Goofy Theory of the Day

From NewKerala.com, via the Thin Green Line:

According to Prof McGuire, in Taiwan the lower air pressure created by typhoons was enough to “unload” the crust by a small amount and trigger earthquakes, reports the Scotsman.

Uh, right.  We don’t know what triggers earthquakes in general, so we certainly don’t know the affect of atmospheric conditions on earthquakes.  This is outrageous speculation from an all night session at the pub, breathlessly reported as actual news.

Let’s do a thought experiment.  A strong typhoon might drop local atmospheric pressure by 0.2atm.  The pressure at the bottom of the ocean averages 200-600atm, and under a few miles of rock is even higher.  I would challenge someone with measurement instruments on a fault to even detect such an atmospheric change.  Even on surface faults, we are talking about gigatons of force held in check by friction — this is roughly the equivalent of a feather landing on the Empire State Building and collapsing it.

I sometimes wonder if we will see a future SAT question whose answer is “climate studies are to science as alchemy is to chemistry”.

  • NEILC

    The global population is supposed to tail off not double in the next 35yrs, since as the developing world develops, and women become enfranchised etc and begin to hold jobs the average number of children born will decrease, especially in places like India. China has already set laws around single child birth. So the population will not simply double in the next thirty-five years as has been the rule in the past.

    “As for when to take action against AGW: I would recommend taking action when it is known, and I do mean KNOWN, not guessed at, that the bad will out weigh the good with global climate change. Until that is certain, there is no logical reason to harm ourselves now.”

    THIS is not logic. Global warming is not a guess, it is a best estimate using a wealth of available evidence, and training of thousands – there is a large difference between the two. YOU are guessing that it is false, not based on the available evidence but consistently choosing to err on the side of skepticism despite evidence telling us not too. Therefore the logical action, in view of the best estimate of the scientific community would be to act now when it would be (relatively) cheap to avert the effects, rather than act later, when we would be positive but when it would be too late and the relative cost much higher. You keep missing this point.
    Nobody in Britain could prove that Nazi Germany was going to go to war in Europe but many called for action to remobilise and re-arm the country, since it was obvious to many that he did not have peaceful intentions. As it was, Neville Chamberlain was reluctant to do so, and as a result it took us several years to get war-ready, by which time Europe had been invaded. In hindsight, it would have been better to get war-ready before war started, since everybody’s best guess envisioned Hitler with militaristic aims and despite there being hard proof.

    If we are to play the waiting game as you would do, at what stage of climate change would you then take action?

  • NEILC

    ‘^there being no hard proof’

  • Shills

    Someone mentioned war in relation to climate. There is a book ‘Climate Wars’ which looks at these possibilities, written by some security expert. Not read it. Don’t know if it is reasonable. A bit of a cash-in somewhat, for sure.

  • phat shantz

    As a chemist I was taught that alchemy — however laughable today — is one of our roots to modern inorganic chemistry.

    “Climate Studies,” on the other hand, are to science as Jell-O is to thermonuclear physics. Except that I can demonstrate the existence of Jell-O and show that it remains coherent under infinitesimal examination.

  • NEILC

    Unfortunately, despite your chemical knowledge, your post is more polarised than a water molecule.

  • ChrisP

    ADiff and Wally,

    Just wondering, for the sake of fairness, can you provide some reading for us so that we can understand your point of view.

    Cheers

  • ADiff

    I’ve already provided extensive references in previous posts. Please refer to these for starters.

    It’s become pretty clear that AGW has been over-stated and exaggerated to a great extent, and that the trend is neither unprecedented, nor in any way likely to produce catastrophic or disastrous impacts. AGW is real, but not dramatic or unprecedented and there’s no evidence of any likelihood it’s “dangerous” on any broad scale and won’t produce any changes normal adaptive processes can’t handle (with the exception of some special cases involving wildlife populations with specific limitations imposed by conditions on their mobility and adaptability where geographically very isolated populations are at risk…such cases are common for many other reasons than current climate trends, too).

  • ADiff

    For the DDT issue (for which alone I’ve provided no refs yet) start with http://junkscience.com/malaria_clock.html … It’s polemical as hell…but hey!…it’s easy to understand outrage at something arguably contributing to over 100 million otherwise avoidable human deaths! No doubt one can actually do one’s own research starting from there, I would hope.

    As for the rest, start with ‘Climate of Extremes’ and review some of the other sources I listed in a prior post in this thread…. The evidence AGW is NOT DANGEROUS & urgent action to address it is just not needed at all grows day by day…..

    There are tons of sources and references, articles, books, and lots of web sites with large numbers of these.

  • Wally

    Shills,

    “The IPCC is my evidence. What do you have?”

    The IPCC is organization, not evidence. And if you actually looked at those reports did you notice the heavily over lapping error bars in their models (http://www.ipcc.ch/publications_and_data/ar4/wg1/en/figure-spm-5.html)? Those are just +/- 1 SD, not even 95% CI. So despite different inputs on assumed CO2 emission increases, all those projections are not significantly different. Thus, you don’t actually know that anything we try to do will help (say comparing B1 to A2), even if we assume these models are correct.

    “Yes you can. Happens all the time. War forces innovations, space race forced innovation, free-market forces innovation… ”

    Hardly. Could WWII have forced the innovation of the atomic bomb without the unforced discovery of E=mc^2? What forced the innovation of aircraft? Throwing money at something can only help so much. You can’t just pay someone to come up with things like E=mc^2. You can pay people to make more incremental improvements, such as many of things seen in the engineering field (slightly better processors, smaller hard drives, etc), but the big discovers come when they come. X amount of dollars isn’t going to find a cure for cancer, nor will Y amount give us the next generation of solar power.

  • Wally

    Neil,

    “The global population is supposed to tail off not double in the next 35yrs”

    Isn’t that what they’ve been saying for almost 100 years now? Those kind of projections are almost always wrong. And much like AGW models, these models have a ranges between growth at logarithmic rates, to actual decreases. To push only the peak then decline model clearly illustrates your bias towards particular conclusions as you ignore evidence to the contrary likely present in the same reports you’re getting your desired information from…

    “and women become enfranchised etc and begin to hold jobs the average number of children born will decrease, especially in places like India. ”

    This is to a large extent already happened, and I was using 2% growth which is similar to most of the western world.

    “Global warming is not a guess, it is a best estimate using a wealth of available evidence, and training of thousands”

    Neil, are you a modeler by chance? I am, and in that business best estimate = guess. They can’t experiment. Thus, they can’t actually PROVE the value of most of their parameters. This leads them to guess. They even have to guess at the which interactions are present, not only the parameters. This could lead to not only incorrect parameters, but missing interactions in individual equations, to missing equations entirely. Its a guess, pure and simple.

    “YOU are guessing that it is false, not based on the available evidence but consistently choosing to err on the side of skepticism despite evidence telling us not too.”

    But you can’t actually point me towards ANY evidence that we could significantly alter the result even if we take the IPCC models at their word. But of course we can dismantle those models six ways from Sunday.

    “Therefore the logical action, in view of the best estimate of the scientific community would be to act now when it would be (relatively) cheap to avert the effects, rather than act later, when we would be positive but when it would be too late and the relative cost much higher. You keep missing this point.”

    No. Your point is to assume that these things are actually “best estimates” and not blind guesses by people creating models to spit out a desired result. Plus you regularly ignore evidence against your point, such as overlapping +/- 1 SD error bars on our climate models. The worst of those models gives a 4 degree window at just +/- 1 SD. And the case of reducing our emissions after some peak in the relatively near future still has an over lapping error at just 1 standard dev. Its ridiculous to base billions to trillions of dollars in possible economic growth on these half cooked mathematical models. So your “logic” relies on a false premise, sir.

    “If we are to play the waiting game as you would do, at what stage of climate change would you then take action?”

    I’ve already answered that question.

  • Wally

    Neil,

    “Just wondering, for the sake of fairness, can you provide some reading for us so that we can understand your point of view.”

    You’ll have to be more specific. Just what parts of our conversation do feel the need to read up on.

  • Wally

    Wait a second, those gray bars are actually -40% to +60%, yet they claim them to be the ‘likely’ range that Neil told us meant 90% CI? This whole report is such crap. Exactly who puts any kind of error bar on a graph that is just -40% and +60% of the mean? Take a look: http://www.ipcc.ch/publications_and_data/ar4/wg1/en/figure-10-29.html

    These are their models individual models they pull this AOGCM composite “model.” They are all over the place. And they just simply take a mean of them? These are “best estimates?” Good lord. So, what they have done is get the mean of a set of models, and made a SD of those model’s results. But what of each model’s uncertainty as well? What of selection bias towards of which models they choose to include in this “meta-analysis?” What are the assumptions of all these models, and why should we treat them as one?

    I knew these things where crap, but man it only gets worse with closer inspection.

  • Shills

    Wally says: ‘The IPCC is organization, not evidence’

    Well of course I mean the referenced reports don’t I?

    And I should prob. clarify that it is not ‘my evidence’ because I don’t have a an understanding of all the evidence, but rather I am interested in this argument with Neil and Chris who understand it, and I’m pretty sure would say the IPCC is the evidence (apart from the stuff post-AR4). Sorry for any confusion.

    So Re. you graph. I would have thought the SD would give you some idea of the confidence intervals? But I really dunno. Why don’t you ask a climate scientist or something?

    Wally says: ‘X amount of dollars isn’t going to find a cure for cancer, nor will Y amount give us the next generation of solar power.’

    So you think we should just leave it to chance/fate? No one is saying that there is any guaranteed recipe for innovation. But money, policy, any soughta motivation, will get peeps focusing on the issue. The more heads on an issue the better. There is no doubt that certain environments are more conducive to innovation than others.

  • Wally

    Shills,

    “I would have thought the SD would give you some idea of the confidence intervals? But I really dunno. Why don’t you ask a climate scientist or something?”

    Ask a climate scientist about a statistics question? I’d rather not. And while the SD is a judge of the variation pressent in the data, it isn’t a good one. The SEM or a CI is a much better judge. When you are just looking at a bar chart (for example) showing SD, those error bars tell you pretty much nothing. If the error bars do or do not over lap, that doesn’t necissarly mean the two samples are significantly different. With CI you can know that.

    “No one is saying that there is any guaranteed recipe for innovation. But money, policy, any soughta motivation, will get peeps focusing on the issue. The more heads on an issue the better. There is no doubt that certain environments are more conducive to innovation than others.”

    But what’s the opportunity cost? How much are you willing to spend knowing you may not find anything, or that you may have found it without spending as much or even nothing? This is why it doesn’t need any more emphasis than anything else. Is it not possible that someone working on a jet engine could make the next big discovery in energy generation, for example? What if you pull money away from aeronautical engineering and kill that possibility.

    Anyway, that’s the problem. We need to fund basic research evenly, because you don’t know what will truly be useful. That’s why its basic research. And if history is any guide it should prove to us that we can’t decide when and where the really game changing discovers and ideas come from.

  • NEILC

    Wally, if you had in fact taken a closer look like you said you had you would realised the -40-60% mean was averaging out several models, this large range is expected when you average out several models, and to be honest i would not use it as evidence myself.
    However, upon real closer inspection you would see individual models plotted against these with means and 5 to 95% ranges of their outputs. Take another look – the AOGCMS are the red dots. Im betting you haven’t read much of the IPCC report very carefully if at all.

    Now it is not particularly useful saying all the scenarios overlap in error bars and so there are no statistically valuable trends. This is not a continuous variable we are measuring when we test different emissions scenarios. Each scenario should be taken separately with their own error bars. There is one model only on that chart who’s minimum range lies within 1 degree warming.

  • Wally

    Neil,

    “if you had in fact taken a closer look like you said you had you would realised the -40-60% mean was averaging out several models, this large range is expected when you average out several models”

    That’s what I said no? That it was -40% to +60% of this multi-model mean? And while we do expect a large range, who exactly said -40% to +60% of the mean means anything at all? What statistical method of error approximation is that? Its complete crap Neil. They pulled that out of their ass. Probably because they don’t know how to create a measure of error when doing a meta-analysis on various models with their own error. Or maybe they do, and they didn’t like what they saw… I don’t know…

    I gathered that about the AOGCMs, and I’ve read parts of the report more closely than others. None of that really matters though does it? We’re discussing this part now. What you’d bet is completely irrelevent. And seeing as you thought “likely” meant 90% CI, when it was actually +/- 1 SD, means we could probably say the same thing about you.

    “Now it is not particularly useful saying all the scenarios overlap in error bars and so there are no statistically valuable trends.”

    I question your scientific training if you actually believe this. If you create a model, input different assumptions, and can’t get significantly different results from those different assumptions, you can’t tell me to 95% degree of certainty that those changing assumptions matter. Now you can interpret that a number of different ways, one way is that it shows you that your models suck.

    “This is not a continuous variable we are measuring when we test different emissions scenarios. ”

    It is actually a continuous variable, you’ve just made several discreate assumptions. You could have just as easily made plot with a continuously changing set of assumptions and responce from them.

    “Each scenario should be taken separately with their own error bars. ”

    Depending on what point you are trying to make at the time. But if you’re attempting to tell me that we need to reduce our emissions, then you have to point to the set of graphs and a comparison between them, not just each one individually. And of course the set show no statistically significant different result from various different assumptions on future emissions, thus you can’t actually support your argument, even if we accept these models.

  • NEILC

    If you look at the reports introduction it does tell you what they mean when they say likely, very likely etc, and it does specifically tell you that ‘very likely’ is greater than 90% confidence.

    I dont see anywhere in your figure reference that uses the term likely or very likely so i am not sure how you think i have tried to mislead you here. And it is very important that the ranges of the individual models mostly fall above the 1 degree warming part, you cannot just ignore that.

    You are misinterpreting me on the overlapping error aspect – of course, if you were measuring a continuous variable, say a concentration of a chemical, then any trends with overlapping error bars would be statistically insignificant, but in this case, where emissions scenarios represent discontinuous variables, it does not quite work the same way: We are not attempting to find any relationship or trend between emissions scenarios and temperature, merely seeing how different scenarios pan out. Does that clarify what i am trying to convey?

  • Wally

    Neil,

    In table SPM.3 they claim the “likely” range is 1.1-2.9 for scenario B1, for example. This is nothing more than .6*mean to 1.6*mean and rounded. I’m not sure if you attempting to mislead me or not, but I find it rather lazy, disingenuous, and suspect that this IPCC report calls this a “likely” window. This is not a statistical measure of the variance in any way what so ever, nor a measure of the confidence in the mean. Its somebody’s stupid idea, nothing more. Later the colored shadings in figure SPM.5 are +/- 1 standard deviation of the set of models’ means (which is meaningless, but we’ll get to that in a second), and gray shadings are this -40%, +60% thing, which of course is prominently displayed in that figure.

    “And it is very important that the ranges of the individual models mostly fall above the 1 degree warming part, you cannot just ignore that.”

    Which ranges? The +/- standard deviation of the means from the different models? Who cares? Firstly, +/- 1 standard deviation only shows where ~68% of your data falls relative to the mean, assuming we have a sample of normally distributed data and all that. It doesn’t give you a measure of confidence in the actual location of the mean. In short, people really shouldn’t be using SD as much as they do, it doesn’t tell you much. We need to be using SEM or CIs. Secondly, the SD of the mean of a handful of models is a meaningless error measurement. Those models come with their own errors. You need a much different analysis to give an accurate measure of the error when doing this kind of meta-analysis.

    These are the kinds of things that are taught in a 200-300 level stats course, maybe even 100 level. I understand that the politicians that are supposed to read these things are not expected to understand such things, but that’s exactly the point. They create these error bars largely knowing the audience they are writing them for won’t know the difference. Thus, the authors should be held to a very high standard, by themselves, as well as you and me. Its too easy to confuse politicians and laymen by throwing up some gray lines and calling it a “likely” window, when how they actually defined that is a meaningless or irrelevant statistic. Its disingenuous and just further proves these people were not trying to be completely honest, but trying to create some ammo for their pet cause. Well, either that or they are completely incompetent.

    “We are not attempting to find any relationship or trend between emissions scenarios and temperature, merely seeing how different scenarios pan out. Does that clarify what i am trying to convey?”

    I don’t think you yourself are completely clear on what you are saying. The vary nature of coming up with a set of predictions based on different assumptions is to compare what happens under those different assumptions. But if your point is to want to say a set of models predicts some range for X amount of emissions great (assuming for a second their error actually means something), that is pretty trivial, right? The idea now is to be able to do something about it, right? So if we do nothing, and this is largely hypothetical to illustrate a point, and a model predicts 4 degrees of warming +/- 2 degrees at 95% confidence. Then the models using emissions predictions under say Cap and trade, or world wide compliance with the Kyoto protocol (or something similar), and predicts 2 degrees of warming +/- 1.5 at a 95% confidence interval. So then the idea is to see how much difference these changes will make in the world temp. And as we can see the range at 95% CI in one case is 2-6 degrees, and the other is .5-3.5. So, and I’ll guess a little again for illustrative purposes, say there is 40% chance doing nothing will be 2-3.5 degrees, and then doing something will be a 50% chance at 2-3.5 degrees. So these models predict a 20% chance that doing nothing will lead to the same window of result as doing something. Now this isn’t the perfect way to do this, but its the basic idea, and why those error bars need to be correct. This is because doing something is going to likely cost us trillions of dollars. And our models can’t even tell us that it will matter. They can only tell us its might matter some, it might matter a lot, or it might not matter at all. There is just way to much error currently present to justify these kinds of costs.

    Now I’m not sure what exactly you’re arguing, but this is the correct way to think about these models’ results and how they should inform our choices. We need to understand what kind of cost is likely to make what kind of difference, and how much that difference will actually matter. I think we can both agree that spending trillions to save tenths of a degree over a century is probably not worth it, but if it is going to save 5 degrees, it maybe is. Its a strict probabilistic cost benefit analysis. And I haven’t seen anyone pin down the likely costs and benefits enough to support any kind of climate legislation.

    And finally, those models are likely to be underestimating the error for the simple fact that they don’t know what they don’t know. So they are only able to estimate the error in the processes they attempt to account for. And again the lack of experimentation leads to a very small confidence in at least establishing that what they don’t know is fairly small. Much less to actually test the validity of their model in the first place.

  • NEILC

    Wally,
    the individual models were plotted with “a mean and 5 to 95% range (red line and circle) from a fitted normal distribution”. So they are not sigma-1 error bars. I am merely trying to correct what you seem to not be reading about this figure.

    And you are also misinterpreting again concerning overlap of error (regardless of the errors), you cannot find any trend (as in a gradient) using discontinuous variables. The figure is not plotting CO2 against temperature but temp with certain CO2 concentrations.

  • Wally

    Neil,

    I know that individual models had that, the AOGCMs did not.

    “And you are also misinterpreting again concerning overlap of error (regardless of the errors), you cannot find any trend (as in a gradient) using discontinuous variables.”

    I’m sorry, but that is complete BS. It is done all the time in every field you can think of.

  • Neil,

    I know that individual models had that, the AOGCMs did not.

    “And you are also misinterpreting again concerning overlap of error (regardless of the errors), you cannot find any trend (as in a gradient) using discontinuous variables.”

    I’m sorry, but that is complete BS. It is done all the time in every field you can think of.