Click images for more details



Recent comments
Recent posts
Currently discussing

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

Powered by Squarespace

Discussion > Let's get real about climate models

Martin A: I remember, a few years back, watching a TV programme that included attempts to model the flow of the Aghulas current, off South Africa (easily monitored by satellite). Despite the simplicity of the concept, and the limited amount of data perceived to be required, it was only 3 days before the predicted flow was completely different from the true flow – in other words, the model did not work. Now, expand that exercise to incorporate the Earth’s atmosphere and couple in the vastly greater number of variables that could be influential, as well as the perhaps even greater number of unknown variables that are influential, and it surely should be obvious to all but the most simple that it is a task that might never be satisfactorily accomplished, and certainly not one that should feature too heavily in any future policy, let alone present-day policy.

Jan 8, 2016 at 8:17 PM | Registered CommenterRadical Rodent

Indeed, RR. That goes to the heart of the problem. The scientific and engineering world is replete with people who have tangled with models of much simpler systems than the whole of the planet, and they know just how improbable it is that climate scientists have discovered the philosophers stone of modeling.

Jan 8, 2016 at 8:31 PM | Unregistered Commentermichael hart

Raff - The model stuff is common knowledge. A validated model is virtually the same as a validated formula or algorithm. It gives the right answer because the relationships are fully understood.

A provisional model is good for speculation and learning but it is not to be used to justify actions because it could be wrong. Models can give the right answer in a particular circumstance but still be completely wrong generally which is why models used for prediction are either validated or scrapped. Only climate models fall into the unique category of being unable to predict properly, invalid and still be used for extremely serious, far reaching and expensive predictions.

There is an accountability question here that has never been explored.

If you are referring to your link, I notice that the anomaly scale on the graph does not have error bars and the entire warming for two decades is about 0.2 degrees. We are only half way through the second decade and the global temperature record so far does not show any significant change in lower tropospheric temperature.

To put my case more simply, I think that a large proportion of the warming during the period when CO2 warming became an obsession was due to very high solar activity and only part of it due to CO2.

This was when the models established the basic relationships and all of the warming was attributed to CO2 and this became enshrined in the models.It was wrong. I personally think it was breathtakingly arrogant to ignore all previous fluctuations in temperature in the pre-industrial period and also ignore all natural drivers of climate in the industrial period. I think the evidence today makes it clear that all of that was wrong, yet climate scientists carry on in some sort of denial of reality.

Jan 8, 2016 at 9:39 PM | Unregistered CommenterSchrodinger's Cat

The climate system is a coupled non-linear chaotic system, and therefore the long-term prediction of future climate states is not possible.
How, exactly, do you propose to create a model that can be guaranteed reliable enough to be used for making policy?
Given that climate is essentially regional rather than global how do you ascertain, in a chaotic system, what policies need to be enacted in any given region?
Given that virtually every one of the existing models has been designed to exaggerate likely feedback effects while recent research is tending towards low sensitivity and that the number of weather events attributable to "climate change" approaches infinity, explain in as simple language as possible what advice you plan to give to governments.
And all that assumes that global warming is not simply a ready made excuse for the eco-warriors to bully us all into not using fossil fuels because they have this totally erroneous belief that we would all be better living in yurts and knitting yogurt while clad in animal skins (except that the vegans would complain — grass skirts in Scotland this weekend, anyone?) and dying at 35!

Jan 8, 2016 at 10:35 PM | Registered CommenterMike Jackson

Am I correct in thinking that to qualify as a CMIP5 model one is required to accept a set of assumptions/parameters/constants/process descriptions in common with the rest of the ensemble?

If so, who decides the requirements? Who tests the set against reality? Is in fact 'the physics' merely a set of formulae rather than derived from any actual basic physics within the model itself? Now, if I AM describing the way things are done, I can see how it would be convenient and pragmatic to do it that way. And defensible to combine ensemble results. But there are big problems with that approach if you are interested in finding out how weather and climate work.

Or am I wrong. I can be, you know.

Jan 8, 2016 at 11:51 PM | Unregistered Commenterrhoda

I don't have the knowledge of all the existing models nor of modelling in general to judge whether what you say is true. But if it were true that even 20 years of testing would not give a result that could be considered reliable enough to use to decide 'policy' you are effectively saying that all of the computing we can now throw at the problem gives us no more than we had 50 years ago. I imagine your response to be, well that is true, climate sensitivity is no more narrowly constrained now after modelling than it was 50 years ago.

If this were true, would you find it comforting or worrying?

Jan 9, 2016 at 12:07 AM | Unregistered CommenterRaff

What have all the climate models ever done for us?

1. They have been used to dictate policy, when they have no predictive skill.

2. I give up. Why doesn't climate science do the same?

Jan 9, 2016 at 1:28 AM | Unregistered Commentergolf charlie


You touch on a very salient point. How much or what type of modelling can be done that could be useful?

Total system modelling is going to be fruitless as this discussion is saying. Subsystem modelling or modelling of certain effects may be better. The idea could be put forward that if you know one effect well and that effect dominates then the system will tend to show behaviour corresponding to that effect. It's not a fullproof idea but at least it provides perspective.

Jan 9, 2016 at 9:37 AM | Registered CommenterMicky H Corbett

Micky H Corbett
Do you need a model to forecast what type of weather patterns will result if and when there is a reversion to previously experienced situations? For example the last few years the Jetstream in the Northern Hemisphere has reverted to the same pattern it had during the late 1970s. The only difference being that last time the changes were blamed on the coming iceage. The fact that models cannot predict when the system will flip back to the alternative state doesn't mean that it won't happen. Once it does then who needs a model to know what the prevailing conditions will be?
1977 Jetstream link from Steve Goddard

Jan 9, 2016 at 1:08 PM | Unregistered CommenterSandyS

As the drive to build progressively more expensive climate models, has come from progressively more numerous and expensive climate scientists, and still nothing useful has been produced, the problems clearly all originate with climate scientists and their alleged science.

The Met Office gave up on Long Range Weather Forecasts, 20+ years ago, because they were rubbish. Why they, and others, thought they were capable of forecasting the climate, is beyond me. We should scrap climate modelling and science. Those climate scientists who think they have something to offer, could be re-employed working for local TV Radio and newspapers on monthly weather forecasts. If they are any good, people will be able to vote for them, or vote them out. "I'm a Celebrity Climate Scientist" "Get him out of here!". If Cooking, baking, gardening etc can be revitalised, and made competitive viewing, why not weather forecasting?

Does anybody know whether ACCURACY of weather forecasting, was one of the selection/elimination criteria, for the BBC's new weather forecaster?

Jan 9, 2016 at 3:22 PM | Unregistered Commentergolf charlie


That's a good example. It would depend on how useful a model would be in helping mitigate damages.
Or maybe it helps provide better discipline in tracking weather events to help gain early warning signs if that's possible. I'm sure a case could be made for some sort of modelling. You would hope it would constrained to budget and reason though.

Jan 9, 2016 at 4:44 PM | Registered CommenterMicky H Corbett

Leave out "climate" and ask rather at what point you would submit your child, healthy or not, to a new invasive medical procedure of unproven utility?

Jan 9, 2016 at 5:52 PM | Unregistered Commenterhunter

As the drive to build progressively more expensive climate models, has come from progressively more numerous and expensive climate scientists, and still nothing useful has been produced,
Jan 9, 2016 at 3:22 PM golf charlie

STILL Epic Fail: 73 Climate Models vs. Measurements, Running 5-Year Means

73 plus and all pretty much in agreement. Even if the first three had some purpose, hard to imagine what benefit to humanity was imagined for the following 70 plus. Except to keep climate scientists and Fortran programmers occupied.

WUWT: How reliable are the climate models?

Jan 9, 2016 at 7:53 PM | Registered CommenterMartin A

Martin A, interesting links!

Since ClimateGate and the revelations of incompetence contained within the 'Harry Read Me' files, billions have been spent, and we would have been better off paying peanuts (and some bananas) to a troop of chimpanzees, each with a coin, one side coloured red for warmer, and the other blue for cooler, and just averaged the results.

Jan 9, 2016 at 9:32 PM | Unregistered Commentergolf charlie

Martin, Spencer is known for fiddling with baselines for best effect so your link has bad provenance. But I'm having trouble reconciling your Spencer graph with this later Spencer graph:

Also you claim to understand that various forcings are not predictable (solar cycles, volcanoes, emissions) by a model and I assume you accept that the forcings that in fact occurred since 2005 were lower than the models in those graphs assumed. So you are showing a graph that doesn't match a later graph and doesn't embed your understanding of how these things work. "golf charlie" liked it so I guess you scored. But he or she probably didn't understand what I just said.

Jan 10, 2016 at 1:28 AM | Unregistered CommenterRaff

Raff, do please show us some "provenance" for your assertion that Spencer is "fiddling with baselines for best effect".

Jan 10, 2016 at 7:22 AM | Unregistered Commenterhunter


How many computer generated climate science models predicted the pause? None.

Have any computer science generated models actually served a purpose? No

Why should tax payers have to pay for climate scientists to play computer games, when they could buy their own 'Playstation', and play games in their own time?

As a surveyor/engineer I can look at all sorts of design failures going back hundreds of years. Computer aided design has made lots of complicated calculations possible at the click of a mouse, but it has reduced the safety margins, and increased the reliance on accurate data inputs. No programmer can claim to have a proven track record in climate models, so why should anyone place taxpayers confidence and money in their output?

"Game over!", "Press Start to Play Again?" is not the mentality for designers playing with other people's money and lives.

Climate models should contain stark warnings on the box. "Reliance on Computer Generated Climate Models can seriously damage the health and wealth of everyone"

Jan 10, 2016 at 11:25 AM | Unregistered Commentergolf charlie

For anyone who didn't see it, John Christy's recent testimony to the US Senate is well worth a read. He shows evidence that the models are wrong.

Jan 10, 2016 at 12:53 PM | Unregistered CommenterSchrodinger's Cat

The late Ronald Reagan came in for criticism when his Wife Nancy revealed she consulted an Astrologer, before he made important US policy decisions.

President Obama has John Holdren, who thinks the odd 00.0010% of carbon dioxide orbiting the planet, is really really important, and we must destroy lives to get rid of it.

Jan 10, 2016 at 2:35 PM | Unregistered Commentergolf charlie

The blocking of SW IR radiation from the troposphere to space by carbon dioxide leads to warming. This may be true, but kinetic energy transfer via collisions will lead to the excitation of the much more abundant water vapour which radiates the excess heat to space.

The carbon dioxide focussed thinking of the climate community does not seem to include the possibility of negative feedback.

Jan 10, 2016 at 2:49 PM | Unregistered CommenterSchrodinger's Cat

Has anything changed in the world of the GCMs since this paper got published in 2013?

Jan 10, 2016 at 2:54 PM | Unregistered Commenterdiogenes

Because if nothing has changed, it suggests that understanding of climate is so poor that the models are not good as a basis for policy. Since they are so consistently wrong on the high side, a rational sceptic would assume that there is nothing to be worried about. That might be an unfounded assumption but the models do not support any other conclusion. In similar situations, Raff is wont to argue that nobody looks at or cares about GCMs - remember his tergiversations on the hockey stick? So why is he now so interested in this scientific cluster fuck?

Jan 10, 2016 at 5:29 PM | Unregistered Commenterdiogenes

Cat, Christie's graphs are clearly marked as showing Mid Tropospheric temperatures. UAH and RSS attempt to measure Lower Tropospheric temperatures. This is what Christie calls an "apples-to-apples" comparison. Do you agree with him? He also doesn't mention that the modeled forcings don't match the actual forcings. Like Martin above, he must be aware of that. So what do you think Christie or Martin achieve with their graphs apart from misleading readers?

Hunter, baseline selection has a direct effect on the graphs presented. Choose a short baseline where the datasets you are presenting differ significantly and you can shift one dataset up or down relative to the other. This is what Spencer does with his 5 year 1979-83 baseline.

Jan 10, 2016 at 5:53 PM | Unregistered CommenterRaff

Raff - His balloon temperatures are presumably mid troposphere and he does claim apple to apple comparisons though he doesn't give details. The graph has been around for a long time and I'm not aware of anyone challenging it apart from you.
I very much doubt that he would present misleading data to the senate since I think that breaks US laws.

I'm sure the modelled forcings don't match anything, this is probably why the model output doesn't match actual temperatures. The models treat IR blocking as equivalent to solar warming of the surface. This may satisfy energy balance but it is nonsense mechanistically and leads to incorrect feedbacks and creates incorrect predictions such as the famous hotspot which has never been found.

Jan 10, 2016 at 6:44 PM | Unregistered CommenterSchrodinger's Cat

So many issues here....

Climate models - will they ever be 'good enough'?

I suspect not, why?

We do not understand the basic processes involved at sufficient detail to create a model.

We do not know the starting data to feed into the model.

The above two issues are each individually fatal.

Here is a reasonable explanation of parameterisation, where whole bunches of interactions are just lumped together because we do not understand the detail to do the job properly.

These three problems are probably unsolvable, due to inaccurate historic data and insufficient knowledge of climate physics and climate biology.

When I wrote simulations of electronic circuits, you would have been laughed out of the lab if you suggested producing a model of a complex pcb ( think NATS ATC, or naval radar signal processing functionality) each of which contained hundreds of complex digital devices, without first verifying EACH component used in the simulation.

Each individual component writing/verification process could take hundred to thousands of man hours to complete.

The upside was, the resulting system simulation was so good, that we could inform the design engineer that his design did not work, and why, before it was even built!

If climate modellers wish to produce a climate model that was useful in some way, then the basic equations used need to be exhaustively proven/verified individually in a 'test bench' where every possible set of good data and bad data can be thrown at them and they always give the correct answer (verified manually). Note: for this to happen we need to 'understand' climate processes at such detail that it may not be worth starting the process off.

Once the basic equations have been verified/proven, they need to be assembled into a climate 'cell' then connected to the test bench and tortured without a single failure. Obviously, just like individual equations, the complete cell functionality needs to be understood so well so that when we run the test bench and feed in inputs and OUR manual predicted outputs, the cell model predicts the same.

To summarise, my view is that climate modeling will remain an interesting activity that advanced states can afford to fund a couple of models as an adjunct to proper physical climate research which may move us forward to understanding more about basic climate physics.

Jan 10, 2016 at 7:53 PM | Unregistered CommenterSteve Richards