simulator output is, you just publish and ask for more money.

Its a great but immoral workflow model.

EM, would you fly in a plane that was the result of a faulty model/simulator?

EM, would you drive a car whose breaking system had been modeled and manufactured via a faulty model/simulation?

It is beyond a joke that people blindly except concepts such a an ensemble of faulty models/simulations.

EM, you can get big bucks in the programming world doing verification and testing because it is: a) essential b) not glamourous c) the brightest and the best do not want to do it, they want to write exciting code.

I would like to say 'trust me' because I have done a lot of this sort of stuff over the years, but you will have no reason to except such a statement.

So I will say, 'do you really really believe that anyone can use a faulty model/simulation for anything?'

I can guess you may say, 'are you telling me that we should not use climate model simulations until they are verified'

Too right I am.

Faulty code is faulty code. No ifs or buts.

I can guess you may say 'the model/simulations are too complex to fix, we do not have a choice but to use them as they are, after all there seem to work a little bit'.

Look EM, if you went to an engineering programming house and asked for a model to be written when you did not understand its total internal mechanism, they could take one of two approaches:

Ethical approach - come back when you can give a complete definition of what you want us to implement

Climate modeller approach - ok, we will do what we can, the bits we do not understand we will guess or estimate, and if it looks a bit off we will keep reworking it while you keep paying us.

The whole world uses the first (ethical) approach, the climate world uses the second approach.

You could educate yourself it program test and verification, that could give you an insight to who important this is.

You could even look up software testing techniques in wiki: https://en.wikipedia.org/wiki/Software_testing

It will give you a flavour of the depth that engineers go into to ensure their product WORKS.

Modern techniques such as extreme programming has the odd concept to climateers, where the test is written before the program code is written! The program code is deemed working if the test passes!

Modern coders use tools to check their code is as fault free before it is even run: static analysis and dynamic analysis.

It is interesting to note that the last time I looked, the market leader for testing systems LDRA did not work with fortran!!!

Millions of pounds/dollars of climate code not checked to the highest testing standards in industry - unbelievable.

Are the climate models that are written in C or C++ tested exhaustively using the most effective tools available? I see no evidence.

All in all, climate modeling is a disaster of epic proportions. Everyone in the commercial computer industry thinks it is a joke.

EM, I hope you are detecting a trend in the above that all is not well in the climate modeling world. That slavish acceptance of what climateers state is extremely risky and with the slightest bit of research falls apart immediately.

Just look into the concept of testing software, ignore climate for a few days, just look at testing.

Testing of web programs, databases, financial transactions, flight control systems etc - then see what the climateers do...

You will be shocked.

I am about as far away from being computer literate as you can get. Yet it is possible to make some form of informed judgement. All sides of the argument agree that the topic under investigation is probably the most complex ever investigated. Another complex subject where mathematical modelling is applied is the economy. I ask myself if this attempt is successful, or if I would risk my pension investing in a computer modelled prediction. If I wouldn't, and I never would, for a lower complexity model, why should I have any confidence that a greatly more complex model predicting climate scores of years ahead will be correct?
Comments written alongside the computer code for a much simpler task that was revealed by Climategate did not confer confidence.