Ian Musgrave has written the post I was going to write on Jon Jenkins' article in the Australian, so I just want to emphasize that fitting a degree six (yes, degree six) to temperature data does not produce a meaningful trend line in any way shape or form. Go read.
Note that if the editors at the Australian had bothered to read their own paper just three days earlier they would have known that the Jenkins' claims about the Oregon petition and global cooling were rubbish.
News Limited blogger Grame Redfearn also pointed out the enormous holes in Jenkins' arguments and talked to Australia's acting chief climatologist Michael Coughlan:
Of climate change contrarians such as Jon Jenkins, Coughlan has this to say.
"We have produced rebuttals of all of these arguments - they have all been addressed. But they just keep trotting them out. No matter how many times you tell them they're wrong, they just keep going. The general approach seems to be - if we keep banging away at an untruth, people will start to believe it".
Let's not forget that these contrarian views are not being expressed on a bit of street press or some fringe web site somewhere - they're being repeated over and over in Australia's only national newspaper. So now comes the revelation - and that is Coughlan's view of The Australian newspaper itself.
"The Australian clearly has an editorial policy. No matter how many times the scientific community refutes these arguments, they persist in putting them out - to the point where we believe there's little to be gained in the use of our time in responding."
Since I already uploaded them, I might as well include my versions of the graphs that Musgrave posted.
Here's the graph that the Australian printed.
It's been copied from Lorne's Gunters' article in the National Post (Canada's version of the Australian), but with this caveat removed:
Moreover, while the chart below was not produced by Douglass and Christy, it was produced using their data
and the ridiculous sixth-degree fit attributed to the University of Alabama at Huntsville.
That graph ends in July was already out of date when Gunter's article was published in late October. Here it is with the latest data (up to November) included.
Note that the latest observation lies right on the linear trend line, completely refuting Jenkins claim that satellite data showed that warming "had completely reversed by 2008".
(Hat tip to Tobias Ziegler for pointing me to the Readfern post.)
Update: Ian Musgrave has more on what is wrong with high order polynomial fits.
- Log in to post comments
So when do temperatures in the lower troposphere hit absolute zero?
You would assume that an "expert on computer modelling" could not possibly have missed the absurdity of using that polynomial fit as a trendline. But have you seen his website?
"We have produced rebuttals of all of these arguments - they have all been addressed. But they just keep trotting them out. No matter how many times you tell them they're wrong, they just keep going. The general approach seems to be - if we keep banging away at an untruth, people will start to believe it".
Yes.
That's the strategy, right there.
Best,
D
As I lay on the beach in Roatan it is hard to believe in global cooling. I am pink all over and loving it. The strange thing is that, while living at 49 degrees north, for many years I noticed that the sun was much stronger than now. In the 90s the sun felt like it was irradiating my skin. sort of felt like I was being deep fried. That is not the case now. Something that was there then, is not there now.
Everwhere I look they say the cold is setting records, but of course that is not a trend ackording to the warmists. The truth is a trend can last a pico-second or a million years. The current trend is not up but down.
This trend and the current climate is what the media is picking up on. It is hard to say it is warm when the pee freezes before it hits the ground. You guys are going to have to wait a few more decades before the data starts to support your beliefs. A broken clock is right at least once a day and in most cases twice a day but only very briefly.
Six more weeks of our Central American trip to go, I just hope the snow is gone by the time we fly home, but I doubt it, It will still be winter.
No disagreement with Tim's article here. kent must think it's correct.
May we float you into space to measure solar irradiance?
As a complete statistical layman, and at the risk of sounding like an idiot, I have a question about the polynomial fit.
Between around 1983 and just before 1985, the difference between the highest temp and the lowest temp is about 0.6 degrees C. Over the same time scale, between 2006 and 2008, the difference is 0.8 of a degree. So why is the huge steepness of the gradient of the latter compared to the former so noticeably out of proportion with the relative difference in temperature?
Cheers.
Erm... that should probably have read "statistics layman". I may be a statistical layman as well, but that's another story...
The primary source for the temperature graph is Roy Spencer's blog, but the good doctor restricts himself to a fourth order polynomial fit because it "smooths out the large amount of monthly variability in the data and helps reveal the underlying 'trends'. How helpful...
Ah, yes... Roy Spencer. For Roy, graphs are more of an art form than a way of presenting scientific results. It's the medium through which he chooses to express his deep, dark hatred toward the climate science community:
http://www.realclimate.org/index.php/archives/2008/05/how-to-cook-a-gra…
I went over there to leave a comment at Roy's blog, but alas I forgot that most Denialist Misinformation Purveying Distribution Channels as a rule don't allow comments.
Best,
D
Bud wondered:
The simplest way to explain it is that for the middle period, there are years both to the left and right that also have to fit so the change over the middle parts get constrained. However, when you get to the end of the time series you don't have any more points to the right so the ending fit whipsaws where ever it wants. That's why adding just a few more months to the data changed the ending fit so much. This problem is exacerbated with higher order polynomial fits, which is why you don't see serious analysts using sixth-degree polynomial fits for data like these.
I agree with Jennifer Marohasy.... "This data does not suggest dramatic global warming. Neither the warming of the late 20th century, nor the cooling since 1998, is of an unusual rate or magnitude."
How can any sensible person suggest that this graph, however it is assessed, shows runaway global warming as James Hansen and other alarmists insist is occuring?
There seem to be a few versions of the polynomial rollercoaster doing the rounds. I think the one at Icecap is earlier than those already posted.
11
Thanks for that clear explanation, Robert.
Lank, quoting Jen,
"...nor the cooling since 1998, is of an unusual rate or magnitude."
Thankfully, Jen clarifies what she means by "unusual" in comments,
"As regards definitions of dramatic and unusual - I suggest we go with common usage and based on our general understanding of the geological record (say the last 6 million years)."
Well, when I talk about something being unusual or dramatic, I routinely consider the last 6 million years for comparison, don't you?
So we should all just relax. The not-dramatic warming isn't much to get worried about as long as we consider it within a time span that pre-dates the emergence of the genus from which modern humans sprang.
An avalanche of stupid and it's.........January 8.
Thanks Robert. I know enough to know it really doesn't look right, but it's good to know why.
Many, many years ago, while studying the computational aspects of photogrammetry, I was advised of two things: use an odd-degree polynomial for preference when doing a least-squares fit to noisy data; and don't go over about 5th degree without a bloody good reason. Having subsequently done some numerical analysis, I have a better understanding of why these rules of thumb were suggested, and also why you should never extrapolate beyond your data points with a polynomial.
The graph looks like an Epic Fail, but I've never expected better from The Australian.
i am not sure, whether everyone does really understand what the sixth-degree fit will do to ANY data.
http://www.purplemath.com/modules/polyends.htm
the end behaviour of a positive degree polynomial will always be the same on both sides. think about it....
Michael (#15) - For most of the past 10,000 years temperatures have been 1 to 3 degrees Celsius warmer than they are today. The 0.6 degree rise in temperatures during the 20th century occurred from the baseline of the little ice age, which saw the coldest global temperatures during the past 10,000 years. Earth has a "rising fever" only if we pretend the little ice age was "normal" and ignore the Earth's subsequent known temperature trends.
Seems to me that a first-degree polynomial is more than adequate in this case.
Considering the long history of huge temperature variation in the earth's climate (ice ages etc.), the 0.6 of one degree centigrade average rise reported by the U.N. "experts" for the entire 20th century (a rise so small that you would be unlikely to detect such a difference without instruments) shows, if anything, that the 20th century was a time of exceptional temperature stability.
Many earth scientists would argue that the current warming trend began 300 years ago, at the end of the Maunder Minimum, a 70 year period when there were very few sunspots on the face of the Sun. Between 1700 and 1735, according to the world's oldest instrumental temperature dataset, the temperature in central England rose by 2.2 degrees C, equivalent to 6.3 C/century, or about ten times the warming of the 20th century.
All this without the help from those nasty man-made CO2 emissions.
Re #19 (Lank): "For most of the past 10,000 years temperatures have been 1 to 3 degrees Celsius warmer than they are today."
Lank, if you can't provide a legitmate scientific source for that piece of drivel (hint: there isn't one), and taking into account the low quality of the rest of your contributions to this thread, my suggestion to Tim is that it's disemvowellin' time.
Lank doesn't argue against Tim's article so obviously he doesn't think there's anything wrong with it. Instead he just makes assertions that he could easily discover are either wrong or strawmen. e.g.
No earth scientists argue that the CURRENT warming trend began 300 years ago. (Any warming trend, if any, greatly accelerated ~100 years ago.) Not only that but only a minority of results say 300 years ago was cooler than 180 years ago. I can only conclude that Lank has an enormous capacity for self-delusion.
the 0.6 of one degree centigrade average rise reported by the U.N. "experts"
Your use of scare quotes there says a lot about the quality of your tactics.
Many earth scientists would argue that the current warming trend began 300 years ago,
Care to name a couple of dozen reputable ones?
Besides which, even if there was some natural warming, that does not mean any subsequent input from anthropogenic CO2 is not a serious problem.
And maybe 300 years ago was about the time we humans started having a significant impact on global vegetation cover. Anybody else here can comment on that aspect?
And maybe 300 years ago was about the time we humans started having a significant impact on global vegetation cover. Anybody else here can comment on that aspect?
Somewhat indirectly, Mr O'Neill has already shown this to be an irrelevant issue.
Tim, to all appearances this fake curve originated not with Gunter, but with Watts last July (here). Even then, various commenters pointed out that it was completely bogus.
When the Gunter article came out a few months later, Watts posted it and the graph with nary a mention of the prior discussion. Leif Svalgaard immediately pointed out the problem with the curve fit, and at his behest one of the other commenters reconstructed the plot and then extended it 15 years, with an amusing result.
To the surprise of none, neither post ever got a correction. I notice that Watts seldom if ever makes one, doubtless because he knows the majority of his readers don't look far enough down in the comments to see his admissions of error.
I'm no expert, but it looks to me as if any difference between the Watts 5th-order curve and the 6th-order one subsequently used in the NP and Oz can be attributed to the NP graphic artist's loose approximation of the Watts curve. In any case the ultimate source seems clear. Gareth in #8 mentions Roy Spencer himself as a possible source, but that seems less likely considering that the Spencer curve is both more recently dated and is a poor match for the others.
"The Australian's War on Science XXIX"
"The Australian's War on Science 31"
Tim, when you switch to ARABIC numerals the TERRORISTS HAVE ALREADY WON
So, Lank, (January 7, 2009 8:12 PM) the temperature rise in the past century was "so small that you would be unlikely to detect such a difference without instruments".
Good point. Let's not worry about anything you can only measure with those new-fangled instrument thingies. It can't be that important.
I think that in the interests of science you have a responsibility to continue standing outside with your shirt off.
Be sure to give us a yell when you notice something, won't you?
Steve Bloom (#23). It is easy to dismiss my comments with rude remarks like "that piece of drivel" but if you care to study climate science 101 you would have discovered plenty of studies that show the cooling from approximately 11 to 10 thousand years ago (called the "Younger Dryas" event) which ended as suddenly at it began, with temperatures jumping 7° C. Since then, the Earth continued warming up until about 6,000 years ago - the (mid Holocene thermal maximum) when the earth was around 1-3° C warmer than today. Since then, it's basically been cooling off - not counting various smaller variations. There are many studies which show this e.g. Richard B. Alley, The Two-Mile Time Machine: Ice Cores, Abrupt Climate Change, and our Future, Princeton U. Press, Princeton, 2002;
or Joe Buchdahl, Mid-Holocene Thermal Maximum, see his figure 5.18 at http://www.ace.mmu.ac.uk/Resources/gcc/5-3-2-2.html
There are some useful, more detailed accounts of these changes in the Arctic Climate Impact Assessment reports here...
http://amap.no/acia/
"This problem is exacerbated with higher order polynomial fits, which is why you don't see serious analysts using sixth-degree polynomial fits for data like these."
you don't use high order polynomial fits unless you have reason to believe the underlying mechanism runs on a high order polynomial basis. this is what the denialists think "computer modeling of climate" is all about, and why they feel justified to condemn it, when they're not indulging in it. the concept of creating a complex mathematical model reflecting the actual physics of the underlying processes which requires a lot of computer time to calculate is apparently completely beyond them.
Gunter's graph is different from Watt's. It has one more month in it and the degree 6 fit is noticeably different from the degree 5 one -- the downturn at the end is steeper and the "trend" goes all the way down to zero. Gunter may well have been inspired by Watt, but his graph produced by the NP.
I switched to the new-fangled Arabic numbering system because a reader emailed that the triple x was causing a net filter to block the post.
Our idiot communications minister [plans to make such filters compulsory](http://scienceblogs.com/evolvingthoughts/2008/12/the_lies_of_the_intern…).
To add to the comments of Robert and sod about polynomials; changing data points at one end of a polynomial will change the position of the line at the other end of the data set. As shown by a graph linked to by Steve Bloom, extending polynomials beyond the data used to determine them quickly gives ludicrous results. My own view is that they should only be used as a last resort.
Lank (#30). If you are going to quote from sources, you should at least make sure that you are quoting correctly.
"Quantitative estimates of mid-Holocene warmth (COHMAP, 1988) suggest that the Earth was perhaps 1 or 2°C warmer than today. Most of this warmth may primarily represent seasonal (summer) warmth rather than year-round warmth."
This is a bit different from what you have claimed. The explanation for this can be found here:
http://www.ncdc.noaa.gov/paleo/globalwarming/holocene.html
Agreed, Tim. I was aware they're not identical and that the NP produced their own graphic, as I noted in the final paragraph of my comment.
Re #30: Jimmy beat me to it. FYI, Lank, +3C globally would put us back into the Pliocene temp range and result in some serious ice sheet melting. That 7C figure for the YD is also completely wrong. Among other things, bear in mind that there's little or no evidence for a YD effect in the SH.
Re #35: Thanks, Jimmy.
Thanks Jimmy - I overstated somewhat and this was my mistake.
Here is a collation of various plotted temperature interpretations - most show that over the last several thousand years the earth is in a general cooling trend and that more recently (last 300 or so years) this trend has flattened and reversed. http://globalwarmingart.com/wiki/Image:Holocene_Temperature_Variations_…
I repeat my earlier comment that considering the long history of huge temperature variations in the earth's climate (ice ages etc.), the 0.6 of one degree centigrade average rise for the entire 20th century shows, if anything, that the 20th century was a time of exceptional temperature stability. So why not recognise this?
Tim @ 33 - our idiot Communications Minister proves that the terrorists have, indeed, won.
Also, Lank, not to beat a dead horse or anything, but remember that you first said:
"For most of the past 10,000 years temperatures have been 1 to 3 degrees Celsius warmer than they are today."
After I objected, you said:
"Since then, the Earth continued warming up until about 6,000 years ago - the (mid Holocene thermal maximum) when the earth was around 1-3° C warmer than today. Since then, it's basically been cooling off - not counting various smaller variations."
Notice how these two statements contradict each other?
Aren't you just a little embarrassed that nearly everything you've said in this thread has something wrong with it?
Steve Bloom(#37) you say "That 7C figure for the YD is also completely wrong". Please give me your version of the YD.
Re #38: Lank, you're treading on air again in that second paragraph.
Steve - maybe that should be hot air! The Inuit didn't seem to mind when the arctic was 'probably' ice free 6,000 - 7,000 years ago! Seems that it was a great help to their hunting and migration! This recent study http://www.sciencedaily.com/releases/2008/10/081020095850.htm doesn't mention the poor polar bears which seemed to have done okay when arctic ice was much less than today...
"Recent mapping of a number of raised beach ridges on the north coast of Greenland suggests that the ice cover in the Arctic Ocean was greatly reduced some 6000-7000 years ago. The Arctic Ocean may have been periodically ice free." http://www.ngu.no/en-gb/Aktuelt/2008/Less-ice-in-the-Arctic-Ocean-6000-…
I wonder how this could have occurred without man made CO2 to increase temperatures.
> Notice how these two statements contradict each other?
But they both contradict Al-Goracular Global Warmism, so the contradiction's OK! Whee!
#43:
Actually, that article says that the Inuit didn't get there until 4000-4500 years ago, when the ice had returned--precisely because an icy Arctic was "essential" to their hunting.
As for the polar bears, we don't really know how well they did. In fact, given how closely related they are to some brown bear populations, I don't think we even know that there were fully-modern polar bears around in this period--they may have hybridized out of existence, then reconstituted afterwards.
And, needless to say, there are a lot of additional pressures on polar bears now that didn't apply then.
Oh, and in answer to your last question:
"However, the scientists are very careful about drawing parallels with the present-day trend in the Arctic Ocean where the cover of sea ice seems to be decreasing."
"Changes that took place 6000-7000 years ago were controlled by other climatic forces than those which seem to dominate today," Astrid Lyså believes."
(Gack!)
So I read Jenkins' mind-bogglingly embarrassing piece...
And, suspending disbelief for a moment and assuming that he hadn't committed every first-year scientific and statistical blunder in the book of "How Not To Do Science or Statistics", I considered the implication of the plunging neckline at the right-hand side of the trend-line he relies upon in That Graph.
At first jump one might ask Jenkins if he is predicting a global ice-age over the next decade, as the trend is dropping faster than the price of Detroit real-estate in a financial crisis. But of course, being the 'computer modeller' that he is, Jenkins would tell us that one cannot extrapolate beyond the independent axis of a plot (unless of course one 'accidentally' loses some of the terminal data, but one wouldn't do that, because it would be scientific misconduct). I assume that Jenkins would in fact tell us that impending data to the right of the graph might show a completely different trend to the free-fall trajectory that one's trendline is currently (conveniently) describing.
And if Jenkins actually had sufficient wherewithal to comment on extra-abscissa extrapolation, he would no doubt mention that describing the trends beyond the bounds of the x-axis is a fraught exercise when using high-order polynomials, precisely because they do not cope at all with data outside of the range with which they were originally derived. High-order polynomials have almost nothing to do with the average real-world limited dataset, and this is of course the very point that many here are commenting upon...
Oo, hang on - am I not now encountering an internal inconsistency?!
So I can only conclude that Jenkins is either extremely incompetent in the modelling of any data beyond two points describing a line, or incompetent in reviewing his accompanying graphs, or that he is in fact massaging his message.
There you go: I have made an accusation. Come and get me.
Sadly for science, I am reminded of several other punters' recent attempts to use polynomials as trendlines. Unfortunately, as I am still on a dial-up link my search engine is not obliging rapidly enough, so if anyone has been keeping tabs on such for posterity's sake I'd appreciate the links in order to refresh our collective memories.
Even more sadly for science, there are hundreds of Denialists crowing at Jenkins' apparent scientific/statistical skill, and proclaiming the infinity of his profound insight.
What hope for Homo sapiens (?!) when so many of its number are so statistically innumerate?
If the data has a sudden downturn very close to the end then the higher the order of the polynomial, the steeper the fit will be at the downturn. Of course, it's hard enough getting statistical significance out of a first order fit to this data let alone anything of a higher order and the higher the order, the harder it is to get statistical significance.
Steve Bloom:
When someone has the capacity for self-delusion that Lank has, he can't afford to be embarrassed.
So what is wrong with using a six degree polynomial fit here? I'm sure there is some valid mathematical reason, but it is not in the least obvious to someone not in the know.
> So what is wrong with using a six degree polynomial fit here?
Well, the real question is "why use a degree-6 polynomial fit at all"? Degree-6 polynomials don't seem to correspond to any physical or stochastic processes which are relevant here, and they don't even produce results which are easy to explain (like, well, degree-1 linear fits).
Um at #47 I should have said "extrapolate beyond the graphed values on the independent axis of a plot".
My bad.
Valhar2000 see #11 and #27 (second link). Although bi hinted at it, you get stupid behavior at the end points. It is the statistical equivalent of crack the whip.
A simple numerical exercise will show why an arbitrary high degree polynomial fit to a noisy signal is nonsense. Taking a toy of example, let us assume that the fit equation is 14.0 + 0.02 t - 0.000001 t^6, for data ranging from t=0 to t=30 (like a 30 year range starting at 1979). The 0.02 is the warming in degree centigrade per year. What would the temperature be in the year 2079 according to this fit, and according to a linear fit? A linear fit would say 2.0 degree warming in a century, the other would be 16 -1,000,000 = -999,984 degrees centigrade. This is 999,711 degrees below absolute zero.
The only reason the highest degree polynomial has a negative sign is becasue the end points have a falloff. What would the fit be if you choose the data from 1981 to 2006? A positive term for t^6.
kent writes:
kent, I don't know if you're aware of this, but when it's winter in the northern hemisphere, it's summer in the southern hemisphere, and vice versa. Global warming is about the mean global annual surface temperature. Global warming doesn't mean we'll never have cold winters again.
You've been told why this is false before, but you just keep repeating it. That's not a mistake, that's either intellectual dishonesty or a stupidity so deep you shouldn't have the brains to continue to breathe.
Once more, for the record, the trend is up. See:
http://www.geocities.com/bpl1960/Ball.html
http://www.geocities.com/bpl1960/Reber.html
Lank posts:
This if false. Where do you get this crap?
Hmm, if Tim is going to start a blog on me I would think he should invite me to contribute. However his email inquiry about my background prompted me to have a look. All his questions about me are answered in my inaugural speech when I entered the NSW Parliament (http://www.parliament.nsw.gov.au/prod/PARLMENT/hansArt.nsf/V3Key/LC2003…). The Greens have already investigated me with vigour Tim!
Tim and others are correct I have no formal climatology background but I have some considerable experience in computer modelling of integro-differential equations, mainly from protein folding but also from other interests in electronics and high frequency circuit simulation which is why I have so many concerns about predictions made from GCM computer models.
Since resigning from the university I have taken an interest in matters climate including downloading Hansen and Schmidt's GCM code base! In view of the comments here some of you should have a look at the way some of the differential equations are parameterised and the order of the equations used!
Now to the science at hand!
The main criticism seems to be the use of the satellite data and the method of curve fitting used. And surprise surprise I agree with many of the comments! The whole concept of curve fitting can be considered "preposterous". In effect you can use any curve fitting method you like to achieve any effect you desire. But this has never stopped the AGW people from using it! The underlying processes of surface temperature are so complex and affected by so many separate equations that no polynomial could replicate them! But more importantly many of the processes are chaotic in nature and CANNOT (and I mean in the mathematical sense CANNOT) be modelled at all.
The point of the curve was that there is a trend upwards from the 70's and a trend downwards since the late 1990's and that does not change no matter what degree (or type) of polynomial you use until you use the longer term linear fits. However if you add in the most recent temperatures across the Northern Hemisphere even the linear fit starts to fall and makes the recent trend downward even steeper for higher order fits.
However I did not in the article, nor would I now, make any PREDICTIONS about future temperatures based on this or any other degree of polynomial fit to chaotic data and nor should anyone else. But for exactly and precisely the same reasons the predictions of the GCM computer models are equally flawed which is really my key point!
Jon Jenkins
Sorry, the above should read "This IS false," of course.
Valhar2000 writes:
You try to fit the simplest curve you can. If the simple curve fits well, or as statisticians say, "explains enough variance," you have no justification for making the curve more elaborate. If you make a curve elaborate enough, you can send it through every point in a graph, but that's a physically meaningless result.
The use of a sixth-order polynomial probably means, "Linear doesn't give the answer I want, and quadratic doesn't give the answer I want, and cubic doesn't, and quartic doesn't, and quintic doesn't, but when I make it sixth-degree I get the answer I want!"
OK, so consider the following statements (statements that can be seen in quite a few textbooks IIRC).
There seems to be a divergence between your statement and the above extract. Can you please elucidate?
The point of the curve was that there is a trend upwards from the 70's and a trend downwards since the late 1990's and that does not change no matter what degree (or type) of polynomial you use
here is my challenge:
please show us a third degree polynomial, that points downward at both ends!
(no need to fit it to any data. juts a graph will do!)
until you use the longer term linear fits. However if you add in the most recent temperatures across the Northern Hemisphere even the linear fit starts to fall and makes the recent trend downward even steeper for higher order fits.
this is FALSE. there is NO linear fit to the entire satellite data, that points downward.
i think you got confused and you were talking about cherry-picked short periods?!?
However I did not in the article, nor would I now, make any PREDICTIONS about future temperatures based on this or any other degree of polynomial fit to chaotic data and nor should anyone else. But for exactly and precisely the same reasons the predictions of the GCM computer models are equally flawed which is really my key point!
so the precisely same reasons, that stop your false use of a incredibly simple mathematical model from giving projections of the future, also stops the complicated models of real climate scientists from working?
sorry, but this claim again is simply false.
Now that you have popped in, Prof Jenkins, perhaps you could help in clarifying a couple of things for us. There are many here that will likely like to know the answer.
If you have read the various comments and followed links, then you will know that there is "concern" over the date at which the data in your analysis were curtailed.
So, when did you submit the piece to The Australian for consideration for publication?
Or perhaps they solicited the piece from you in the first instance. If so, what date was it solicited and when was the piece prepared by you and the date submitted?
TIA.
Jon Jenkins @56
Sir. you are wrong, wrong wrong on this count.
I have spent my career modeling complex, and in many cases chaotic systems. I have worked on Numerical Weather Prediction models, which do an excellent job at predicting the future state of the weather (a chaotic system) when used correctly. These models are used by weather agencies across the world and have demonstrable skill.
I have worked on models for turbulent flow in atmospheric flows (and more general flows) and turbulent flows around aerofoils. The flow around these objects is complex and has many of the These state of the art models are now so precise that they can be used by engineers for design work. I've also worked on dynamical models of machinery, which can have chaotic behaviour. Modelling these systems is an intriguing mix higher mathematics (fun with topology and differential geometry), physics and numerical analysis.
This is just the experience I've gained in a relatively short career. The modeling of complex systems is difficult, and the presence of chaotic behaviour means that the dynamics and initialisation of the model must be treated specially. However, it is a flat out flasehood to say these systems cannot be modeled.
As for climate modeling, if you've studied the subject you should have learned that certain statistical behaviour of the climate systems is not chaotic at all.
I suppose the oft used example is that I can predict that January in Australia will be (on average) warmer than June. Another example is the average global temperature of the last 30 years can be modeled adequately by:
T = linear trend + ARMA noise process.
A GCM is not suppose to forecast the weather on June the 14th, 2030. It's suppose to give an estimate of the statistical mean state of the atmosphere for the winter of 2030. Many of these variables (for example those found in the zonally averaged momentum equations) can be shown by analysis to exhibit no chaotic behavior once transients have died out.
Your statement is wrong, and your graph is misleading. I hope you can admit that.
"So, when did you submit the piece to The Australian for consideration for publication?"
A week or so ago.
"Or perhaps they solicited the piece from you in the first instance. If so, what date was it solicited and when was the piece prepared by you and the date submitted?"
No they did not solicit, I sent it after the article by Mike Steketee.
What has any of this got to do with the science? Why is it the AGW devotees tend to go after the man rather than debate the issue. If you really want ammunition against me personally the best bit is disclosed in my departing speech from the Parliament: I was diagnosed with a brain tumour in 2007 and need to take anti-seizure medication. Does this mean that the science I present is wrong?
As I said above, NO fit, linear, polynomial or anything is relevant when you are talking about complex process with so many degrees of freedom and so many unknown feedbacks. You can fit curves to any time scale of your choosing to give the end scale gradient you desire. However to imply that you can make useful predictions out of this is wrong no matter what side of the AGW fence you sit! All we can really say is that the climate warmed slightly from the mid 70's to the lat 90's and the climate cooled slightly form the last 90's till today.
As to chaos theory, the Russian Academy of Science has been using chaos theory for their climate predictions for some time. And I'll leave it to you to do the research as to what they predict but I would get out the warm woollies if you believe them. I have about as much confidence in that prediction as I do in the GCM ones.
Nice of you to visit this site, Prof. Jenkins.
You have by now been alerted to the fact that your graph, which purported to show warming "had completely reversed by 2008", is several months out of date.
You will also be aware that the UAH data, on which the graph is based, show the global temperature anomaly rose in September, October and November.
By November it was +0.254, right on the 30-year linear trend and pretty much exactly what one would expect after ENSO conditions returned to neutral following a significant La Nina event.
What is your response to this?
By the way, may I offer a suggestion on your expository style?
If you want people to take you seriously, it might be a good idea to avoid language like this: "The warmaholics, drunk on government handouts and quasi-religious adulation from left-wing environmental organisations..", " the fraud of the IPCC" "the infamously fraudulent 'hockey stick'" etc.
You are obviously a man of science. Such inflammatory language, however, could give the reader the unfortunate impression that you have rejected the science not because of a sober and rational assessment of the facts but because of some overriding psychological motivation like, for example, hypothetically speaking, a deep-seated ideological prejudice against government intervention in markets or a fear that your 4WD might be confiscated by faceless bureaucrats.
Another suggestion would be to offer some data and/or sources to back your assertions.
Like this one, for instance: "Cataclysmic volcanic eruptions have often placed more greenhouse gases into the atmosphere in (a) few minutes than man induces in a decade."
Atmospheric CO2 is rising by about 19 ppm per decade at present. Can you give me an example of evidence showing a volcano has ever (let alone "often") managed to do more than that in a "few minutes"?
Oh, and I'd suggest you steer clear of those polynomial curves. They're way too flexible near the end point. They can turn around and bite you right on the bum.
I was at my polite best, with temperate, courteous language in #60. And what do I get in reply, but:
Please point out to me where my supposed ad hominem attack is on you and where my request is for ammunition against you personally in my comment #60.
I see nothing of the sort. Precisely nothing! So why did you raise that nonsense and accuse me of attacking you personally?
Also, in that quotation above, the diversionary tactic of conflating the two separate issues of your illness and whether that means the science you present is wrong is, I think, contemptible. I'm not too sure whether it's an "appeal to pity" type fallacious argument or some other documented type (can't be bothered to run it down at the moment), but it is certainly a red herring. Shame.
But let's now go back one sentence from the above-quoted personal slur upon me. You also said:
Sorry, but I expect more from professors. As someone who has worked in academia and who has almost certainly submitted work to journals for peer review, you must surely be aware of the importance of research logs/lab books and document history (dates when work is carried out, submission/revision dates and the like). The questions about submission timing of your article were pertinent because of the implicit question in para 2 of #60, namely:
You appear to want to dodge around that issue in your diversions above (at least you didn't answer it above), so let me make it explicit.
Now that we know the time history on the piece (for which I thank you for supplying), given the gestation of the piece is "a week or so ago", why did you curtail the data where you did, when updating the data (freely available) to November (or even October) would have been the correct thing to do? What is the scientific basis for stopping in ... was it July? I look forward to your reply.
Prof J, please re-read your article in The Australian.
You may find that it is not just "the AGW devotees tend to go after the man rather than debate the issue."
May I, again, remind you of this passage you wrote: "The warmaholics, drunk on government handouts and quasi-religious adulation from left-wing environmental organisations...", not to mention the accusation about the IPCC authors: "...44 scientist mates who have vested interests in supporting IPCC computer modelling.." etc etc.
Having said that, I'm sure the question was genuine: many people are genuinely mystified about the apparent decision at the Australian to launch a campaign against climate science/scientists, using the same old recycled (and discredited) talking points over and over.
Any enlightenment on this point would be gratefully accepted, I'm sure.
Now, back to the science.
You claim "All we can really say is that the climate warmed slightly from the mid 70's to the late 90's and the climate cooled slightly form the last 90's till today."
First, I'm glad that you now aparently have confidence in the instrumental data pre-dating the satellite data which started in December 1979.
Second, please define "today". In other words, just precisely which cherry are you trying to pick here?
I am rather baffled by this discussion about fitting a polynomial curve to a set of known data points.
A single term polynomial fit is an average.
A two term polynomial fit is a trend line.
People do not have any problems when averages and trend lines are shown on a graph of measured data points.
A four term polynomial fit is starting to show the little wiggles in the actual data.
If you add enough terms to a polynomial fit, the original data points will be reproduced exactly.
Where people go wrong, is when they attempt to extrapolate a polynomial curve into the future.
Some people have gotten upset, because they have extroplated this curve, and know that such a drastic drop in Earth's temperate is not possible. They are correct, but their arguments are based upon a very flawed concept.
Please pay attention to this rather simple lesson about extrapolation.
Recycling
Steve Huntwork, I think the point (made by someone else here I think) is more that a high order polynomial is more heavily influenced by the data points at the end of the data series than a linear function is.
For example, both the La Nina in 1989 and the Pinatubo volcano in 1991 were associated with large drops in temperature but the polynomial curve does not fall - it had to take account of the subsequent rises in temperature. Yet there is a similar fall at the end of the series, and the curve drops like a stone.
If, for the sake of the exercise, you add 5 degrees to the latest value, the polynomial fitted curve changes direction and heads toward the sky. In contrast, the gently upward sloping linear trend line through the same data will not change very much.
The bottom line is that if you have a data set that dips down near the end and you wanted to emphasise the new downward "trend", you'd choose a high order polynomial to do it.
Playing the man is attacking specific individual on a non scientific basis (i.e credibility or status or personal issues) rather than attacking the process which is what I did. Example: the continued use of the hockey stick AFTER it was shown to be invalid including the refusal of the IPCC the retract it in later reports is a perfect example. The involvement of a select few people who have vested interests in perpetuating the outcomes of the IPCCC process i.e. more funding and kudos but mostly because it is not open to independent review!
I do not retreat from my comments about the process whereby only a few scientists have the authority to change the world by declaring "we did it" without any form of independent review or question whatsoever!
Back to the science:
Steve Huntwork said it succinctly in his post above but perhaps an explanation for the more mathematically oriented is in order, this is hard to do in this forum because I can't post symbols etc. But here goes...
The solution (i.e. T) to the climate equations of the earth is complex, it is a giant set of differential equations perturbated by cyclic, impulse and random drivers, some of which we can parameterise (i.e. solar, day/night, seasons etc) and some of which we can't (clouds, El Nino, thermohaline etc). The outcome is the solution of the Laplacian equations and can be shown by Heaviside's Rules (and related expansion theorem) to be the ratio of polynomials of greater order than the input. I will try to do this in text but it may not come out, if it doesn't make sense get a maths text and look up Heaviside's Expansion Theorem.
let g(s) be the climate equations,
if g(s) = N(s)/p(s) (where N(s) and P(s) are polynomials in (s), which they have to be because that is how the equations are parameterised, see **Note below)
then
L^-1g(s) = Sigma[r=n..r=1] [n(sr) e^(srt) / p'(s)]
Where n = number of zeros of p(s)
in other words the solution will be a massive polynomial.
(Again, see note below about real GCMs)
It follows that the T curve will also be a massive polynomial in (t) and will be best approximated by the largest polynomial possible.
However we all know that curve fitting to a section of a curve is dangerous because even if higher order polynomials fit it better they tend to oscillate around the data at the extremities. But the opposite is true of smaller polynomials which tend to ignore the accumulation of smaller poles and zeros and linearise at the extremities!
So I say again: NO curve fitting technique can claim to be more or less valid than another!
But if the AGW side are going to use curve fitting to make a point then so can anyone else!
** Note on real GCMs: Actual GCMs work differently in practice, the integrals are usually solved by quadratures, unconstrained meshes and/or finite elements. Integrals are typically approximated by cubic spline interpolation (sometimes higher order interpolations are used depending upon the order of differential). Initial values are averaged from historical data and interpolated (there's that word again) from real data sets. The issue of boundary conditions is pretty much ignored except that they must match at the cell boundaries! (see http://climatesci.org/ for technical description)
Example: the continued use of the hockey stick AFTER it was shown to be invalid including the refusal of the IPCC the retract it in later reports is a perfect example.
*snork*
The involvement of a select few people who have vested interests in perpetuating the outcomes of the IPCCC process i.e. more funding
Chortle.
only a few scientists have the authority to change the world by declaring "we did it" without any form of independent review
heehee.
NO curve fitting technique can claim to be more or less valid than another!
(silence)
if the AGW side are going to use curve fitting to make a point then so can anyone else!
sigh.
Have the comedy writers making on-line parody characters all found work, and all that's left is washed up, unfunny hacks making these parody characters?
I'm getting a comedy Jones and I need a fix! The parody character Mr Jenkins is not meeting my comedy needs!
Huh? Whassat?
Um...excuse me a minute, please.
Ahem.
Erm, apologies. Someone tells me Mr Jenkins may be real.
Terribly sorry. Apologies.
Gotta go now...um, well, buh-bye!
Best,
D
Jenkins, that was an impressive load of scienobabble, but there's one problem: you throw out lots of equations and conclude that
> It follows that the T curve will also be a massive polynomial in (t) and will be best approximated by the largest polynomial possible.
but your equations don't have T(.) in them. So what I'd like to know is, according to your understanding, T(.) is the solution to... what equation? I'm assuming that t (small letter) denotes the time period. But what's s?
Do climate modellers really use the fitting of higher order polynomials? As I understood it, they used equations based on sound physical principles. Some may have been derived by fitting curves to experimental results but these would have been more in the way of linear or logistic equations, not sixth order polynomials. Or have I got completely the wrong idea?
Jon, the hockey stick was not shown to be invalid. Read the [fine NRC report](http://scienceblogs.com/deltoid/2006/06/nas_report_on_hockey_stick_rel…).
Certainly it is true that a higher-order polynomial can better approximate the data, but that doesn't mean that a higher-order polynomial gives a better idea of the trend, else we would just set the degree to the size of the data and get a curve that goes through all the data points. For temperature data a linear trend seems to be the best idea. If you think the slope has changed over time, rather than a higher-order polynomial, you could use local regression (loess).
Prof. J, I guess we can agree on one thing - I've said on this site a long time ago that I think the debate about trends is misplaced, because it really devolves, at least implicitly, into a debate about whether the temperature can be described as a function (linear or otherwise) of the date (as opposed to energy inputs, atmospheric composition, etc).
However that does not get around the more important question: has the temperature become warmer and if so has the recent warming been largely the result of greenhouse gas emissions by human activity and can therefore be expected to continue?
Unfortunately you're yet to say anything of substance on that point.
in other words the solution will be a massive polynomial.
i am neither to deep into climate models, nor does your 2 line introduction into the subject contain enough information for me to fully understand them.
but: each of your lines seems to include a ratio of polynomials, making them not polynomials!
It follows that the T curve will also be a massive polynomial in (t) and will be best approximated by the largest polynomial possible.
i asked you above, to provide a third degree polynomial that heads downward at both ends. (when you claimed that the degree doesn t matter)
you ignored that question.
now you claim, that the largest degree is best.
Tim provided this graph with both the linear and your "trend".
http://scienceblogs.com/deltoid/2009/01/06/uahtlt5.2.png
i fail to see the advantage of the higher degree polynomial. instead both "curves" tend to be better approximations at different times.
However we all know that curve fitting to a section of a curve is dangerous because even if higher order polynomials fit it better they tend to oscillate around the data at the extremities. But the opposite is true of smaller polynomials which tend to ignore the accumulation of smaller poles and zeros and linearise at the extremities!
sorry, but 1998 surely is an extreme point. both curves fitted give exactly the SAME approximation for it!
So I say again: NO curve fitting technique can claim to be more or less valid than another!
did you ask anyone with some knowledge of statistics on this?
i have some massive doubts, that he will agree!
"Do climate modellers really use the fitting of higher order polynomials? As I understood it, they used equations based on sound physical principles. Some may have been derived by fitting curves to experimental results but these would have been more in the way of linear or logistic equations, not sixth order polynomials. Or have I got completely the wrong idea?"
Yes you have the right idea just the wrong method! We would all love to be able to solve massive differential matrices (the physical/chemical equations) directly but, unless I have missed some revolution in mathematics/physics, we can't :-(
So we have to use "tricks" to put the equations into forms which can be solved numerically. There are all sorts of "tricks" used depending upon the equation at hand. The we have to use all sorts of other trickery to solve the remaining equation sets, I have even seen one group try to use harmonic balancing which is novel seeing as the climate never reaches a stable state i.e. it is always being "driven" by some new input.
Climate models are based on a cell size of 2x4 (lat by long) (that was last time I looked at the Hansen/Schmidt code. How can we simulate the climate 100 years form now based on cell sizes of 10,000's of cubic kilometres? What about mixing and turbulence i.e. a cyclone!!
Oh and Fortran should be banned!!).
However that does not get around the more important question: has the temperature become warmer and if so has the recent warming been largely the result of greenhouse gas emissions by human activity and can therefore be expected to continue?
I wish I could answer those two questiona! The satellite data says the Northern Hemisphere may have warmed slightly but the SH has not! It also shows, depending on which curve fitting method you use, that recent cooling may have occurred. And both of these "beliefs" are borne out anecdotally (i.e. a few years ago NH was hot now they are currently freezing with record lows!)
As to the second question the AGW camp rely on the GCM outputs to say yes but for myself I simply have had too much experience with computer models of much simpler and more constrained systems (i.e. protein folding in my case) to believe this. I KNOW that we cannot even predict the structure of simple 20 AA proteins in controlled environments when we know a great deal about the intermolecular forces involved. I then have to question how we can predict the climate in 100 years when the environment is many orders of magnitude more complex.
In the end, what I am saying regardless of whether the GCMs are right or wrong we MUST become energy self sufficient within ~200 years and good basic science is the only way to do this. "Feel good" stuff won't cut it when the oil/gas/coal run out and the problem is that schemes like ETS are useless UNLESS the money goes back to the science.
Jon
Jenkins:
> So we have to use "tricks" to put the equations into forms which can be solved numerically.
As far as I know, none of these "tricks" involve starting with mathematical expressions like g(s), N(s), and t(s), and then suddenly concluding something about the temperature record T(t) out of nowhere.
So I ask again: According to your understanding T(.) is the solution to... what equation precisely? And what's s?
Or are you just throwing out mathematical gobbledygook in order to confuse?
sod:
> i asked you above, to provide a third degree polynomial that heads downward at both ends. (when you claimed that the degree doesn t matter) you ignored that question.
Heheheh.
Someone up there said:
"I have spent my career modeling complex, and in many cases chaotic systems. I have worked on Numerical Weather Prediction models, which do an excellent job at predicting the future state of the weather (a chaotic system) when used correctly. These models are used by weather agencies across the world and have demonstrable skill."
Hmm did any of the computer models predict the current cooling ....ahh NO (http://www.trac.org.au/images/temps_simple2.gif)
Tell you what make you a deal; El Nino transfers about 5PW of energy around the planet (about 50% of the total energy budget). Tell me (and the rest of the world) when the next El Nino event is going to occur and if you are correct you win the debate.
Shorter Jon Jekins:
GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING! GLOBAL COOLING!
Professor Jenkins,
I asked earlier what you meant by "today" when you claimed the temperaure had cooled slightly between the late 1990s and "today".
You've given no answer yet, and now you say "The satellite data says the Northern Hemisphere may have warmed slightly but the SH has not!"
Again, to what periods are you referring, precisely?
Do you take the November average as your latest reading, the past year, the past decade, the latest point on your 6th order polymonial, or what?
On a more general and probaly more important note, what do you think of the physics involved in all this, you know, greenhouse gases (GHGs) absorbing energy and all that.
I seems that you'd be too intelligent to insult us with the old "how could such a teeny weeny bit of life-giving cuddly carbon-dioxide do such a mean and nasty thing" line we get so often. (PLEASE don't disappoint me!!)
So do the GHGs absorb the energy? If not, why don't you think that? If they do, where does it go if the atmosphere and oceans aren't heating up? Is there some negative feedback that completely offsets both the GHG forcing and any positive feedbacks? If so, what makes you think they exist and what do you think they might be?
I'm intigued, because, despite repeated claims to the contary, it's not just the models that lead to the conclusion that antheropogenic global warming is a reality, but the physics as well.
We'll leave that to Roy Spencer, shall we?
Deltoid et al.
The bottomline is that the global temperature, however it is measured but in particular by satellite, shows cooling and this is not how you have predicted it should be.
Indeed you (Tim Lambert) and your mates including John Quiggin, hitched your reputations some years ago to the idea that the world would continue to warm.
You did this, not because you understood the physics, but because that is what leaders at the institutions you believe in were saying would be the case.
But it is not proceeding to plan. Indeed despite a continual increase in atmospheric levels of carbon dioxide, global temperatures are falling.
For more information, click here: http://jennifermarohasy.com/blog/2009/01/29-years-of-global-temperature…
re Jenkins...
>I asked earlier what you meant by "today" when you claimed the temperaure had cooled slightly between the late 1990s and "today".
The earth cools slightly but still temperatures rise.
If i was a person living in 1882 i could be witnessing the temperature cooling. Would I then expect that in 1912 that temperatures would be rising and that by 1950 temperatures would be higher than 1882.
If Jenkins believes that climate scientists can not predict the future with computer models. I hardly think it is appropriate that Jenkins should be proposing that a reduction in temperature is also a good indicator of future temperatures!
I find it a bit hypocritical. But then sceptics aren't out to prove what is actually happening, they are out to dismantle the science for political reasons.
Oh dear Jennifer, you really don't know what you're talking about, do you?
From the Hadley Centre:
Try the same exercise with the satellite data. It will still show warming.
Marohasy:
>But it is not proceeding to plan.
Ah, that old paranoid conspiracy plan!
That would be Plan 9 or 10?
paul: "Ah, that old paranoid conspiracy plan! That would be Plan 9 or 10?"
I think it's Plan 9, Paul. Here's the link:
http://au.youtube.com/watch?v=6-kCC8WUKYk
You can see Prof Jenkins at 1:56 and Jennifer Marohasy at about 2:20.
No conspiracy. Just a couple of facts:
1. Tim Lambert has for years been claiming global warming.
2. The globe is not warming.
Jon Jenkins is not only pointing this out. But the national newspaper is daring to publish him. So Tim is angry.
Socratic BS!
Readers of this discussion should find Professor de Jager's homepage very helpful. It is rich with careful science and thorough analysis. I strongly recommend it to those interested in understanding the world.
Here it is: http://www.cdejager.com/
Here is an extract from his CV:
He was director of the Utrecht Observatory, founder and first director of the Utrecht Space Research Laboratory, and founder of the Astrophysical Institute of Brussels Free University.
He was general secretary of IAU (International Astronomical Union), president of COSPAR (Intl. organization for co-operation in Space Research) and president of ICSU (Intl. Council for Science). He founded and was first editor of the journals `Space Science Reviews' and 'Solar Physics'. He is member of various learned societies, among which the Royal Netherlands Academy of Arts and Sciences, the Royal Belgian Academy of Arts and Sciences, the Academia Leopoldina (Halle, Germany), the Indian Science Academy, Academia Europaea, etc. He received honorary doctorates in Paris and Wroclaw. He was recipient of awards and distinctions among which the Gold Medal of the Royal Astron. Soc. (UK), the Hale Medal of the Amer. Astron. Soc. (for solar research, US), the Jules Janssen Medal (for solar research, France), the Karl Schwarzschild Medal (for astrophysics, Germany), the Gagarin Medal and Ziolkowski Medal (space research, S.U.), the COSPAR medal for international cooperation, etc. He is honorary member of SCOSTEP, the international organization for solar-terrestrial physics.
"2. The globe is not warming."
Anything less than a trend is meaningless. The trend, as shown by the graphs in Tim's post, is positive. I can't believe you (or anyone else) doesn't understand that.
So, to correct your comment:
1. Tim Lambert has for years been claiming global warming.
2. On any meaningful timescale, there is currently global warming.
I look forward to a post on your blog acknowleding these basic facts.
Jon Jenkins.
My bemusement remains.
Given a dataset with a discrete and limited (in a climate context) range of time values accompanied by temperature values, what justification can anyone use for the fitting of a 6th order polynomial with which to represent the interval's trend (especially as there is no consideration of any of the myriad of other parameters relevant to global climate/temperature in this graph)?
It is a simple question.
If the ordinate had been any other mundane parameter would you allow your undergraduates to use such a fit? And what of the fact that there are decades of preceeding data not included? Especially, what of the fact that including even some of these data in a refit will give a completely different 6th order polynomial? How does this simple fact justify the use of such a trendline format?
Where in common usage do other researchers employ 6th order polynomials to describe the trend in a time series?
If you seriously think that you have the advantage on both time-scale selection and on trend-fitting I hope that you might consider posting over at Tamino's. You would surely have something to teach the analysts there...
**Update:** Ian Musgrave [has more on what is wrong with high order polynomial fits](http://astroblogger.blogspot.com/2009/01/whats-up-with-polynomial-fits-…).
Professor Jenkins said
Well let's see. Using the Met Office's Unified Model, Smith et al. (Science 317: 796-799), in a "forecast" to 2014, said:
Well, long way to go, agreed; but they're doing OK so far.
BTW, you still haven't answered my polite question at the end of #64 (and implicit in my even more polite #60). I'm still looking forward to your answer. TIA.
Jon Jenkins is not only pointing this out. But the national newspaper is daring to publish him. So Tim is angry.
Tim is not angry. he is correcting errors.
Jenkins spews
"in other words the solution will be a massive polynomial.
(Again, see note below about real GCMs)
It follows that the T curve will also be a massive polynomial in (t) and will be best approximated by the largest polynomial possible."
HORSE SHIT.
It might be represented by a power series but it sure as hell isn't going to be a polynomial. The fact that you are pretending that they are the same is all the proof we need to know that you are a liar.
Furthermore, we are not getting the the actual numbers here, we are getting data which will always contain some noise. Fitting a sixth degree polynomial to noisy data is much worse than anything Mann ever did. You sir are a fraud. If you want to wander over to Open Mind, Tamio(the pro on time series) will hand you your head.
The fact is that all climate models, if you run them long enough have 5-10 years that go against any linear trend. You can see this as far back as Hansen et al 1988 if you take out your magnifying glass and look from ~1973-1983.
Jenkins accidentally blunders into a fallacy, that the climate system has infinite degrees of freedom. Because most variables have huge amounts of correlation, the actual number of degrees of freedom is relatively small especially over a global average (more averaging over space and time reduces the number of dof). Here is an example
Jenkins systematically blunders into a worse fallacy aka the stamp collector's approach that the best description of a thing, is the thing itself. Statistics is the escape from that dead end, and statistics coupled with physics allows us to ask what is producing the observations and produces a reasonable extrapolation. As we have seen Jenkins approach is incapable of extrapolation and thus useless.
Dr. Jenkins:
I can understand that yo umay not be up on the climatology processes, why the various national academies, royal societies, etc. may identify this as an issue, so I'll just ask you about the statistics.
1. Climate data is strongly autocorrelated at short intervals. What are your adjusted degrees of freedom, to take that into account?
2. What are your fit data, level by level, and your incremental changes? When do they stop being statistically significant (i.e., when do you start fitting more noise than data?)
3. What does inclusion of data beyond your endpoints do to your graph? Does your polynomial trend better fit the data for September 2008 on (or from 1978), or does the simple linear trend (or statistically significant lesser polynomial trend) fit better? (A quick plot of predictions vs. residuals will be fine).
As a low-level statistician/psychologist, I'm interested in knowing these things. I'm sure you wouldn't want to fit curvy lines to data without any interest in the actual meaningfulness of this. And of course, we know that trendlines mean nothing without associated significance levels (taking autocorrelations into account) and fit to data outside the range. I'd hate to assume that you and those making use of your work are producing meaningless propaganda.
Here's an open question for this group: How long should one beat a Red Herring before it becomes Herring Sauce?
No one denies that the planet is warming. It's been warming for about 12,000 years, with small dips here and there caused by a variety of things, but in general it is going to continue to warm until it decides to cool again and we are again under a mile of ice.
The alarmists tell us that man's production of CO2 is causing it to warm unusually fast, because the CO2 is catching the predominant 14.77 micron wavelength IR photon and causing the atmosphere to heat even more than it is by conduction and convection alone. And, we are told, the more CO2 we pump into the air, the more the air will be heated.
I guessing the folks who are saying that ( the alarmists ) didn't stay awake very well in physics class, and don't understand the mathematical term we call "non-linear".
So, here's a simple analogy.
Let's suppose you set up a baseball pitching machine on the pitcher's mound and it starts tossing balls in the general direction of the batter's plate. The pitching machine has a wobbly base, so the balls don't all go in the same exact direction but tend to fan out in an arc.
Next, let's put a catcher behind the plate. After a few pitches, we learn than one catcher can only catch about 10% of the pitches thrown, so we add a second catcher. Catcher #2 manages to catch 10% of the thrown balls that were not caught by catcher #1, or about 9% of total balls thrown. Adding yet another catcher, we find it capable of catching 10% of the balls not caught by #1 and #2, about 8.1 %. At this point the 3 catchers combined are snagging 27.1% of all the balls thrown.
Starting to get the picture? By the time we have 10 catchers at work, most of all the balls thrown will be caught.
And at that point, adding more catchers won't materially affect the number of balls caught because that number is dependent on the number of balls thrown, which is fixed. Let me reemphasize that: adding more catchers beyond a certain point can not significantly increase the number of balls caught.
Thus it is with CO2 in the atmosphere. At our current CO2 levels of about 385 PPM, the CO2 is "catching" pretty near all of the bandwidth being "thrown" at it. We call that "near saturation". Adding more CO2, even doubling the current atmospheric content, can't significantly increase the atmospheric temperature beyond the current levels, because present levels are absorbing most of the IR band associated with the molecular absorption of CO2. The "greenhouse effect" was really poorly named... it should have been called the "blanket effect".
The simple physical reality is that the CO2 molecule does not have a permanent dipole moment - eliminating a rotational mode - and leaving only a single bend mode with a
resonant wavelength of 14.77 microns, and over 99% of the IR energy radiated from the Earth in this band that can be captured by CO2 has already been captured. This leaves only 1% of the greenhouse effect already achieved by CO2
left to be achieved by further increases in CO2 .
Adding more CO2 can not increase the amount of IR radiation being projected by the warm planet, and since most of the appropriate bandwidths are already being absorbed, there's no reason to continue to worry about the effects of pumping more CO2 into the atmosphere. And the only effect from that will be your turnips growing bigger.
And you can stop beating on that Red Herring now....
Note: This, like most simplified analogies, is obviously not numerically (precisely) correct. For a more exact development of the mathematics, you have to go to something like http://arxiv.org/PS_cache/arxiv/pdf/0707/0707.1161v3.pdf
Jim Peden:
speaking of beating, "your" long-ago refuted "arguments" were long ago beaten down, refuted, put to rest, moldered, broken down into soil and replanted. Many of "your" "points" have been numbered for easy reference [1. , 2. ] and some of "your" "points" have been made into a fun game [ A.].
No one knows why you trot them out here, unless you want us to play the game to see how many points you won.
HTH.
Best,
D
If CO2 in the atmosphere had never been raised as an issue by anyone, do you think the temperature graphs would be cause for alarm?
Do you think these would appear pretty normal over any decadal span?
I tend to think that people would focus their attention to more important things- like cricket matches - if Al Gore had chosen an alternate method to be a con artist.
Sod in 94 erroneously claims:
Tim is not angry. he is correcting errors.
Silly sod.
Don't you know that pointing out errors makes Tim a HAYTUR? He is angry. He hates America...er...Australia. He's shrill. He hates freedom.
You warmers in the AGW camp are too busy being Cassandras to see the truth in front of your eyes.
Best,
D
Hmm, Gerlich and Tscheuschner eh?
See Eli here and
here (and possibly elsewhere) and see Arthur Smith here.
Gerlich and Tscheuschner eh? Don't think so.
Eli Rabbet - now THERE'S a trusted name in scientific thought. Another Isaac Newton!
Everyone ought to pay heed to what Eli has to say.
Ad hominem says it all really.
And what mud would you like to sling at Arthur Smith and his analysis of Gerlich and Tscheuschner?
Jon Jenkins is on a noble, Nobel mission to "Debunk Junk Green Science" so almost any abuse of science and mathematics can be rationalised away. He's been published by The Australian. Gaze on his works ye mighty and despair.
Why not run whatever polynomial you like through data and draw your conclusions afterwards? Jenkins illustrates why not, and Lambert gives the smoking figure for it.
In science, as I was mentioning in my blog on deciding climate trends, we try to avoid 'choice'. Objectivity is important. High order polynomials have the problems mentioned. So, as we see here, the conclusion drawn depends on whether you run a 6th order polynomial rather than a 1st order polynomial. Jenkins conclusion depends sensitively on behavior of his polynomial near its end points -- where we know that they are much less reliable. Over the central portion of the record, the fit show an even higher rate of warming than the linear trend.
The method for making an objective decision is already well-known, though apparently not to Jenkins (if we grant honesty to him). That is, you take your data and first fit a 1st order polynomial (straight line) to it and see whether the fit is statistically significant. If not, then straight lines can't be used. Either way, you may then look at a second order polynomial. See whether that has a statistically good fit. If it does, but so does the 1st order, then you have to make another test: "Is this 2nd order polynomial a significant improvement over the first order?" -- This is a standard test. If the answer is no, then you don't use the second order polynomial. Repeat this process until your higher order polynomial doesn't explain a significantly greater amount of the variance.
Go back and look at the 6th vs. 1st order fits Tim shows. The 6th is nearly the same as the 1st over most of the range of the data. It explains little additional variance, and does so at the cost of using many more parameters.
This is where we objectively say 'don't use a 6th order polynomial to draw your scientific conclusions about this data set'.
Jennifer Marohasy.
You are, according to your CV, a biologist. As such you should presumably have employed techniques such as mark-recapture estimations of population size and of survival, and you should therefore be aware what happens at the terminus of such time-dependent data. For a variety of mathematical and physical reasons, endpoints in all sorts of data series are a nuisance.
As another example, you are no doubt also aware of boundary effects in ecosystems, or indeed in any biological system including cell culture plates and nasal passages.
So why the hell are you supporting (at #87) the use of a 5/6th order polynomial to describe the trend in a limited time series?! And supporting it you must be, because without this absurd trendline Jenkins has no case at all.
Even with it he still has no case, because the trendline is complete mathematical rubbish.
Are you seriously hitching your scientific star to this wheel-less wagon of imbecilic ignorance?
Jon Jenkins:
Not exactly. The solution would be a massive number of combined exponential functions of time with a massive number of time constants, real and imaginary. The polynomial is in the s-domain, not the time domain.
Not exactly. It follows that the T curve might be best approximated by the largest combination of exponentials possible.
Climate models are not intended to predict El Ninos any more than probability theorems predict rolls of a dice. Even though probability theorems can't predict the next roll of a dice, they can predict statistics of dice rolls. In a similar way, climate models can't predict individual El Ninos but they can predict statistics of weather.
I think Jon Jenkins should try to debate his ideas with a real climate scientist so that he can clarify his statements for us.
Bingo, Chris O'Neill.
Anyone who has even a kernel of methematical/statistical knowledge will see why Jenkins' statements are a scientific embarrassment, and anyone who doesn't understand your point is demonstrating that they do not have the nouse to be involved in the discussion in the first place.
Odds are that this won't stop them though...
>No one denies that the planet is warming. It's been warming for about 12,000 years,
You need to get out more.
There are plenty of people that deny the planet is warming.
And, of course, even if a curve *was* represented by a polynomial, it doesn't follow that a polynomial of the same degree would be remotely good for fitting to it once error's taken into account. If you take data points from a degree-100 polynomial curve, add a little noise to them, and then try to fit another degree-100 polynomial to those points, you may well get absolute garbage.
i wonder if Jenkins will return for a comment.
his take at the graph has been seriously DEBUNKED!
Totally and utterly agree with Robert Grumbine in 106.
"triple x was causing a net filter to block the post.
Our idiot communications minister plans to make such filters compulsory.
"
You should try doing breast cancer outcomes research
"but when it's winter in the northern hemisphere, it's summer in the southern hemisphere, and vice versa"
which reminds me... from the retinue of australian deniers telling me about the record cold in europe and the US as proof of global cooling... are you guys having a cool summer down there, or are the Usual Suspects just doing their usual thing? i can't find a lot of commentary on aussie weather from up here.
"In effect you can use any curve fitting method you like to achieve any effect you desire. But this has never stopped the AGW people from using it! "
and that's the problem in a nutshell. that a person in a position of authority, one who apparently has a background in computer analysis of biochemical science no less, does not know the difference between modeling and curve fitting, is really depressing. it's a basically huge difference in science which predates computers by a few centuries.
let me try it again, for those who requested clarification and for those who unfortunately need clarification but shun it. there is modeling, and there is curve fitting, and the two are not at all the same. curve fitting is where you take a bunch of data and find a "predicted" curve on the sole basis of what fits the points best. it makes no assumptions regarding the underlying mechanism, nor does it result in any conclusions. the curve could be anything you want; polynomial, power series, exponential, any combination, or just something made up: "Y=Xsquared+1 except where X is evenly divisible by three, then Y =0". whatever makes the curve fit the points best. the purpose is typically to allow you to mathematically predict what Y would be for a value of X without having to go read it off the graph. for instance, if you are trying to calibrate an instrument to a sensor which has a nonlinear response curve. instead of having to go look up the graph and find what 7 millivolts means in terms of whatever you are measuring, you can use your fitted curve to calculate what the value is for each meter reading, and make a new meter face which, instead of 7 millivolts, says "3 gigawatts per cubic furlong" or whatever.
polynomials have very specific properties, as discussed at length; every time you up the degree (x squared to x cubed to x to the fourth, etc.) you add another bend into the curve; obviously for x points, if you use a polynomial of x-1 degree, you can fit every single point perfectly; but what the fitted curve does in between the points can be pretty crazy, and at the ends it's completely unconstrained. raising the degree of the polynomial is like wrestling with a bump in a carpet; you can't make the error go away, you just make the curve fit the points better and shove all the error into the spaces between the points and at the ends. since you've done such a good job fitting to the datapoints and you don't have any estimate of the error between the points, you claim victory. so you use high degree polynomials with about a ton of salt, and only if you must.
whereas, modeling goes the other way; you have an ideal of how what you are looking at works, and you develop the equations that describe that, then you adjust the parameters of those equations to fit the data. it's the basis of all science, finding generalizable mathematical formulae which describe observed behavior; computers only enter into at as an improvement over slide rules and hand calculation. for instance, you note that falling objects have a constant acceleration; you therefore model that distance fallen = half the "acceleration" mutiplied by the square of time falling [d=(at^2)/2]; so you fit that equation to your measurements of distance d and time t and adjust the value of "acceleration", a, to find the value that best fits the curve to the data; then you publish this as your estimate of a real physical parameter, the acceleration of gravity. in the day of newton, galileo, et al this was laboriously worked out by hand; nowadays, you use Excel or something similar. voila, "computer" modeling.
the difference here is that with a model, you are really saying something firm and predictive about the behavior of the system, even at values of X where you don't have data. nobody's got time/distance data gathered over 24 hours of free fall, but if you plug that amount of time into d=(at^2)/2 with the value of a calculated from the data via the previous paragraph, any physicist will bet his life on the answer. (yes, it turns out acceleration a varies as a function of the distance between the two masses; as that was found by similar modeling process, it emphasizes the point)
and of course, the modeling procedure(s) for climate prediction is discussed at length all over the literature and on the net, and it's modeling not curve fitting. individual measurements of all the myriad components that go into atmospheric behavior and oceanic behavior and the interface between the ocean and atmosphere from years of observations and calculation of these simpler processes are assembled into a grand picture; before computers this would be way too complex to dream of solving, but with computers even if you can't find each parameter precisely, you can calculate a range of results which are realistically probable.
to argue against modeling "But this has never stopped the AGW people from using it!" by conflating it with curve fitting on the grounds that "you can use any curve fitting method you like to achieve any effect you desire" is like arguing that you can't predict the rate of growth of money in a bank account which has a guaranteed interest rate and is federally insured, because "look at the stock market! financial predictions are totally unreliable!"
"many of the processes are chaotic in nature and CANNOT (and I mean in the mathematical sense CANNOT) be modelled at all.
"
yes; for instance the movements of individual molecules in the atmosphere are extremely chaotic in nature, therefore all those people who think they can speak of fictions like "air presure" and "diffusion rates" are deluded, at best.
Lank:
Jim Peden:
I wish you guys could get your act together and tell the same story. You might achieve more than zero credibility if you do.
"the Russian Academy of Science has been using chaos theory for their climate predictions for some time. And I'll leave it to you to do the research as to what they predict "
do you mean khabibullo whatshisname the space scientist's prediction of solar cooling? or the 6 guys' prediction of cosmic ray mediated cooling? or both? or another one? is one of them official? which one uses chaos theory?
"The bottomline is that the global temperature, however it is measured but in particular by satellite, shows cooling and this is not how you have predicted it should be. "
See also:
"if the olympics doesn't set a new record for running the mile, this is proof that the continual progression of better athletic performance by humanity since Roger Bannister has begun to reverse itself"
"At our current CO2 levels of about 385 PPM, the CO2 is "catching" pretty near all of the bandwidth being "thrown" at it. We call that "near saturation". "
irrelevant. if all the earth's radiation is absorbed within the first mile, say, of atmosphere, then you have a new radiating body consisting of the much warmed atmosphere at that level; if you absorb that radiating energy in the upper levels of the atmosphere it warms the atmosphere. if it is completely absorbed in the next, say 5 miles of atmosphere (miles 2-6) then that level of atmosphere becomes you new warming body. etc. all the way up. the question is does IR escape from the topmost layers of the atmosphere? and it does, so that is the opportunity for increased CO2 to reduce heat radiation.
look at it another way; given your model, the actual radiating surface of the earth is not the outside of the dirtball, but a level of atmosphere below which all the IR is captured. by definition, essentially. as you add more CO2, the height at which that level of atmosphere is situated becomes higher. higher levels of atmosphere have lower temperatures, therefore lower radiated energy.
"For a more exact development of the mathematics, you have to go to something like
"
hmm..
"1. There are no common physical laws between the warming phenomenon in glass houses
and the fictitious atmospheric greenhouse effect, which explains the relevant physical
phenomena. The terms "greenhouse effect" and "greenhouse gases" are deliberate misnomers.
2. There are no calculations to determinate an average surface temperature of a planet"
Oy.
Chris O'Neill (January 11, 2009 10:31 PM) #118
"I wish you guys could get your act together and tell the same story. You might achieve more than zero credibility if you do."
I'm sure Jim Peden will get back to you once he's counted the number of times "water vapor/vapour" is mentioned in the IPCC's latest report, so he can re-word this claim (on another site): "Curiously enough, the UN IPCC reports don't even mention water vapor, since it is technically not a 'gas' in the atmosphere."
Could take a while though - it's mentioned a heck of a lot of times.
There might be a further delay while he tries to work out what all those baseball catchers do with their balls, which is surely the big problem with his "simple analogy".
which reminds me... from the retinue of australian deniers telling me about the record cold in europe and the US as proof of global cooling... are you guys having a cool summer down there, or are the Usual Suspects just doing their usual thing?
Mixed bag. Spring was well above average, but the one month of summer so far was below average.
Spring 2008 Temperatures (Sep-Nov):
Both daytime maximum and overnight minimum temperatures for the season were well above normal with seasonal anomalies for Australia of +0.85°C (11th highest since 1950) and +1.01°C (4th highest) respectively.
http://www.bom.gov.au/climate/current/season/aus/summary.shtml
Summer Temperatures (December 2008 only):
National maximum temperatures were 0.37°C below the long-term average (25th lowest of 59 years)
http://www.bom.gov.au/climate/current/month/aus/summary.shtml
And for the first quarter of 2009:
The national outlook for daytime temperatures averaged over the March quarter (January to March) shows a moderate shift in the odds favouring warmer than normal conditions over the tropical north. Overnight temperatures are also likely to be higher than normal in the tropical north, as well as for most of the rest of WA.
http://www.bom.gov.au/climate/ahead/temps_ahead.shtml
Overall outlook for summer 2009 seems to be about average.
Latest UAH satellite data show an anomaly in December of +0.411 for the Northern hemisphere.
Of the period covered by the satellite data (December 79 to December 2008) that's higher than all but two months prior to 1998 and 0.08 higher than the average for the period from January 1998.
When were those cold records set?
BTW the SH anomaly was -0.045 (seems to be tracking ENSO which is still a bit on the la Nina side of things but what would I know).
The global anomaly was +0.183 in December after +0.251 in November, so I guess this means we're heading for another ice age. Again.
And it could be a bad one!
The 6th order polymonial "trend" shows the global anomaly will be -21.5, more or less, by the end of 2020.
Sigh.
Z @ 117:
Or, in terms closer to the esteemed professor's field: in a molecular dynamics model of a protein, the movement of the thousands of water molecules involved, and their interaction with the protein, is entirely chaotic. Yet the model still manages to maintain the protein's structure.
z at #116, and Tristan at #126 have added to the pile of homework that Jon Jenkins has to do in order to even begin to justify his position.
I hope that, in the spirit of scientific progress, he will attempt to do so.
Re #127 ... as might one also hope of Jennifer Marohasy, don't you think (in light of #107)?
Jennifer Marohasy writes:
No it doesn't.
http://www.geocities.com/bpl1960/Ball.html
http://www.geocities.com/bpl1960/Reber.html
I hope that, in the spirit of scientific progress, he will attempt to do so.
Yer dreamin'.
Best,
D
"many of the processes are chaotic in nature and CANNOT (and I mean in the mathematical sense CANNOT) be modelled at all. "
No, that is not actually the problem. The problem is in mathematics itself.
Almost all chaotic process can be modeled very well by some differential equation.
The problem is with differential equations. Any complicated differential equation is going to be chaotic. The problem with predicting the weather is not the lack of a good model. The weather equations are a very good model. The problem is that the weather equations are every bit as chaotic as the weather.
I should say what I mean here. Any solution to a differential equation depends on initial conditions (or a boundary conditions). That is to say we need the current state of the system. Chaotic means that very very small change in the initial condition (or boundary condition) can lead to a very large change in the solution. Since there is always error measuring the state of a system, this is a real problem. You can say what the solution would be if your initial value had no error, but it does so you are left with many possible solutions, which diverge from one another fairly quickly.
Thus we are pretty good at predicting the weather 2 days from now, but very bad at predicting the weather 2 weeks in advance (by two weeks the possible solutions have diverged very far from one another).
When this blog was started I had hopes for some real discussion of the science of climate modelling, it started well with a discussion about curve fitting and the various merits and predictive capability (or lack of!). Then we got onto the "hockey stick" and opinions as to whether the two Macs, the Wedgman Report and "absence" of the Med Warming/Little Ice Age has affected its credibility ( I believe they do and others don't). But like the IPCC process which I also have issues with it may be interesting but it is not science.
But then we descended into the usual personal stuff. The initiator of the blog insists on attacking me personally to discredit me somehow. To be brutally honest it is irrelevant that I have studied medicine, science, have an honours degree, a PhD and non degree enrolments in other stuff. And yes as ashamed as I am to admit it (perhaps it was the drugs...) I also have also dabbled in the humanities and have a Dip. Ed. It is irrelevant because I claim no special expertise in climatology and do not ask nor care the background of the other contributors to this blog (including Tim) other than the quality of their contributions to the debate. So I had given up....
Then post #116 came up and got to the real guts of the issue about curve fitting versus the prediction. The poster "z" is spot on: CURVE FITTING IS USELESS FOR PREDICTING RATHER IT IS ALL ABOUT THE MODELS AND WHAT THEY PREDICT!!
So if Tim and any others want to have a real discussion about climate change then let's discuss the models and their skill at predicting the future. The only publicly available source code is NASA (Hadley and the Germans do not release their code). I would really like to get a look at the CSIRO code but unfortunately they will not release it to the public so if anyone has access?
I am happy to participate in either anecdotal discussions about the IPCC collection of models and their relative merits including their previous success (or lack of!) at predictions. But I would rather have a detailed technical discussion about the models and mathematics themselves. So if you are up for it lets go.....
Jon Jenkins,
First I'd like to know if you understand that GCMs are physical models and not statistical models, and if you understand what the difference is?
Second, do you understand that the utility of global AOGCMs is that they can produce projections based on unpredictable human inputs and aren't useful for discrete predictions, and, again, what the difference is?
Just saying yes isn't enough. I'd like to hear some explication in order to estimate your degree of understanding.
But I would rather have a detailed technical discussion about the models and mathematics themselves. So if you are up for it lets go.....
To get up to speed, here. From the phraseology and argumentation, one suspects there's a learning deficit.
Best,
D
Jon Jenkins:
I'm certain that the professional climate scientists (including university lecturers) who write for realclimate are "up for it" so why don't you ask them about your concerns about climate models? They actually have an active thread on climate models running at the moment. I'm sure they will answer your questions about climate models much more clearly and concisely than anyone who usually writes on this blog. So if you are up for it then do it.
Chris:
prediction: this will be like the amen chorus at CA when I offered to sign them in to the dendrochronology listserv to argue their opinions.
Best,
D
Jon Jenkins:
> When this blog was started I had hopes for some real discussion of the science of climate modelling,
You started by throwing out a bunch of equations that made absolutely no sense. When I asked you what s meant and how you managed to conclude something about T(t) when neither T(.) nor t was mentioned in your previous working, you simply ignored the question.
> But then we descended into the usual personal stuff.
Oh, but of course it's our fault. Boo-hoo.
But then we descended into the usual personal stuff.
J. Jenkins
And of course your article in The Australian was an exemplary model of ad hominem and ideology free civility. For example:
warmaholics (several times)
They are then further doctored by a secret algorithm to account for heat-island effects. Reconstructions such as the infamously fraudulent "hockey stick" are similarly unreliable.
The warmaholics, drunk on government handouts and quasi-religious adulation from left-wing environmental organisations
the fraud of the IPCC,
Etc.
Pathetic.
It is indeed interesting to read the âdiscussionsâ in these many forums on climate change. I am heartened by the suggestion that the current list of bloggers might be interested in a proper scientific discussion on the models used to obtain the projections put forward by the IPCC. That could well be followed by a second discussion on the action of CO2 which has already been partly addressed by âzâ at #121. I will make a comment on one of zâs points which is to say that the rise in the height at which the green house gases reverse their role and become âcoolersâ rather than âwarmersâ, implies a colder radiating region. This reversal amounts to collisional excitation followed by radiation some of which goes out to space, rather than absorption of radiation followed by collisional heating of the surrounding air at the lower levels. Am I correct in assuming that this is because you are assuming that the radiation from the higher air sample is similar to that of a black body and therefore radiates significantly less according to a T^4 power law? This is not necessarily so, since gases do not, and in fact cannot, radiate directly as black bodies. They can only radiate at frequencies determined by their internal structure, at rates which are determined by the so-called âoscillatorâ strength of the various transitions available from each upper state, which, in all cases, has a population determined by the Maxwell-Boltzmann law from quantum statistical mechanics and which it is true depends on T but not in the same way as the black body function (Planckâs Law). So the rate of radiation both at each frequency and in total is much more complex than a simple black body calculation.
Back to the curve fitting versus modelling debate which is also very interesting. My first point is that to use any polynomial to fit what is essentially a randomly behaved function with an underlying but unknown trend, is to court nothing but disaster. There is no point in doing this unless one is trying to demonstrate the correctness of an hypothesis which defines a functional expression that is expected to correctly represent that trend. One then uses a function with as many variable parameters as necessary to suit the hypothesis and which allows a âbestâ fit in terms of minimising the residuals using something like a Levenberg-Marquet algorithm to obtain the values of those parameters at that minimum.
If the functional behaviour of the global temperature for instance is unknown, but is not totally random or chaotic over the long term, it must be assumed that it is at least dependent on a set of slowly varying drivers, be they increases in green house gases, changes in the earthâs orbit or variations of the solar constant. Thus in any short period of say two years, or whatever length of time seems reasonable given the known possible variation periods, these drivers of climate or global temperatures must be fairly constant over this short time, but will probably bear no relationship whatsoever, to their values at ten years on either side, let alone 100 years. Thus taking an average of the global temperatures over each two year period in the form of distinct averages or running averages, must surely provide the best available approximation to the variational function. Using any assumed function such as a polynomial, must invariably impose a constraint on the so derived mean temperature at every point, where part of that constraint is imposed by the value of a point which may, in the case of this temperature curve, be 50 years away. This is the type of error, although not quite the same, which James Hansen made in deriving his so-called, and unfortunate, âHockey Stickâ graph.
With regard to models and curve fitting, while it is true that they are by definition quite different, the differences in the cases of the climate models are probably not all that well defined. This is because of the nature of the problem, which is obviously very complex, and the fact that a lot of the input parameters required to make the models work at all, are unknown and must be derived by attempting to âtuneâ their values (see IPCC AR$ 2007 Chapter 8) so that the resultant output from the models fits known data. This is not all that different from the curve fitting process where the coefficients in a polynomial function are also derived from fitting the thing to known data. In fact, the outcome is a little more positive in favour of the curve fitting, because, although I have criticised them above, at least there is no attempt to extrapolate to a date beyond the known data. In the case of the models, the values of the parameters obtained from âfittingâ, if you like, to known data from earlier years, are then used in the models to project beyond the known period and in this case out to 50 or 100 years. This not a criticism, it is just a fact and must be done since there is no other method available to the modelers for deriving these unknowns.
Another relationship to curve fitting will also be found in the solutions to the many hydrodynamic and other second, or higher order, differential equations which have to be solved. It is extremely unlikely that any of these equations will admit of analytical solutions and the numerical processes used in finding solutions will undoubtedly involve the use of functional approximations to a solution from which parameters will be obtained from the boundary conditions. And so onâ¦..
OK. You can now all tell me that I donât know anything about the solutions to the models. Go for it. I was a bit disappointed, actually, to find that some people ducked Jon Jenkinâs challenge to become involved in a debate and simply referred him to âRealClimateâ or somewhere for advice which he may or may not need.
John Nicol
> some people ducked Jon Jenkinâs challenge to become involved in a debate
Jon Jenkins ignores the simple questions we pose to him, therefore it's our fault. Boo-hoo, boo-hoo, boo-hoo...
WotWot:
> And of course your article in The Australian was an exemplary model of ad hominem and ideology free civility.
But WotWot, the phrase "The warmaholics, drunk on government handouts and quasi-religious adulation from left-wing environmental organisations" is the "We hold these truths to be self-evident" of Liberal Fascism! Therefore it's OK.
John Nicol:
So James Hansen derived the âHockey Stickâ graph. Looks like we learn something every day.
Yes it's so disappointing that Jenkins won't debate with people who know what they're talking about.
Dear bi -- IJI,
You said:
"Jon Jenkins ignores the simple questions we pose to him, therefore it's our fault."
bi, I don't believe I mentioned it was your fault. I simply observed that people had not engaged in a debate which could have been interesting to observe and take part in. There seemed to be a number of respondents to Jon Jenkin's posting who obviously are very well credentialled in climate science. It would be great if they could expand on their various comments and provide us with some insight into the difficult science and modeling that is used to project the values of global warming in 50 or 100 years time. This is very important I believe, since there will soon be very large economic changes arising from the ETS and from restrictions which many councils are now placing on coastal developments because of predicted climate change, including sea level rises.
Unfortunately, for reasons known only to themselves, our publicly funded CSIRO refuses to divulge just how they arrive at the many predictions they are making concerning, for example, increases in storm surges on North Queensland's coast line, causing the Townsville City and other North Queensland Councils to modify their town planning. What we get from CSIRO is nothing but a disclaimer assuring us that these predictions are only the best they can do at the present time and if they turn out to be wrong, as they well might, then we can't blame them. This is OK perhaps for blue sky science where the research is simply aimed at a better understanding of a concept with the expectation that further publications and debate will lead to the correct solution. However, even there, it is necessary for the details of an experiment or theory to be published to allow anyone else to repeat the experiment or analyse the theory.
The projections by CSIRO are being made for the purpose of giving advice to government, with the clear expectation that notice will be taken of what has been said, yet no one is allowed to check their processes. Fortunately this is not the case with many others of the modeling fraternity who are very happpy to share their knowledge and even computer programmes. It is therefore doubly encumbent upon CSIRO - public funding and advice which has very serious economic consequences - to be open in providing the background to their results and complete details of their processes, to anyone who may be interested in checking.
I have asked them for a list of their peer reviewed publications on the subject - Silence!
Not good enough. There was a time when CSIRO scientists actually interacted openly with other scientists and the public. After all, it is the public who pays for both their salaries and their research funding. This seems to indicate to me at least that they have an obligation to provide proper information in a transparent and open way.
In the absence of information being available from the modelers, I had hoped that less formal advice might be available from someone else who knows all about the modelling process, CO2 etc.
-----------------------------------------------------
Chris O'Neill, you have said:
âClimate models are not intended to predict El Ninos any more than probability theorems predict rolls of a dice. Even though probability theorems can't predict the next roll of a dice, they can predict statistics of dice rolls. In a similar way, climate models can't predict individual El Ninos but they can predict statistics of weather.
I think Jon Jenkins should try to debate his ideas with a real climate scientist so that he can clarify his statements for us.â
What makes El Ninos different from any other weather phenomenon. Remembering that El Nino is just one peak in a particular combination of cyclical behaviours of the atmosphere and oceans associated in some way with the Southern Oscillation, the other extreme being the La Nina. In between, the system is in some intermediate state which will have characteristics associated with El Nino or la Nina which you are saying climate models are not intended to predict. While the IPCC admits that the models can't predict them, I don't ever recall reading that they would not be intended to do so. It is just a matter of gaining enough experience and knowledge I believe. I wonder if your analogy of a dice whose values appear as random numbers for which there are very clear statistics as you point out is really very apt. Any random event obeys similar statistics but the El Nino is not believed to be random - it is just that we don't yet know what drives it and determines when it will occur.
The El Nino climate phenomenon is clearly driven by various agencies such as heat, currents and winds which can also effect other aspects of weather (in the short term) and climate (in the long term). If the models were ever likely to be successful, it seems reasonable, from direct scientific principles, that they should first be able to master one of the strongest influences known. It is not as if the characteristics of the El Nino and La Nina are not well known, and the changes in temperature distributions, winds and ocean currents well documented. No such documentation exists for the roll of a dice! There are also probably other more subtle phenomena which are not yet recognised but which also have a significant influence on climate, though are not currently being included in the models. After all, according to the IPCC AR$ 2007 Chapter 8, the models are not able to handle clouds properly either, which means they are not capable of properly including the effects of probably two of the most powerful influences on the climate as they stand today. This must surely place severe limits on the expected accuracy of projections 50 to 100 years hence, when it is also admitted in Chapter 8, that they cannot predict the climate in 2008, using the initialising parameters obtained from the earlier 100 years before say, 1960. Even more importantly, they point out, again in Ch 8, that even if they could, it would not yet be known what this success would mean in terms of their ability to predict further into the future. There is no doubt that the models are very well constructed and are called on to provide information from a very, very complicated system. Nevertheless, the acknowledgement of their limitations included in the most recent of the IPCC reports, goes a long way towards demonstrating clearly that any information provided by the models, must be treated with extreme caution. Perhaps, and hopefully, increases in computer size and speed in the near future together with a better understanding of the physics of the action of green house gases, will soon lead to a much better modelling regime where many, if not most people, will have well justified confidence in the future climate projections provided by the models.
John Nicol
John Nicol:
Nothing.
You're reading too much into what I said. At the very least I didn't say that El Ninos were random. I'm just saying that predicting a particular El Nino is different from predicting say, average temperature over the time that several El Ninos are expected to occur.
sorry Jon, i can see why you would want to change the subject. your claims about your graph have been torn apart.
multiple posts above would just require a one line answer from Jon Jenkins, that he doesn t give, because he can t answer it.
why not simply admit, that it was an error to use the 6th degree polynomial?
or will you finally give me a 3rd degree one, that shows a similar result?
John Nicol,
What one would like from a very good dynamic model, is that it exhibit El Nino and La Nina behavior, and that the frequency and intensity of them corresponds to what is seen in the real world. That is climate.
What one doesn't need, and doesn't expect, is that the model exhibit exactly the same weather, with each storm, each bit of 'noise,' and each El Nino/La Nina matching exactly in time and intensity to each one we see in the real world.
IOW, we don't expect the models to predict the weather in the real world. We do expect frequency and intensity of weather in the models, averaged over time, to match what we see in the real world averaged over time.
That is why the modelers do ensembles - so they can average them, and therefore "average out" the weather events and leave the climate signal.
John Nicol writes:
Gosh darn that James Hansen! It was really sneaky of him to change his name to Michael Mann, too.
Yes. I accept that Chris OâNeil. Sorry if I misrepresented what you had said.
---------------------------------------------------
Lee, It is true that one doesnât expect the models to represent the much more random weather events. And from what you have just said, I believe we agree that the models should be able to account for the fairly long term (decadal) changes in âclimate/weatherâ represented in the Southern Oscillation which cycles between La Nina and El Nino conditions. Without being able to do that, they will not be very useful as climate predictors. Thanks for your reponse.
----------------------------------------
Barton Paul Levenson
OK Michael Mann derived the Hockey Stick. Sorry and apologies to any one named James Hansen.
John Nicol
But the values that come up from rolling the dice are not random - they're the result of a chaotic system, just like the weather is. The value that comes up on top depends on the way in which the dice sit in the hand, the angle and force of the throw, the amount of spin imparted, the distance from the surface they land on, the type of surface they land on, etc., etc.
Since they're chaotic, we can't predict what the next value rolled will be. But we can predict with a very high degree of confidence what an ensemble of a large number of rolls will look like. Similarly, the molecular movements at each step of a molecular dynamics model won't match those in any real system (even if we could measure such), but results are starting to come in that show that, once run long enough, they will predict quite closely the final tertiary structure of simple proteins. Again, we can't predict whether it will be raining in Memphis at 4pm on Tuesday next month, but the models do appear to represent quite well the long-term aggregate behaviour of the climate.
once run long enough,
Do you meant once iterated enough times?
meant = mean
Jon Nicol - "...I believe we agree that the models should
be able to account for the fairly long term
(decadal) changes in âclimate/weatherâ
represented in the Southern Oscillation which
cycles between La Nina and El Nino
conditions."
1) Do you mean the the models should account for changes in frequencey of El Nino or La Nina avents, something which which may happen on a decadal scale, or the events themselves as they occur on an inter-annual scale?
2) What do you mean "should account for". Do you mean the models should explain why the El Nino/La Nina events occur, what their effects are, why they beome more or less frequent, or actually predict discrete events (as opposed to simulate them)?
3) What if the models can't do one or more of these to your satisfaction?
Would some uncertainty over the nature of these events lead you to the conclusion that anthropogenic global warming is not happening?
Or do you envisage some possibility that a better understanding of the ENSO phenomenon might provide evidence that anthropogenic global warming is not happening?
4)If the answer to either the second or third questions in point 3 is "yes", what do you think might be happening to the heat energy absorbed by greenhouse gases?
If not, why is modelling ENSO so important, aside from demonstrating why no-one should get to excited about the odd hot year (like 1998) or cooler year (like 2008)?
I'll leave to someone else to ask what you mean by "climate/weather".
Cheers.
I guess congratulations are in order for David Evans.
He seems to have gotten a retrospective promotion.
It seems now it wasn't just computer programming he did back in the day, he's actually "a former adviser to the Australian Greenhouse Office, the precursor to the Department of Climate Change".
http://www.theaustralian.news.com.au/story/0,25197,24914441-11949,00.ht…
jade: "But have you seen his website?"
Classic. Don't you love the tagline? "Dedicated to debunking junk green science." The best PR heads couldn't workshop one better than that. Is it implying that all "green science" is junk science? Or is it actually closer to the truth -ie. Jenkins doesn't like greenies, so he only attacks science that appears to side with them?
"As one of the original "greenies" it causes me great pain..."
Yeah, right... Jenkin's it part of the huntin', fishin', four-wheel-drivin' Outdoor Recreation party. Can't get much more green than that, can ya?
I ran a 6th degree best fit over GISTEMP data from 1880-2008. Funnily enough, I didn't see a downward curve at the end, but a distinctly upward one.
Oh, and I extended the line to 1700 -> 2100, and I notice that the planet will be 7 degrees warmer in 2100, and that there was a global ice-age before about 1700. Nifty graphs here (down the bottom, last two images).
Another problem with the polynomial fitting is seen if you apply the graph above to the 161-game American baseball season -- and apply a 6th degree polynomial curve to that point in the middle of the season when your team has just won eight games in a row. If you do this, your trendline will "prove" that your team will win all of the rest of the games that season and every season thereafter. If your team losing eight games in a row, it might as well disband because the trendline clearly shows your team will never win another game in a billion years.
Actually, your baseball team could lose 159 games in a row--but if it wins the first and last game, a 6th-degree polynomial fit over the whole season will still prophesy victory for all future games...
Readers of this discussion should find Professor de Jager's homepage very helpful. It is rich with careful science and thorough analysis. I strongly recommend it to those interested in understanding the world.
Here it is: http://www.cdejager.com/
Here is an extract from his CV:
He was director of the Utrecht Observatory, founder and first director of the Utrecht Space Research Laboratory, and founder of the Astrophysical Institute of Brussels Free University.
He was general secretary of IAU (International Astronomical Union), president of COSPAR (Intl. organization for co-operation in Space Research) and president of ICSU (Intl. Council for Science). He founded and was first editor of the journals `Space Science Reviews' and 'Solar Physics'. He is member of various learned societies, among which the Royal Netherlands Academy of Arts and Sciences, the Royal Belgian Academy of Arts and Sciences, the Academia Leopoldina (Halle, Germany), the Indian Science Academy, Academia Europaea, etc. He received honorary doctorates in Paris and Wroclaw. He was recipient of awards and distinctions among which the Gold Medal of the Royal Astron. Soc. (UK), the Hale Medal of the Amer. Astron. Soc. (for solar research, US), the Jules Janssen Medal (for solar research, France), the Karl Schwarzschild Medal (for astrophysics, Germany), the Gagarin Medal and Ziolkowski Medal (space research, S.U.), the COSPAR medal for international cooperation, etc. He is honorary member of SCOSTEP, the international organization for solar-terrestrial physics.
It is indeed interesting to read the âdiscussionsâ in these many forums on climate change. I am heartened by the suggestion that the current list of bloggers might be interested in a proper scientific discussion on the models used to obtain the projections put forward by the IPCC. That could well be followed by a second discussion on the action of CO2 which has already been partly addressed by âzâ at #121. I will make a comment on one of zâs points which is to say that the rise in the height at which the green house gases reverse their role and become âcoolersâ rather than âwarmersâ, implies a colder radiating region. This reversal amounts to collisional excitation followed by radiation some of which goes out to space, rather than absorption of radiation followed by collisional heating of the surrounding air at the lower levels. Am I correct in assuming that this is because you are assuming that the radiation from the higher air sample is similar to that of a black body and therefore radiates significantly less according to a T^4 power law? This is not necessarily so, since gases do not, and in fact cannot, radiate directly as black bodies. They can only radiate at frequencies determined by their internal structure, at rates which are determined by the so-called âoscillatorâ strength of the various transitions available from each upper state, which, in all cases, has a population determined by the Maxwell-Boltzmann law from quantum statistical mechanics and which it is true depends on T but not in the same way as the black body function (Planckâs Law). So the rate of radiation both at each frequency and in total is much more complex than a simple black body calculation.
Back to the curve fitting versus modelling debate which is also very interesting. My first point is that to use any polynomial to fit what is essentially a randomly behaved function with an underlying but unknown trend, is to court nothing but disaster. There is no point in doing this unless one is trying to demonstrate the correctness of an hypothesis which defines a functional expression that is expected to correctly represent that trend. One then uses a function with as many variable parameters as necessary to suit the hypothesis and which allows a âbestâ fit in terms of minimising the residuals using something like a Levenberg-Marquet algorithm to obtain the values of those parameters at that minimum.
If the functional behaviour of the global temperature for instance is unknown, but is not totally random or chaotic over the long term, it must be assumed that it is at least dependent on a set of slowly varying drivers, be they increases in green house gases, changes in the earthâs orbit or variations of the solar constant. Thus in any short period of say two years, or whatever length of time seems reasonable given the known possible variation periods, these drivers of climate or global temperatures must be fairly constant over this short time, but will probably bear no relationship whatsoever, to their values at ten years on either side, let alone 100 years. Thus taking an average of the global temperatures over each two year period in the form of distinct averages or running averages, must surely provide the best available approximation to the variational function. Using any assumed function such as a polynomial, must invariably impose a constraint on the so derived mean temperature at every point, where part of that constraint is imposed by the value of a point which may, in the case of this temperature curve, be 50 years away. This is the type of error, although not quite the same, which James Hansen made in deriving his so-called, and unfortunate, âHockey Stickâ graph.
With regard to models and curve fitting, while it is true that they are by definition quite different, the differences in the cases of the climate models are probably not all that well defined. This is because of the nature of the problem, which is obviously very complex, and the fact that a lot of the input parameters required to make the models work at all, are unknown and must be derived by attempting to âtuneâ their values (see IPCC AR$ 2007 Chapter 8) so that the resultant output from the models fits known data. This is not all that different from the curve fitting process where the coefficients in a polynomial function are also derived from fitting the thing to known data. In fact, the outcome is a little more positive in favour of the curve fitting, because, although I have criticised them above, at least there is no attempt to extrapolate to a date beyond the known data. In the case of the models, the values of the parameters obtained from âfittingâ, if you like, to known data from earlier years, are then used in the models to project beyond the known period and in this case out to 50 or 100 years. This not a criticism, it is just a fact and must be done since there is no other method available to the modelers for deriving these unknowns.
Another relationship to curve fitting will also be found in the solutions to the many hydrodynamic and other second, or higher order, differential equations which have to be solved. It is extremely unlikely that any of these equations will admit of analytical solutions and the numerical processes used in finding solutions will undoubtedly involve the use of functional approximations to a solution from which parameters will be obtained from the boundary conditions. And so onâ¦..
OK. You can now all tell me that I donât know anything about the solutions to the models. Go for it. I was a bit disappointed, actually, to find that some people ducked Jon Jenkinâs challenge to become involved in a debate and simply referred him to âRealClimateâ or somewhere for advice which he may or may not need.
John Nicol
Readers of this discussion should find Professor de Jager's homepage very helpful. It is rich with careful science and thorough analysis. I strongly recommend it to those interested in understanding the world.
Here it is: http://www.cdejager.com/
Here is an extract from his CV:
He was director of the Utrecht Observatory, founder and first director of the Utrecht Space Research Laboratory, and founder of the Astrophysical Institute of Brussels Free University.
He was general secretary of IAU (International Astronomical Union), president of COSPAR (Intl. organization for co-operation in Space Research) and president of ICSU (Intl. Council for Science). He founded and was first editor of the journals `Space Science Reviews' and 'Solar Physics'. He is member of various learned societies, among which the Royal Netherlands Academy of Arts and Sciences, the Royal Belgian Academy of Arts and Sciences, the Academia Leopoldina (Halle, Germany), the Indian Science Academy, Academia Europaea, etc. He received honorary doctorates in Paris and Wroclaw. He was recipient of awards and distinctions among which the Gold Medal of the Royal Astron. Soc. (UK), the Hale Medal of the Amer. Astron. Soc. (for solar research, US), the Jules Janssen Medal (for solar research, France), the Karl Schwarzschild Medal (for astrophysics, Germany), the Gagarin Medal and Ziolkowski Medal (space research, S.U.), the COSPAR medal for international cooperation, etc. He is honorary member of SCOSTEP, the international organization for solar-terrestrial physics.
It is indeed interesting to read the âdiscussionsâ in these many forums on climate change. I am heartened by the suggestion that the current list of bloggers might be interested in a proper scientific discussion on the models used to obtain the projections put forward by the IPCC. That could well be followed by a second discussion on the action of CO2 which has already been partly addressed by âzâ at #121. I will make a comment on one of zâs points which is to say that the rise in the height at which the green house gases reverse their role and become âcoolersâ rather than âwarmersâ, implies a colder radiating region. This reversal amounts to collisional excitation followed by radiation some of which goes out to space, rather than absorption of radiation followed by collisional heating of the surrounding air at the lower levels. Am I correct in assuming that this is because you are assuming that the radiation from the higher air sample is similar to that of a black body and therefore radiates significantly less according to a T^4 power law? This is not necessarily so, since gases do not, and in fact cannot, radiate directly as black bodies. They can only radiate at frequencies determined by their internal structure, at rates which are determined by the so-called âoscillatorâ strength of the various transitions available from each upper state, which, in all cases, has a population determined by the Maxwell-Boltzmann law from quantum statistical mechanics and which it is true depends on T but not in the same way as the black body function (Planckâs Law). So the rate of radiation both at each frequency and in total is much more complex than a simple black body calculation.
Back to the curve fitting versus modelling debate which is also very interesting. My first point is that to use any polynomial to fit what is essentially a randomly behaved function with an underlying but unknown trend, is to court nothing but disaster. There is no point in doing this unless one is trying to demonstrate the correctness of an hypothesis which defines a functional expression that is expected to correctly represent that trend. One then uses a function with as many variable parameters as necessary to suit the hypothesis and which allows a âbestâ fit in terms of minimising the residuals using something like a Levenberg-Marquet algorithm to obtain the values of those parameters at that minimum.
If the functional behaviour of the global temperature for instance is unknown, but is not totally random or chaotic over the long term, it must be assumed that it is at least dependent on a set of slowly varying drivers, be they increases in green house gases, changes in the earthâs orbit or variations of the solar constant. Thus in any short period of say two years, or whatever length of time seems reasonable given the known possible variation periods, these drivers of climate or global temperatures must be fairly constant over this short time, but will probably bear no relationship whatsoever, to their values at ten years on either side, let alone 100 years. Thus taking an average of the global temperatures over each two year period in the form of distinct averages or running averages, must surely provide the best available approximation to the variational function. Using any assumed function such as a polynomial, must invariably impose a constraint on the so derived mean temperature at every point, where part of that constraint is imposed by the value of a point which may, in the case of this temperature curve, be 50 years away. This is the type of error, although not quite the same, which James Hansen made in deriving his so-called, and unfortunate, âHockey Stickâ graph.
With regard to models and curve fitting, while it is true that they are by definition quite different, the differences in the cases of the climate models are probably not all that well defined. This is because of the nature of the problem, which is obviously very complex, and the fact that a lot of the input parameters required to make the models work at all, are unknown and must be derived by attempting to âtuneâ their values (see IPCC AR$ 2007 Chapter 8) so that the resultant output from the models fits known data. This is not all that different from the curve fitting process where the coefficients in a polynomial function are also derived from fitting the thing to known data. In fact, the outcome is a little more positive in favour of the curve fitting, because, although I have criticised them above, at least there is no attempt to extrapolate to a date beyond the known data. In the case of the models, the values of the parameters obtained from âfittingâ, if you like, to known data from earlier years, are then used in the models to project beyond the known period and in this case out to 50 or 100 years. This not a criticism, it is just a fact and must be done since there is no other method available to the modelers for deriving these unknowns.
Another relationship to curve fitting will also be found in the solutions to the many hydrodynamic and other second, or higher order, differential equations which have to be solved. It is extremely unlikely that any of these equations will admit of analytical solutions and the numerical processes used in finding solutions will undoubtedly involve the use of functional approximations to a solution from which parameters will be obtained from the boundary conditions. And so onâ¦..
OK. You can now all tell me that I donât know anything about the solutions to the models. Go for it. I was a bit disappointed, actually, to find that some people ducked Jon Jenkinâs challenge to become involved in a debate and simply referred him to âRealClimateâ or somewhere for advice which he may or may not need.
John Nicol
Readers of this discussion should find Professor de Jager's homepage very helpful. It is rich with careful science and thorough analysis. I strongly recommend it to those interested in understanding the world.
Here it is: http://www.cdejager.com/
Here is an extract from his CV:
He was director of the Utrecht Observatory, founder and first director of the Utrecht Space Research Laboratory, and founder of the Astrophysical Institute of Brussels Free University.
He was general secretary of IAU (International Astronomical Union), president of COSPAR (Intl. organization for co-operation in Space Research) and president of ICSU (Intl. Council for Science). He founded and was first editor of the journals `Space Science Reviews' and 'Solar Physics'. He is member of various learned societies, among which the Royal Netherlands Academy of Arts and Sciences, the Royal Belgian Academy of Arts and Sciences, the Academia Leopoldina (Halle, Germany), the Indian Science Academy, Academia Europaea, etc. He received honorary doctorates in Paris and Wroclaw. He was recipient of awards and distinctions among which the Gold Medal of the Royal Astron. Soc. (UK), the Hale Medal of the Amer. Astron. Soc. (for solar research, US), the Jules Janssen Medal (for solar research, France), the Karl Schwarzschild Medal (for astrophysics, Germany), the Gagarin Medal and Ziolkowski Medal (space research, S.U.), the COSPAR medal for international cooperation, etc. He is honorary member of SCOSTEP, the international organization for solar-terrestrial physics.
It is indeed interesting to read the âdiscussionsâ in these many forums on climate change. I am heartened by the suggestion that the current list of bloggers might be interested in a proper scientific discussion on the models used to obtain the projections put forward by the IPCC. That could well be followed by a second discussion on the action of CO2 which has already been partly addressed by âzâ at #121. I will make a comment on one of zâs points which is to say that the rise in the height at which the green house gases reverse their role and become âcoolersâ rather than âwarmersâ, implies a colder radiating region. This reversal amounts to collisional excitation followed by radiation some of which goes out to space, rather than absorption of radiation followed by collisional heating of the surrounding air at the lower levels. Am I correct in assuming that this is because you are assuming that the radiation from the higher air sample is similar to that of a black body and therefore radiates significantly less according to a T^4 power law? This is not necessarily so, since gases do not, and in fact cannot, radiate directly as black bodies. They can only radiate at frequencies determined by their internal structure, at rates which are determined by the so-called âoscillatorâ strength of the various transitions available from each upper state, which, in all cases, has a population determined by the Maxwell-Boltzmann law from quantum statistical mechanics and which it is true depends on T but not in the same way as the black body function (Planckâs Law). So the rate of radiation both at each frequency and in total is much more complex than a simple black body calculation.
Back to the curve fitting versus modelling debate which is also very interesting. My first point is that to use any polynomial to fit what is essentially a randomly behaved function with an underlying but unknown trend, is to court nothing but disaster. There is no point in doing this unless one is trying to demonstrate the correctness of an hypothesis which defines a functional expression that is expected to correctly represent that trend. One then uses a function with as many variable parameters as necessary to suit the hypothesis and which allows a âbestâ fit in terms of minimising the residuals using something like a Levenberg-Marquet algorithm to obtain the values of those parameters at that minimum.
If the functional behaviour of the global temperature for instance is unknown, but is not totally random or chaotic over the long term, it must be assumed that it is at least dependent on a set of slowly varying drivers, be they increases in green house gases, changes in the earthâs orbit or variations of the solar constant. Thus in any short period of say two years, or whatever length of time seems reasonable given the known possible variation periods, these drivers of climate or global temperatures must be fairly constant over this short time, but will probably bear no relationship whatsoever, to their values at ten years on either side, let alone 100 years. Thus taking an average of the global temperatures over each two year period in the form of distinct averages or running averages, must surely provide the best available approximation to the variational function. Using any assumed function such as a polynomial, must invariably impose a constraint on the so derived mean temperature at every point, where part of that constraint is imposed by the value of a point which may, in the case of this temperature curve, be 50 years away. This is the type of error, although not quite the same, which James Hansen made in deriving his so-called, and unfortunate, âHockey Stickâ graph.
With regard to models and curve fitting, while it is true that they are by definition quite different, the differences in the cases of the climate models are probably not all that well defined. This is because of the nature of the problem, which is obviously very complex, and the fact that a lot of the input parameters required to make the models work at all, are unknown and must be derived by attempting to âtuneâ their values (see IPCC AR$ 2007 Chapter 8) so that the resultant output from the models fits known data. This is not all that different from the curve fitting process where the coefficients in a polynomial function are also derived from fitting the thing to known data. In fact, the outcome is a little more positive in favour of the curve fitting, because, although I have criticised them above, at least there is no attempt to extrapolate to a date beyond the known data. In the case of the models, the values of the parameters obtained from âfittingâ, if you like, to known data from earlier years, are then used in the models to project beyond the known period and in this case out to 50 or 100 years. This not a criticism, it is just a fact and must be done since there is no other method available to the modelers for deriving these unknowns.
Another relationship to curve fitting will also be found in the solutions to the many hydrodynamic and other second, or higher order, differential equations which have to be solved. It is extremely unlikely that any of these equations will admit of analytical solutions and the numerical processes used in finding solutions will undoubtedly involve the use of functional approximations to a solution from which parameters will be obtained from the boundary conditions. And so onâ¦..
OK. You can now all tell me that I donât know anything about the solutions to the models. Go for it. I was a bit disappointed, actually, to find that some people ducked Jon Jenkinâs challenge to become involved in a debate and simply referred him to âRealClimateâ or somewhere for advice which he may or may not need.
John Nicol
Why use a old fashion polynomial. I have found that a Fourier series will more accurately model the data and based on my initial calculations it has shown that there is nothing to worry about it is just a bunch of oscillations.
Chris, it's my understanding that a Fourier series can be designed to fit any data, no matter how noisy. (I may be wrong, as it's been a while ...)
However, it's really only an appropriate method if you start from the assumption that you're modelling a bunch of oscillations.
@jade
>
>So when do temperatures in the lower troposphere hit absolute zero?
>
Hopefully they won't ever reach zero. I'm not sure what that would mean for humanity if the troposphere were that hot. If you mean the anomalies, they regularly pass through zero.
@Tim
As you have both maths and computing background I'd like to see a bit more info on what curve is a good fit to that data. Linear trend is simplistic and I seem to remember from my Uni days when we were finding curves to fit the el-nino cycle we had identified after processing a huge bunch of sea level air pressure data that the important criteria were finding something representive of the data and maybe ignoring the end points ??
I can understand why someone wouldn't like that (4th degree isn't it ? did I see some one say it was 6th ?) fit line. But to my eye it doesn't look that bad.
I guess it depends on what you want.
The linear trend is perhaps useful for determining whether there is an increasing tendency is it not ??
Another type of curve (sine ?? polynomial ??) might be good if one was investigating evidence of cycles in the data would it not ???
@Tim
nvm mind about fit curves I've followed some links and found some good ones.
Just for the record, this graph got re-used again by the Heartland Institute in January 2009, p.5 ... only they made it worse yet. That one says:
"Since 2005, global temperatures have given back most of the warming that had occurred since 1980."
It also says, at bottom left, in minuscule characters:
SOURCE: THE UNIVERSITY OF ALABAMA AT HUNTSVILLE and at bottom right: ANDREW BARR / NATIONAL POST
and adds:
"According to NASA satellite dataâthe only timely global temperature record--the global temperature in 2008 is no higher than it was in 1979."
Not only some words, but gridlines are different, suggesting that someone regenerated the graph with different labels.
OOPS, sorry, broken link.
See Jan 2009 E&CN, p.5.
All of you can rattle on about "science" and "consensus" and "objectivity", but the elephant in the room is "human nature"; i.e. the lust for power and money; enthusiasm for the big lie when it means your side gets rich and dominates by force.
The "global warming" context has never had anything to do with science, but about using the force of law to determine who makes money and who pays. It's like any other crisis structure (war on poverty, war on drugs, war on terror, socialized medicine, and on and on).
If you're going to dedicate yourselves to science, top wasting ordinary folks' toil and time; investigate whey charlatans like Albert Gore go about bleating and beating, and promoting the greatest fraud of our time. Again, the "global warming" context has nothing to do with science -- it's all about money and power. If you can't recognize the historical patterns, you ought to find another line of work.