Bishop Hill The big news down under
Jun 17, 2014 Climate: solar I hope everyone is reading the series of posts by David Evans and Jo Nova about their new hypothesis on why variations in solar irradiance apparently have such a limited effect on the planet's temperature. It's probably fair to say that many sceptics have scratched their heads on this subject from time to time, but the team from down under have gone the extra mile, coming up with what is starting to look like a fascinating explanation, namely that there is a delay between the change in irradiance and subsequent changes in temperature. They hypothesise further that this may be something to do with changes in the Sun's magnetic field.
It's too early to say whether this all holds up of course, but I'm certainly going to be keeping a close eye on it.
Reader Comments (109)
"Whenever an argument against solar forcing has been made by the Warmists it has always been predicated on an immediate reaction to TSI or whatever measure being discussed. That never made any sense to me..."
Yeah, that's why a change in TSI (night time versus day time) doesn't affect the temperature either! ;-)
Don't over look the fact that we have not seen the full detail of David Evan's work ( well Lubos may have seen it all). There are quite a few episodes to go yet.
A theory that requires a mysterious "Force X" to explain something that can't be detected, which is required to justify the overall theory, is not on a good footing from the word go. In other words, an ad hoc 'Force X' is required, that is exquisitely balanced, that cancels out all short terms changes that one would expect to be attributable to TSI. Well it's possible, in the sense that lots of things are possible, but if occam's razor were to be applied, it would easier to conclude that there is no effect, because we can't measure one, and therefore no need for "Force X". Or in other words, is this another example of a curve fitting exercise?
Beware rebirth of the notoriously famous and influential Elliot Wave Principle coming from a gold speculator. Beware too of the Timewave Zero primal scream plots of Terrence McKenna types who rekindle the real science of old school Tim Leary, but trumped up beyond its means, from mere psychology into parapsychology. Are we observing the world or just ourselves projected out into it, heroically seeking social status and thus female sexual choice in our favor?
David Appell's response is curious, in computing the response I notice he uses a no-feedback climate response.
You too are ignoring my caveat "to first order."
"I think this should be treated with caution. I'm always suspicious of "hunting" for cycles. There are many ways to get cycles or frequencies of 11 or 20 or so years. It's a bit like numerology!" --timheyes
Yes, and possibly more than a bit. My concerns: (1) This may turn out to be subtly disguised wiggle-matching; (2) It's based on a revolutionary new technique; (3) Log-log plots are involved. (4) What if the notch is an artifact of the noisy TSI signal or its processing?
Did these "other scientists" publish work refuting Robitailles published work?
Scientists rarely respond to cranks, nor waste their time publishing papers refuting them. And journals wouldn't waste the valuable page space.
To me lubos's best argument is that a perfect 11 year notch filter should display an extra response when the solar cycle is 10 or 12 years but it doesn't. So either the filter somehow tunes itself to the sun or it doesn't exist. My money is on the 2nd option.
Planet X kidnaps Godzilla to test it's own new beast:
http://youtu.be/PFRYG4NYFc0
I've asked before if we could have, ahead of time, a CO2 centric explanation for any cooling at the end of the current 'pause'. It looks as though this solar centric explanation would do that. A predictive skill test awaits both.
@ ianl8888 12:07 AM
You are missing the point completely. I don't want to divert from the current topic, so we could continue this on the discussion section.
Your adjectives are uncalled for and inaccurate. Any individual or group of people can realize that a better publishing system is possible, and they are very free to voice their opinions. Even more, they should voice them loudly, so that other people become aware of the situation and alternatives. It does not follow that they should have the resources to implement such a system. As for the practical methods, you are wrong again, I gave you a very good example: http://arxiv.org/.
So is this more Von Neumann's elephant curve fitting or will this one have any predictive quality?
Oh and I do wish everyone would have "correlation does not imply causation" stuck on their computers (or the clisci version "even without correlation I'm 97% certain there is causation").
"To me lubos's best argument is that a perfect 11 year notch filter should display an extra response when the solar cycle is 10 or 12 years but it doesn't."
Look at the graph, Evans isn't saying it's a perfect notch filter, the bandwidth of the attenuation is pretty wide.
I have had a read of the David Evan's posts and also read Lubos' post. As some of you may know, I am a geophysicist so I am very comfortable with the analytical techniques being used and I use them everyday in my professional work.
As Lubos points out, David is treating this as a convolutional problem. He defines the output signal as the temperature data and the input signal as the TSI. Note (again as Lubos points out) that the two signals can be anything you like. The operation that David is performing is a deconvolution in order to deduce the mathematical function that maps the input signal to the output signal. In the frequency domain, deconvolution is performed by dividing the amplitude spectra of the two signals. In other words, at any frequency the amplitude of the function is the ratio of the output divided by input signal amplitudes at that frequency.
The problem is that because this analysis can be performed on any two signals over a similar domain (here, over a similar time range), no physical relation is required or implied - the two signals can be completely unrelated and a transfer function will still exist that would map one signal to the other. As Lubos points out, the notch is actually telling you that there is a peak in TSI with a period about 11 years that does not exist in the temperature data set, according to this analysis. That suggests that TSI variations over short periods have no measureable effect on Earth temperature. (Note the word "measureable", ie not detectable with the available data)
The second problem is the deduction of the "delay" to the notch filter. David Evans has not looked at the phase spectra of the two signals (we would be looking at the difference in their phase spectra through deconvolution), stating this cannot be done with the available data. Herein lies a significant problem. David's initial analysis assumes no delay - what I would call a zero phase signal. From the type of notch filter deduced, the filter time domain response dies off reasonably quickly. David rightly points out that for the "transfer function" to be physically real it must be causal ie the sun must cause the earth to warm. The Earth cannot start warming before the sun causes it to warm. So he simply shifts (delays) the filter by enough to make it causal, discovering a time delay of around 11 years (and somewhere between 10 - 20 years. This argument is a tautology. You can shift the output from any
deconvolution of two unrelated signals in this way and then claim the result is causal. The only way you can demonstrate you have discovered a real world physical effect would be to directly estimate the delay from the phase spectra of the two series. As David states this cannot be done with the available data, then there is no proof that the "transfer function" is any more than a tautology at this stage.
If David can use his OFT method to robustly determine the cross-correlation trace of the TSI and the temperature data and the result shows a causal filter with a similar amplitude spectrum to the notch filter shown already, then I would agree he was on to something. In the meantime, the argument is fatally flawed and I agree with Lubos.
None of the carbon dioxide dominate models predicted the pause of global warming, so it is logical to look more closely at internal variability and the sun. Even if Evans' hypothesis is not the final answer, this is how science advances.
Thanks for pointing us to the Lubos thread. He has made a couple of important mistakes (it is a big theory, and there is a lot to read). We will be replying soon on joannenova.com.au to clarify them. In the meantime we're answering questions as fast as we can on the site.
Its unfortunate that Lubos did not contact me first. The errors in his post could have been easily fixed before he published it. Nonetheless, it's good to see debate being generated. Hopefully we can make it a more productive exchange in future.
David Evans,
If you are reading this thread: in order to demonstrate causality, how do you estimate the phase spectra difference between TSI and temperature datasets?
We would be expecting a linear phase shift between temp and TSI of -2*PI*t where t is eleven years (in suitable units, of course) if the causality argument were to hold up in the data analysis. Otherwise I cannot see how the delay can be defended as anything other than a tautology. If you cannot estimate the full operator ie amplitude and phase spectra, I cannot see how your arguments can be supported from the data.
If you have, or are about to, explain this elsewhere a link or comment to that effect would be helpful.
Regards,
ThinkingScientist
"I've asked before if we could have, ahead of time, a CO2 centric explanation for any cooling at the end of the current 'pause'."
First, there hasn't been any cooling. Not on the surface, and certainly not in the ocean. The seas continue to rise, which indicates ice is still melting and the ocean is still warming.
There's been a slowdown in surface warming, but then, the rise in surface temperatures is not expected to be monotonic with increasing CO2. Other factors influence it - aerosols, ENSOs, volcanoes, solar variability, ocean cycles, etc.
There have been slowdowns before, and there will be slowdowns again. They hardly mean CO2 isn't a greenhouse gas, and the slowdowns say nothing about the ultimate warming CO2 creates.
David Appel: So, you believe that unpublished comments from "scientists" outweigh a peer reviewed published article.
And you are a science writer?
What unpublished comments? There have been lots of published papers examining the slowdown and its causes.
David Appel: I was not talking of the "pause". Please read a little more carefully. I have only been talking of Robitaille's paper on black body radiation.
But, now that you mention the pause. Your:
"the rise in surface temperatures is not expected to be monotonic with increasing CO2."
Please show one model that did NOT have a monotonic rise in temperature, on a decadal scale, before the pause.
Currently, only about 5% of models can show the pause. That means that 95% of models show a monotonic rise in temperature, on a decadal scale.
In a couple of years, that 5% will fall to zero %.
David Appell: "First, there hasn't been any cooling."
In response to my post -straw man. Hang your carbon-centric ideas on another peg matey.
Robitaille is obviously a crank -- if there really were problems with Planck's law or Kirchhoff's law, it would have been apparent long ago and physicists would be all over it, understanding the problem. They aren't.
Hansen's famous 1988 paper did not have a monotonic increase.
By the way, it's far from clear that the data models have been accurately portraying surface temperatures. It depends on how underrepresented areas are infilled. Cowtan & Way, which uses kriging to infill temperatures -- a method some thing is superior -- shows 0.19 C of warming over the last 15 years.
David Appell:
1. Robitaille published a peer reviewed paper. You say this paper is wrong, but can't give any peer reviewed paper that refutes him. For what it is worth, Robitaille shows that changing what the cavity is lined with, changes the radiation. Technically, he is saying Kirchoff (and Planck) are correct, but only if you use garphite as the lining of the cavity.
2. Sorry, Hansen 1988 shows that temperature rise is monotonic at decadal scales. Scenario C goes flat after 2000, but only becasue emissions go to zero. Scenario A and B both show, on a decadal scale, a monotonic rise in temperatures.
http://www.skepticalscience.com/Hansen-1988-prediction.htm
David Appell: C&W show a warming of 0.119, not 0.19, since 1997.
http://onlinelibrary.wiley.com/doi/10.1002/qj.2297/full
This rise is also still outside the AR5 envelope.
http://climateaudit.org/2013/11/18/cotwan-and-way-2013/
Again: if there was anything to Robatille's claims, physicists would have found them decades ago.
He is a crank -- and quite clearly a crank -- and that makes him a waste of time. As such, it's up to him to prove his cranky ideas are right, not for others to prove they are wrong -- because there is an endless supply of cranky ideas, and disproving them all would be a full time job.
Hansen's model, now almost 30 years old, clearly doesn't show a monotonic rise in surface temperature.
Again, it's not clear whether the models that provide the surface temperature anomaly are accurately capturing all the surface warming. Here is the amount of warming for recent periods (NEITHER of which is representative of climate):
15 years:
HADCRUT4 - 0.11 C
Cowtan & Way - 0.19 C
20 years:
HADCRUT4 - 0.21 C
Cowtan & Way - 0.29 C
and these are for periods influenced by nonclimate features, especially (oceanic) weather of ENSOs (big El Nino in 1997-98, big La Nina in 2010-11).
Periods that aren't representive of climate can't be used to prove or disprove claims about climate.
Models solve boundary value problems, not initial value problems. So they are not good in the short-term where the initial state matters a great deal. But they are good in the long-term, where equilibrium is determined by energy balance.
Cowtan and Way's numbers give a warming of 0.15 C since December 1997.
A period that, again, is not representative of climate change, but of short-term factors like ENSOs, PDOs, AMOs, etc.
David Appell: Nope. C&W show a trend of 0.119, Jan/1997-Dec 2012 . table 3. This is about 0.07 above Hadcrut4
Hybrid .119 ± 0.076
David Appell:
He is a crank -- and quite clearly a crank -- and that makes him a waste of time. As such, it's up to him to prove his cranky ideas are right, not for others to prove they are wrong -- because there is an endless supply of cranky ideas, and disproving them all would be a full time job.
As Einstein said, it only takes 1 paper to prove Robitaille's paper wrong. Do you have that paper?
Hansen's model, now almost 30 years old, clearly doesn't show a monotonic rise in surface temperature.
Hansen's paper clearly shows, on a decadal scale, in scenario A and B, a monotonic rise. Please look at that chart a little more carefully.
Your data stop in Dec 2012. Mine do not.
And it makes no sense to combine C&W with HadCRUT4, since they are different methods. That's the whole point of Cowtan & Way -- an attempt to make up for HadCRUT4's shortcomings.
If you think Robatille is right, go ahead and believe him. No one else does, so don't be surprised when people dismiss you as a fool.
Hansen's graph does not show a monotonic increase. And it's over 25 years old by now -- climate models have advanced enormously since then.
You keep focusing on short-term intervals that are not representive of climate. Climate models calculate *climate*, not noise.
Focus on the noise if you want -- most of us are focusing on climate.
david appell: My data? No, C&W data. Please try and be a little more accurate.
If you are using your own calculation, you should have said so. So, at least you admit you are wrong on the 0.19 number. Now show the calculation for your 0.15 deg rise since 1997, please.
Who said combining C&W with HadCRUT4? I said that C&W gives a number about 0.07 deg C over the HadCRUT number for the same period.
Also, please point out on the decadal scale, where hansen's A&B scenario are not rising monotonicly.
And could you point me to a paper that disproves Robitaille's paper?
David Appell: Perhaps you do not understand the definition of "climate".
According to the WMO, climate does need to be measured in 30 year chunks. (Section 4.8.1)
The optimal period for temperatures is often substantially shorter than
30 years, but the optimal period for precipitation
is often substantially greater than 30 years
and discussing reference periods and normals:
In general,
the most recent 5- to 10-year period of record has as
much predictive value as a 30-year record."
http://www.wmo.int/pages/prog/wcp/ccl/guide/documents/WMO_100_en.pdf
David Appell:
If you think Robatille is right, go ahead and believe him. No one else does, so don't be surprised when people dismiss you as a fool.
Without contrary evidence, I will assume that Robitaille is correct. If given contrary evidence, I would re-evaluate.
A fool is someone who believes that something is wrong, with nary a shred of evidence to back up that belief.
Give C&W a rest please.
Les: My data comes from Cowtan & Way -- they have been updating it all along, and it's now current to March 2014.
http://www-users.york.ac.uk/~kdc3/papers/coverage2013/series.html
should be:
According to the WMO, climate does NOT need to be measured in 30 year chunks. (Section 4.8.1)
"Without contrary evidence, I will assume that Robitaille is correct."
Have you looked for the contrary evidence?
"As Einstein said, it only takes 1 paper to prove Robitaille's paper wrong."
And it took many to prove Einstein right. That continues to this day.
Who has proved Robitaille right?
Les: Short-term noise isn't representive of climate, but of noise -- effects like ENSOs that, over the long-term, cancel out. This is why climate scientists use longer periods when discussing climate change.
What the WMO said was, "...the mean is an incomplete description of the climate." Section 4.8.1 is about establishing a baseline, not making conclusions about climate. Note that none of the major datasets uses less than 20 years to establish a baseline, and all but RSS (I think) use 30 years.
david appell: Have you looked for the contrary evidence?
Yes, I have. I assume so have you, but are unable to find any.
David Appell: Who has proved Robitaille right?
Right stick, wrong end. Robitaille is right until proved wrong, as Einstein said.
Remember Einstein also said that 1000 papers would NOT prove him right, but it would only take one paper to prove him wrong.
"...about establishing a baseline, not making conclusions about climate".
17-18 years is long enough a time period to be significant at climatic scale. It is well longer than a decade. If we look back, the IPCC used just one multidecadal time block - 70s-'98 for its headline conclusions.
"Robitaille is obviously a crank." --David Appell
"Copernicus is obviously a crank. --theologians
"And Galileo, too." --ibid
"Vavilov is obviously a crank." --Lysenko
"Wegener is obviously a crank." -- geophysicists
"Warren & Marshall are obviously cranks." -- gastroenterologists
The faster the ad hominem arguments spew, the weaker the science. But give me an honestly deluded crank over a deliberate employer of fallacies, anytime.
"First, there hasn't been any cooling. Not on the surface, and certainly not in the ocean. The seas continue to rise, which indicates ice is still melting and the ocean is still warming."
Melting sea ice doesn't cause sea level rise. (sigh)
And ocean levels can rise through thermal expansion (ice melt is not required) and have done so for 20,000 years. Unless we've entered a new ice age, one wouldn't expect sea levels to suddenly stop rising or even by stable. Why would one expect their behave to be so perfectly tuned?
I think if you have to use silly arguments and misdirections to support your case, you don't have a very good case.
Lubos further comments in an unrelated blog post discussion thread:
http://motls.blogspot.com/2014/06/obamas-commencement-speech-and-illusion.html#comment-1439139985
"In effect, he's just calculating the response function needed to transform the observed total solar irradiance to the global mean temperature - it's tautological, pretty much Fourier transform of the data to the parameters of his model. It's guaranteed to work well enough because it's Fourier transform back and forth. In this sense, the model is maximally fine-tuned i.e. unlikely - the value of the response function for every frequency has to be a particular number for the "model" to produce the right results.
The response function "looks" similar with various choices of periods etc. so it might confuse you into thinking that it is a law of physics. However, if you think about the patterns that are universal and the accuracy with which they are universal, you will realize that it's mostly the fact that the 11-year frequency is suppressed that is universal. Not surprising - one needs this suppression to erase the 11-year signal from the global temperature data."
@ Patagon 7:43am
>I gave you a very good example: http://arxiv.org/.<
And I've said that http://joannenova.com.au/2014/06/big-news-part-v-escaping-heat-the-three-pipes-theory-and-the-rats-multiplier/ is a much better one, as it is accessible to a much larger audience
This thread has two points - Evan's new hypothesis (still unfolding, perhaps fatally flawed) and his new publishing method (quite brave)
I understood you exactly
End of my end of story
Lubos wrote:
[The response function "looks" similar with various choices of periods etc. so it might confuse you into thinking that it is a law of physics. However, if you think about the patterns that are universal and the accuracy with which they are universal, you will realize that it's mostly the fact that the 11-year frequency is suppressed that is universal. Not surprising - one needs this suppression to erase the 11-year signal from the global temperature data."]
I don't know what point he's trying to make here.
All Evans has said is that if the input to the climate black box is the sun's input and the output is the global temperature then the transfer function has a notch, centred around 11 years. He then goes on to explore how this transfer function could be happening.
Lubos is just agreeing with the first part, he isn't refuting anything AFAIUI.
Jo and David have posted a response to Lubos. To me, it looks like Lubos is wrong and didn't really read the paper carefully.