Jones:

I’ve been told that IPCC is above national FOI Acts. One way to cover yourself and all those working in AR5 would be to delete all emails at the end of the process

[Response: Bradley and Cook are entitled to their opinion about anybody's papers. Barnett is overstating the degree of agreement in the CMIP3 20thC runs and is wrong about the nature of the tuning that occurs. Jones is correct in both statements - models are all wrong (but the question is whether they are useful), and FOI legislation does not cover the IPCC and is not a document retention law. - gavin]

Comment #54 http://www.realclimate.org/?comments_popup=9931

http://www.spiegel.de/international/world/0,1518,686697,00.html

Finally, didn’t those wall st banks have all sorts of sophisticated models? Worked well for them.

Climate models, regardless of sophistication and the the amount of computing power, cannot provide credible projections of climate for ten years or longer periods. The influence of known unknowns can be guessed, but correct outcomes are unlikely since the unknowns are not well understood.

Modeling the unknown unknowns requires great imagination and incredible luck to even identify their possible existence. Successful modeling is remotely possible. Failure is inevitable.

Why climate models have any credibility whatsoever can only be explained by Barnum’s axiom.

“Scientists’ use of computer models to predict climate change is under attack” (Washington Post)

The following is an excerpt from the above Washington Post article:

“But scientists say that, during this time, they have only become more certain that their models work.

Put in the conditions on Earth more than 20,000 years ago: they produce an Ice Age, NASA’s Schmidt said. Put in the conditions from 1991, when a volcanic eruption filled the earth’s atmosphere with a sun-shade of dust. The models produce cooling temperatures and shifts in wind patterns, Schmidt said, just like the real world did.

If the models are as flawed as critics say, Gavin Schmidt, a climate scientist at NASA’s Goddard Institute for Space Studies said, “You have to ask yourself, ‘How come they work?’ ”

“The Iowa Method, that’s why they work,” would have answered my late friend, Senior Master Sergeant Robert Kenneth Clough, US Air Force. According to Sgt. Clough, the Iowa Method is a brilliant model of simplicity and accuracy: you start with the desired conclusion, and then figure how you got there.

In short, the Iowa Method is infallible.

Apparently that’s the way climate scientists like Gavin Schmidt feel about their computer models.

Do you want an Ice Age? Take the one that occurred, and build a model that you declare replicates it. The model will always create the Ice Age that occurred, although it may not create the one that’s coming.

Does the computer model of the last Ice Age begin with a precipitous drop in atmospheric carbon dioxide? It should, if the computer model predicting current rapid warming based on rising atmospheric carbon dioxide is also the Ice Age model.

The same is true for a volcanic eruption in 1991. It may seemingly explain the subsequent cooling, but what explains the past decade of cooling without volcanic eruptions? Kevin Trenberth in a Climategate e-mail says it’s a “travesty” that our climate science (based on computer models) doesn’t explain the current lack of warming.

A major segment of climate models is based on paleoclimate reconstructions by Michael Mann, Keith Briffa, Phillip Jones, et al. In essence their studies, heavily weighted to analyses of tree rings, show very little variation in global temperatures for the past thousand years, then a rapid warming in the last half century corresponding to increased atmospheric carbon dioxide.

Inconveniently, their tree-ring reconstructions show cooling after 1960 instead of warming, so researchers conveniently discarded that portion of their research and replaced it by grafting on instrumental records to “hide the decline.”

Doing this obviated the need to explain why the tree-ring proxies for temperatures were good until 1960, then not good thereafter.

While ethically problematical, their approach satisfied the dictates of the Iowa Method: show that global warming did not begin until atmospheric carbon dioxide increased.

Non-scientists like Albert Arnold Gore, Jr. grabbed hold of this “correlation” by first stating that current warming was “unprecedented” for the past thousand years, then by doubling this claim to the time of Jesus Christ.

Again inconveniently, Über-climate alarmist Phil Jones recently admitted that the Medieval Warm Period (900 to 1400 AD) may have been warmer than present, and that global climate had cooled recently.

However, Jones ended by proclaiming what can only be regarded as inconsequential, that January 2010 was the warmest on record, the record having begun in 1979. I doubt any reputable scientists would regard one month in a 31-year record as proof of anything, up to and including considering it to be just cause for reordering energy production and the economies of all nations.

Although the Iowa Method would approve.

“The risk models employed turned out to be merely formulaic descriptions of the past and created an illusion of precision.”

Tom Moriarty has been criticizing some recent work by RealClimate’s Stefan Rhamstorf for being exactly this kind of modeling. Rhamstorf believes that he can project future sea level based on a complex formula relating temperatures to sea level-this was published in PNAS, but it turns out that the model can describe the past, but as Tom points out, it is unlikely to be an explanatory model, because it produces nonsensical results:

http://climatesanity.wordpress.com/2010/04/12/rahmstorf-2009-off-the-mark-again-part-4-parallel-universes/

In the first post of his series, tell me if this sounds familiar, Tom says:

“It is very important to understand that VR2009’s model (equation 2) is put forth as more than just a description of sea leve [sic] rise for the last 120 years. Rather it is an explanation for that sea level rise…But the difference between a formula that describes and a formula that explains is essential to understand, For example, if you were to drop a rope on the ground in a random fashion you could come up with some kind of formula for the elevation of the rope at each point along the first half of its length. Perhaps you would fit the elevation to an 10th order polynomial that mimics the pattern that the first half of the rope made. But that formula would have no power to tell you the pattern of the second half of the rope. Your formula would be a description of the pattern of the first half of the rope, but not an explanation. Also, if you lifted your rope and dropped it in a new pattern, your formula would now be useless to predict the elevation of the first half of the rope again.”

If you read all of them, it gets really juicy, turns out one can feed all kinds of scenarios for temperature change into Rhamstorf’s model, and the projected sea level can end up making zero sense.

“The excesses in subprime lending in the United States were fed by an excessive amount of faith in technically sophisticated approaches to risk management and a misguided belief that mathematical models could price securitized assets, including securities based on mortgages, accurately. These valuation methodologies were so technical and mathematically sophisticated that their utter complexity lulled many people into a false sense of security.

In the end, the complexity proved hopelessly inadequate as an all-encompassing measure of risk, despite its frequent advertisement as such. The risk models employed turned out to be merely formulaic descriptions of the past and created an illusion of precision. Such approaches could not and cannot replace the forward-looking judgment of a seasoned professional.”