RE: DocMartyn

"If a model is not a hypothesis, in the classical sense of being falsifiable, then what exactly is it? How can a model be deemed useful if it can tested against reality and found wanting?"

I think you made a very good point with your post and certainly the idea of a model being a hypothesis and therefore falsifiable is a good one. However, I think we have to go a little further in that a model can contain two parts to it:

(a) The model contains an algorithm based on some physical laws or principles by which output from inputs are calculated
(b) The model must be parameterised at run time and the results may be sensitive to the choice of parameters

It is possible for the model physics to be correct but still end up with a mismatch with real world data when the model is run because the input parameters, start conditions or constants have been incorrectly set. However, this then leads to the problem with models: a model can be "tweaked" in order to obtain a match with known data by modifying its parameters. However, this cannot be then taken as a demonstration of the correctness of physical processes built in to the model. This is the classical logical fallacy/tautology about "proving" things with models.

Where the physics is known from first principles then the "tweaking" should not required. This can be true of many models, eg laws of motion (at low speeds ie not where relativity becomes important).

Regarding the term "lurker" I'd suggest there's a possible improvement to "reader", many people read a blog without actively taking part. eg I read this blog but I'm not an active participant in most discussions.

However "lurker" is a very well understood term, it does the job and it's not worth quibbling over.

Omnologos: I preface this statement with a big smiley face to indicate that it is not meant completely seriously

:o)

To borrow your formula "Unless the BH commenters grow up from this _demonising_ attitude, where every Met Office bod is seen not as an individual but as part of some Evil group, I'm afraid much of the dialogue will be meaningless."

Back to serious. Such inaccurate generalisations could go round in circles for a long time, so it's worth noting that we posted the emails here because we think this is a place where a dialogue is possible.

DocMartyn: You are kind of making the same point I was and I like your analogy of the bicycle tyres, which makes the point more vividly than I did. Ed showed that the global average sits within the model envelope, but when I looked at the data in a different way (using an older set of model runs which complicates things) I found that regional disparity. The point of bringing it up at all was twofold. First, getting the global average right is a necessary but not sufficient criteria. Second, we have far better observations for the 60S-60N region that we do in the Arctic.

Noticing that disparity is just the first step down a much longer path. It was an analysis I did in an afternoon, just to see what I could see. That's why you are seeing it discussed in an email rather than in a more formal paper. What my little observation means in terms of REJECTING something, or otherwise, is a lot harder to say. If we do consider a model as a complex hypothesis (putting my Popper hat on) then

-We need to consider that we have not one, but several models in that envelope. Ideally we would have thousands of runs of each, but we don't. At best I had access to a handful of runs from each model.

-We need to consider the observations lying outside the envelope as a "falsifying hypothesis", which can itself be falsified, for example if a systematic error is discovered in the observations, or if there was an important difference between what was modelled and what was measured.

-We need to consider which if any of the sub-hypotheses contributing to the complex hypothesis are falsified. To give a specific example, the forcings used in the model runs for the past decade could be wrongly specified, or one subunit of the model could be incorrect while the model was otherwise perfect (I consider that latter option unlikely by the way).

-We need to consider the fact that this is a probabilistic question and that this could simply be an unusual event with a small but finite probability (for example, if I roll three 6s with three fair dice I don't reject the hypothesis that they are fair dice simply because the probability of getting three 6s is less than 0.01). This question is clouded further by the difficulty that this particular run of years is being examined precisely because it does seem unusual.

Taking the Popper hat off (they're somewhat unfashionable nowadays I'm told) there are a whole other set of arguments about how models can be useful without being perfect (Tamsin Edwards blog is particularly eloquent on this subject http://allmodelsarewrong.com/), what the differences between projections and predictions are. How might one go about estimating uncertainties based on imperfect models? We also have to ask whether this was a question that the model experiments were even designed to answer? How do we effectively balance the evidentiary value of 5 years of 'fresh' data against 155 years of data that everyone is familiar with?

I'm not offering any of these up as a defence of models per se. I'm just listing the things I would find it apposite to consider before taking the analysis further. The key thing as far as I understand it (and I would be the first to admit that I'm no expert when it comes to models) is learning how these differences between observations, model output and other hypotheses can help improve our understanding of the real world.

John, thanks for the reply. I am an experimentalist where possible, a modeller when needed and a statistician only in the case of desperation.
However, as a part-time optical spectroscopic I cannot understand why you do not use difference spectra when doing model vs reality.
Any examination of a sequential time series of the difference (model-reality) allows you to observe small components against a background. Difference spectra allow you to spot information against noise, and should make more statistical analysis rather easier.
With regard to Popper, he basically said it all.
The history of science, especially my field, is littered with examples rather nice sounding hypotheses delivering real human tragedy; eugenics, vestigial organs, the Dalkon Shield, Opren (Oraflex in the US) ...........
Bitter experience has taught the biomedical's that the sequence of models;
tissue culture, mouse, dog, young healthy male back-packer, initial patient population and then general population still fails with regularity.

Bizarre question though it may seem, Richard, do you recall when you first heard, and first used, the word "redacted"?

Bizarre question though it may seem, Richard, do you recall when you first heard, and first used, the word "redacted"?

I'm not Richard, but wasn't it when the dodgy dossier came out?

Splitpin,
See my post on the word "redact/redacted" at Climate Audit after the Norfolk Constabulary announced the closing of the case. The word was used by "FOIA" in their editing of the Climate-gate II emails.
http://climateaudit.org/2012/07/19/scotland-yard-interviewed/#comment-343752

It proves nothing, of course, but I thought that it was worth mentioning. I think few people using the British-English (!) dialect would have casually chosen to use the word at that time.

Richard Betts

Am I missing something here or did nobody ask you informally for this information?

I have a couple of questions for Dr's Betts and Kennedy.

1) Why do you think that the measurement of temperature between 1980 and 2008 has changed by 0.0037 degrees per year, when one compares today's GISS data with a GISS dataset that was downloaded in spring 2009?

http://i179.photobucket.com/albums/w318/DocMartyn/HowGISShaschangedin3years.jpg

2) How can you model a moving target? That is, if you tune your model to the data from, say 1950-2005, when you know that in a few years time the model will be wrong because the past temperature has changed?

I was mistaken about GISS due to a website navigation error. The download I did today was from "Our traditional analysis using only meteorological station data" and is running 30% higher than the GISS Global temperature I had previously used.
They did not outrageously cheat in the manner I though, I was wrong, they cheated by a smaller amount. 0.005 degrees per century, but every little helps.

http://i179.photobucket.com/albums/w318/DocMartyn/ProperGISSJuly2012andSpring2009.jpg

Excuse me for asking, but I seem to have lost my bearings on this so I'll put my understanding down on paper here:

Did not Rob Wilson do what was supposed to be a series of blogs at BH?

Did he not claim in the second blog that he'd had "nasty" emails from posters?

Did not posters, including myself, ask him to name and shame the posters who'd insulted him?

Did not Martin A, presumably having seen climate scientists in the past claim they'd had death threats only to find they'd received emails saying things like, "You're not very nice", put an FOI in to see the emails which Rob refused to show?

So my question is why are we seeing an email exchange between Richard Betts, Ed Hawkins, John Kennedy and Rob Wilson. I know for a fact that Richard and Ed are so nice that they have been prescribed "nasty" pills to be taken three times a day to balance their personalities, and from the emails it looks like John Kennedy is in the same support group. Rob, I'm sorry, but you come across as thinking you're yummy.

As much as I like to see these four fine chaps communicaing behind the scenes, it's not what piqued my, and others' interest, it was the "insulting" posts, and we're being denied them. Why? If Rob has been villified, threatened or insulted by any of our regular posters let us all see the evidence then we can draw our own conclusions.

Martin, thanks for the explanation, I'd missed the bit about FOI ing the emails of Richard, Ed, John and Rob (doesn't really have the same ring as John, Paul, George and Ringo does it). They're fine fellows to a man, but all believe that humans are warming the earth and that it will be dangerous.

It's not a knew phenomenon, except for the warming bit, there have always been humans who believe the human race is intrinsically evil in some form or another and its actions have to be controlled by a group of people who have access to the "secrets".

I am continually amazed that clearly likable individuals don't see the horror of what they're doing in terms of the human suffering caused by high energy prices. Nor do I follow the rational that we should let people starve or die of hypothermia now to save the planet for future generations.

I'm not saying that's the view held by the Fab Climate Scientist Four above, just that they seem unaware that their fruitless propositions that we can reduce CO2 output will cause untold misery. You would think wouldn't you that someone in the scientific hierarchy would by now have pointed out that we are emitting the same amount of CO2 in year that China does in three weeks so that our government can make do something to reverse the insane energy policies foisted on us by WWF activists.