IPCC, models, HadCRUT, and cherrymandering

Dang, I read a silly comment by sharparoo, I set out to articulate why it is, go back and scan SB11, line out my thoughts, start writing a bit, decide to refresh the page only to learn I’m too slow. 🙁
So, Sharp!!! What they said!!!
But, honestly, Sharp, I don’t think Dr. Spencer considered the possibility of the screams of cherry picking to begin with. It wasn’t the purpose of the paper to show all of the possibilities and find one that may have accidentally coincided. As a net admin, I run into a similar problem. I’ve colleagues and friends that are incredibly good hackers. I’ve as much knowledge as they, but my mind doesn’t work in the manner theirs do. So, when I set up a network or a node that needs to be secure, I ask one of my friends to break into it. They let me know how hard it was and whether or not I forgot a certain approach. Dr. Spencer needs to hire an alarmist cynic to read his papers first before submitting. But even after that, it would be impossible to cover every possible potential criticism regardless of how little rationale was behind the criticism. Maybe that’s what he’s using A.D. for.

Greg (September 22, 2011 at 4:01 pm), good one!

James Sexton: rather than wasting money on a warminista, we need to develop a AGW version of the old 8-ball. You know the one that would float an answer to a question in a window? Just take all the objections we’ve run across and put them in there – the AGW crowd is so predictable that we’d probably anticipate the objection at least 60% of the time.

Wouldn’t the title be more apt if it read “Spencer b-slaps pedantic critics!”?

If we consider Dessler’s purported rebuttal of Spencer, and Spencer’s rebuttal of Dessler in this post, we necessarily conclude that 2 or 3 of the 13 models used by the IPCC in their climate forecasts do a better job of reflecting recent temperatures; these are the ones which better reproduce ENSO. None of the models faithfully reproduces world temperature trends, but at least these 2 or 3 reproduce the timing (though not the magnitude) of the heat loss following a warm maximum.
This relative superiority of 2 or 3 models does not and in principle cannot prove the “truth” of these models, but does demonstrate that they are superior to the 10 or so less competent models which are currently embedded in IPCC predictions. At the least, one would want to know what future climate trends are predicted by these better models, and whether and how they differ from current IPCC predictions. Whatever they predict may not be any more accurate than a ouija board, but at least have slightly improved credibility compared to those from demonstrably worse models.
Theo Goodwin (at 2:44 above) expresses scepticism with assertions that using these 2 or 3 better models would be a “better model of Earth’s climate.” If his point is that even the best of these models has not been demonstrated to be reliable at predicting climate, I completely agree. I wouldn’t bet a dime on the predictions of the best current models, much less the economic future of the developed world. But I do think that eliminating demonstrably worse models from consideration in favor of slightly better models represents a modest scientific advance. Science can never prove the truth of a proposition, but it can prove the falsity of one; and it can discriminate between a better explanation and a worse one.
If nothing else, the Spencer paper and ensuing controversy should lead to the dumping of the worst models currently used by the IPCC.

I wonder if any of the following limitations of the IPCC models provided in Chapter 8 of the 2007 Report Working Group I “The Physical Science Basis” have anything to do with the problems identified by Spencer?
“Due to the computational cost associated with the requirement of a well-resolved stratosphere, the models employed for the current assessment do not generally include the QBO.”
“In short, most AOGCMs do not simulate the spatial or intra-seasonal variation of monsoon precipitation accurately.”
“The spatial resolution of the coupled ocean-atmosphere models used in the IPCC assessment is generally not high enough to resolve tropical cyclones, and especially to simulate their intensity.”
“Blocking events are an important class of sectoral weather regimes (see Chapter 3), associated with local reversals of the mid-latitude westerlies. There is also evidence of connections between North and South Pacific blocking and ENSO variability…but these connections have not been systematically explored in AOGCMs.”
“Unfortunately, the total surface heat and water fluxes (see Supplementary Material, Figure S8.14) are not well observed.”
“These errors in oceanic heat uptake will also have a large impact on the reliability of the sea level rise projections.”
“Due to the limited resolutions of the models, many of these processes are not resolved adequately by the model grid and must therefore be parametrized. The differences between parameterizations are an important reason why climate model results differ.”

DM says:
September 22, 2011 at 6:40 pm
“Climate models have to be based on some physical theories and hypotheses, otherwise you couldn’t put together the code for them to run.”
OK, then, do you have some formulation of the physical theory, the set of hypotheses, that was used to create the code so that the model would run? I have been asking every Warmista for that set of hypotheses. Where is it? Is it in the deep oceans with Trenberth’s missing heat?
Computer code has no cognitive content of its own. All it can do is spew out simulations which are just series of numbers. Someone then has to take those numbers and say “Oh, look, this segment of numbers tracks the observational data for La Nina.” The computer code, having no cognitive content of its own, cannot in any way be about La Nina, represent La Nina, or even sing La Nina.
I began managing models for large corporations about forty years ago. I can assure you that no computer model, however complex, amounts to a theory. After all, there are beautiful computer models of the stock market. And they are incredibly useful analytic tools. But surely you don’t think for one minute that someone actually has a theory of the stock market that consists of reasonably well-confirmed hypotheses that enable accurate prediction of stock market behavior.
What the Warmista are doing with their models is essentially the same thing that someone does when he takes a set of existing graphs for past years of something like Arctic sea ice extent and from them projects the extent for this coming Halloween. They gussy it up with all kinds of interesting differential equations and similar items and convince themselves that it amounts to a theory. It doesn’t. If you will read what I have written above, in earlier posts, you will see that a model is a pale shadow of a theory.

Barry Elledge says:
September 22, 2011 at 6:18 pm
“This relative superiority of 2 or 3 models does not and in principle cannot prove the “truth” of these models, but does demonstrate that they are superior to the 10 or so less competent models which are currently embedded in IPCC predictions. At the least, one would want to know what future climate trends are predicted by these better models, and whether and how they differ from current IPCC predictions.”
Nope, sorry, but all the models show major disagreement with some set of observational data or another. That means that all of them are toast. The fact that some of them agree well with ENSO is of interest to modelers as they revise their models. But it should be of no interest to scientists. Unlike sets of physical hypotheses, aka physical theories, the individual components of a computer model have no cognitive content of their own. You cannot isolate the computer code that generates the segment of numbers that are treated as tracking ENSO and do further investigation on that code alone. Nor can you port that code to another model and treat it as the same code. Each model has to stand or fall as a whole because it is only as a whole that a model generates a simulation and has empirical import.
Please do not use the scientific word ‘predicted’ in reference to models. Instead, say that the model user “extrapolated” a result from his simulation, as one extrapolates about future sea ice from old graphs of sea ice. To do prediction, as physical hypotheses do, your tool must specify all the observable facts. Physical theory does this by implying the facts. Computer code has no way of doing this. The numbers that make up a simulation can be treated as tracking observable facts but they have no logical or methodological connection to the observable facts. The connection is totally in the mind of the model user, unencumbered by scientific method.