People send me stuff.
Tonight I got an email that contained a link to a paper that takes on the wonky claims related to barycentrism and Earth’s climate, specifically as it relates to Nicola Scafetta’s 2010 and 2012 papers. This new paper taking on the Scafetta claims will be published in the Journal of Atmospheric and Solar-Terrestrial Physics, April 2014. The author is Sverre Holm Department of Informatics, University of Oslo, Norway.
Abstract, some graphs, and discussion/conclusion, along with a link to the paper follows.
Abstract
It has recently been claimed that there is significant coherence between the spectral peaks of the global temperature series over the last 160 years and those of the speed of the solar center of mass at periods of 10-10.5, 20-21, 30 and 60-62 years. Here it is shown that these claims are based on a comparison between spectral peaks in spectral estimates that assume that the global temperature data contains time-invariant spectral lines. However, time–frequency analysis using both windowed periodograms and the maximum entropy method shows that this is not the case. An estimate of the magnitude squared coherence shows instead that under certain conditions only coherence at a period of 15-17 years can be found in the data. As this result builds on a low number of independent averages and also is unwarranted from any physical model it is doubtful whether it is significant.
…
Discussion and Conclusion
Scafetta (2010) claimed the global temperature series for the last 160 years to have
spectral lines at 21, 30 and 62 years. Time–frequency analysis shows that the lines are
time-varying (Figs. 1 and 2) and very different from the nearly constant lines in the
time–frequency plot for the speed of the center of mass of the solar system (SCMSS)
(Fig. 3).
The supposed periodicity around 30 years in Scafetta (2010) is not really
present in the climate series at all and could be an artifact due to a combination
of model overfitting and smearing due to the time-invariance assumption which has
been forced on the data. The claimed spectral peaks by Scafetta (2010) for the global
temperature series are therefore not reproducible if proper consideration is taken of
the time-varying nature of the data. The only significant coherence between the cli-
mate series and the sun’s movement that was possible to find was at 15-17 years (Table 1). However, both the low number of independent averages that it builds on as well as the lack of a physical explanation for this coherence, makes us hesitate to claim that it is significant.
===============================================================
Looks to me like “game over” for claims of Barycentrism controlling Earth’s climate. Clearly this was a case of pulling a signal from noise that is just an artifact of the process, much like Mann’s special brand of math that made hockey sticks from just about any red noise input data.
Full pre-print of the paper here: http://arxiv.org/pdf/1307.1086.pdf
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
“Anthony Watts says:
March 12, 2014 at 9:56 am
@climatereason ditto. If somebody can reproduce his work, show why it’s either right or wrong, I’ll adjust my opinion accordingly.”
If we’re talking of the paper Scafetta, “Empirical evidenceforacelestialoriginoftheclimateoscillations
and itsimplications” http://www.fel.duke.edu/~scafetta/pdf/scafetta-JSTP2.pdf then I have been able to reproduce all figures in that paper but one. That was his Fig. 8. (Power spectra of the speed of the Earth relative to the Sun and of the speed of the center of mass of the Earth–Moon system relative to the Sun).
Reproduction or not is not the point of my criticism of the paper, it is the method used and the assumptions that have been made.
TonyB:
Well there is a cycle which has ~60 year components to it. So the answer is yes, but not in a sine wave sort of way.
http://climatedatablog.files.wordpress.com/2014/02/cet-monthly-with-full-kernel-gaussian-low-pass-annual-15-and-75-years-filters-with-a-15-year-savitzky-golay-projection.png
Anthony you say “Periodicity is one thing, I have no qualms with Milankovitch cycles, as they are easily calculable and broadly reproducible to show the variation in watts/m2 on the surface of the Earth over those time scales.”
I think we do well to look for the patterns i.e periodicities in the data itself without worrying too much about the underlying processes. That is where the modelers go wrong – they assume they know how the system works and build their models accordingly.
The chief uncertainty in my cooling forecasts is where we are with regard to the 1000 year periodicity. I think we are just a little past the peak . This is supported by Figs 3 and 4 in the last post at
http://climatesense-norpag.blogspot.com
I would appreciate your opinion on this working hypothesis.
Steven Mosher says:
March 12, 2014 at 8:58 am
In the end you either have to re do all the work yourself– or trust someone else… some paper, some post, some words ..some figures.
. who you gunna trust? Leif. Why? its simple. He shows his work, he shares his data. When I check it I get the same answers. Is Trusting Leif an appeal to authority? No, its practical experience at work.
______________________
Excellent, Steven. Not only is Leif’s work verifiable, but his statements in re scientific reasoning and methods, bolster confidence in his works (not that confidence adds to the science.)
For instance, someone recently praised Dr. Svalgaard’s “position” about an issue as being correct. Leif responded that a “position” is meaningless and only the data counts.
RichardLH
http://www.vukcevic.talktalk.net/PF-latest.gif
Steven Mosher says:
March 12, 2014 at 10:13 am
“There is no cycle IN THE DATA.
the data is the data
There is no trend in the data
There is no mean in the data
The data is the data. nothing more and nothing less. It is what it is and nothing more.”
True(ish) – but definitely misleading.
There is a daily cycle in the data.
There is a monthly cycle in some data.
There is a yearly cycle in the data.
There may well be more than a yearly cycle in some data.
How you treat the data to display it will serve to hide or emphasise those cycles. Normally it is common to use low pass filters to conceal both the daily and yearly signal.
If you use slightly longer low pass filters you can observe a ~60 year signal in the data. All that is being done is reveal what is present in the data
http://climatedatablog.files.wordpress.com/2014/02/hadcrut-giss-rss-and-uah-global-annual-anomalies-aligned-1979-2013-with-gaussian-low-pass-and-savitzky-golay-15-year-filters1.png
You want the R or spreadsheet methodologies for that? That has been done my me and others elsewhere but can be done here if required.
Dr Norman Page says:
March 12, 2014 at 10:16 am
“I think we do well to look for the patterns i.e periodicities in the data itself without worrying too much about the underlying processes. That is where the modelers go wrong – they assume they know how the system works and build their models accordingly.”
I agree strongly with that statement. Modellers presume that they know the answer to all the factors they need to consider then plug in values to make it all work. Does make the assumption that they have correctly identified all of the factors though.
ren says:
March 12, 2014 at 10:21 am
“http://www.vukcevic.talktalk.net/PF-latest.gif”
Curve fitting is not something I am particularly comfortable with. It makes a wild assumption that some mechanism exists that gives it some meaning. Until that is shown then it is not a very useful exercise.
Martin Lewitt says:
March 12, 2014 at 6:06 am
cumulative influence on the sun.
——————————————-
Cumulative is a key thought. That is how the sea ice increase in Antarctica appears to me. I tried to explain that to some warmists in relation to what is happening in Antarctica holds greater significance to the globe than the changes in the Arctic. The southern sea ice increase is a cumulative affect which will be felt over time.
@RichardLH
And I think this is a great loss of information, daily temperatures have the cooling response of the surface when the Sun goes down everyday at every location we have a station, and the yearly data shows how temperatures “reset” to some minimum every year, land temperatures have little to no memory of previous years warming.
Some of the comments here typify the best of WUWT, which exemplifies the practice of the scientific method as I was taught it. Scaffeta has a barycentric hypothesis, gets some data, appears to confirm his hypothesis. Others point out two problems. One, inconsistency with well understood and well tested Newtonian gravity (or,mid you will, implausibly large effects from implausibly small forces without a plausible amplification (or resonant) mechanism. Two,Mathis new paper reanalysing the data and showing much of the prior result is apparently an artifact rather than real. Neither disproves completely the Scafetta hypothesis, but together suggest it very unlikely to be true. And so research properly moves on to more fruitful topics.
Lord Monckton points to a very important one, the source of the seeming 60 years cycle Akasofu and Loehle point to. To mutter about ocean PDO and AMO is to perhaps quantify the observation, but not explain it. Stadium wave is a similar quantification, but which goes further toward an explanation in coupled ‘resonant’ phenomena. All of which, failed hypotheses and sharpened questions both, is far advanced from the rather pathetic model based science is settled meme of the IPCC and mainstream climate science a la Dessler positive cloud feedback with r^2 of 0.02, Trenberth’s missing hiding heat, shindell’s if we fiddle the model forcings enough we can save high sensitivity contrary to what Lewis and Crok have shown, and on and on.
WUWT, Climate Etc., Climate Audit, and their equivalents are better with more severe ‘peer review’ than Science, Nature, Nature Climate Change, and all the other journals that Climategate emails specifically said would be coopted and corrupted.
The Internet changes everything for the better.
RichardLH says:
March 12, 2014 at 10:04 am
From a statistical point of view you are correct. There are more things than statistics though. A simple low pass filter will remove higher frequencies and display any cycle longer than 15 years which is what I have attempted to show.
This discovers an ~60 year signal in the data. It is not curve fitting, it is data reduction.”
We are not debating whether there are ~60 year cycles in the data sets. That is well known.
The issue is whether there is coherence between the data sets at that period and others. That could be the first step towards finding out if one is caused by the other. But correlation/coherence is not the same as causality, so it is only a first step.
Scafetta thinks there is coherence and causality. I didn’t find coherence despite being able to reproduce all the relevant plots in his 2010 paper.
I’m not remotely qualified to discuss the math here, but one thing confused me in the abstract of the paper, namely that it was analyzing the “speed” of the sun’s center of gravity over time. “Speed” is the magnitude of velocity, but either has to be measured relative to some reference point or reference frame. What’s the frame of reference for the speed of the sun’s center of gravity in this context?
My other confusion perhaps comes from an incomplete grasp of the barycentric theory for solar perturbations. As I understand it, the issue is not just that the velocity vector changes magnitude, but direction as well, and the center of mass of the system (which is what the sun itself is orbiting around) follows a highly eccentric path, ranging from passing through the sun’s nucleus to being as much as the radius of the sun outside its limb). Doesn’t the major hypothesized effect on solar activity involve this eccentric orbital motion in some way perturbing the sun’s body and internal processes?
Apologies if this has been covered amply elsewhere, but the reference to “speed” seemed odd to me, and limiting possible mechanisms and correlations to variations in speed vs motion of the orbital center seemed to be missing a good part of the point of the Barycentrists.
Thanks for any light anyone can shed on this for me!
New Zealand climate variability
New Zealand climate, particularly rainfall and wind patterns, shows systematic variations at different time periods. Due to the short instrumental record, most of the identified variation is strictly weather and not climate, and includes: a quasi-biennial oscillation (QBO) associated with sea-level pressure and meridional (north-south) flow around New Zealand; the ENSO pattern with periods of 3-8 years; a decadal pattern strongly correlated with the 11 year Schwabe sunspot cycle; and cycles with periods of 18-22 years that also correlate well with the Hale magnetic solar cycle. A 70-80 year pattern linked to the PDO is also evident, which some have correlated to the 60-120 yea Gleissberg cycle that is associated with modulation of the Schwabe cycle amplitude. Proxy data also suggest
the presence of a 200-220 year de Vries solar cycle (also known as Seuss Cycle).
http://www.grassland.org.nz/publications/nzgrassland_publication_67.pdf
Is it possible to ignore that which is actually happening? Polar vortex moves per a few days over the Europe.
““I think we do well to look for the patterns i.e periodicities in the data itself without worrying too much about the underlying processes. That is where the modelers go wrong – they assume they know how the system works and build their models accordingly.”
I agree strongly with that statement. Modellers presume that they know the answer to all the factors they need to consider then plug in values to make it all work. Does make the assumption that they have correctly identified all of the factors though.”
###################
Both wrong. This is not how modelers work. There is no evidence that they work this way so your theory about how they work is un supported.
When we look at how they really work ( when you actually get off your ass and read their code, go to their conferences, listen to their talks, read their papers) you will see that
1. They do not PRESUME they know all the factors. They know there is missing physics.
2. You dont assume you know how “the system works”, you model what is known.
The earth rotates. you model that. It orbits the sun. you model that. The earth has oceans
you model that. It has land. you model that. it has mountains. You publish a paper saying that you cannot model the andian mountain range and you note how your model diverges from reality. The atmosphere has gases. you model that taking knowledge from the engineering
of radiative transfer. You note that you cannot use a LBL model and model transfer using a band model. You run comparions between LBL models and band models. For years you study this.
At every turn you note what you dont know. The earth has clouds. You try to model that. you fail.
You plug in a parameterization. An old engineering trick.
Bottom line neither of you has any idea how modlers work
Start here. dont be stuck on stupid
http://mitgcm.org/
The bottomline. You dont start assuming that you know everything. you start with what you know.
first principles physics. you turn that into code. you run the code and you find out that you
dont know everything. You find out that you need to add chemistry. you find out you need to model volcanos. If you dont understand something ( say cloud microphysics) you have to leave it out. yu know you leave it out. you actually DOCUMENT what you cant do and what you can
I am intrigued by this: “The coherence plots in Figs. 4 and 5 and also in Table I are based on shorter [than 60 year] windows. ”
Does that work?
Presumably, the coherence has been calculated using Fourier transforms? If so, how can you expect to find anything useful about the 60 year period if you do not include any 60 year cycles? My memory of such things is hazy these days but won’t that just lump more “noise” in with the signal and lower the coherence?
graphicconception’s comment about Fourier transforms reminded me of something important from my signal- and image-processing days: If you just grab a random time series, starting and stopping the interval abruptly, you’re going to get all kinds of artifacts from the rectangular “window function” you’ve applied. There are all sorts of window functions (Hamming, Hanning, Blackman, etc, etc) that attempt to minimize this by essentially fading in and then fading out the signal gradually within the interval of interest, to avoid spectral leakage caused by the window function itself. I didn’t see any mention made of what windowing was performed on the data as part of the spectrogram processing.
A modeler explaining what a model Cant do
RichardLH
‘Modellers presume that they know the answer to all the factors they need to consider then plug in values to make it all work. ”
Wrong. He can’t even be bothered to check whether his proclamation is true.
Of course its false. Just listen to a modeler describe the shortcomings of models
When you falsely portray your opponents of claims to knowledge,
When you accuse them of hubris,
When you accuse them of making stuff up.
be careful. you are likely spotting your own flaws.
Sverre Holm says:
March 12, 2014 at 10:49 am
“We are not debating whether there are ~60 year cycles in the data sets. That is well known. ”
Actually you will find that there are significant number of people who will dispute that the ~60 year signal exists at all. I get many such complains as though it is not there even though the graphs clearly show its presence. I apologise if you do accept it.
“The issue is whether there is coherence between the data sets at that period and others. That could be the first step towards finding out if one is caused by the other. But correlation/coherence is not the same as causality, so it is only a first step.”
Indeed. That is why I extended the set to include the Shen PDO reconstruction which does appear to continue the series out to the 1400s. Now what is needed is something that matches to that sort of periodicity and with a plausible mechanism.
Steven Mosher says:
March 12, 2014 at 11:34 am
The models allow for very little (if any) natural periodicity in their workings. They are completely unable to reproduce the observed ~60 year signal in the data. They make big assumptions that CO2 is the main driver for what has been observed. So far reality has not followed that path.
Monckton of Brenchley says:
March 12, 2014 at 12:21 am
It is not improper to look for patterns in physical observations, for they may (or may not) reveal a physical law.
…
And it is climate science that first gave us the notion that has come to be known as mathematical chaos – the observation that in certain objects, the climate arguably among them (Lorenz, 1963, Giorgii, 2005), even the most minuscule perturbation of the initial conditions at any chosen t-zero can exert a disproportionately large influence on the evolution of the object over time.
Thanks Lord Monckton for a good summary of the case. The chaotic-nonlinear Lorenz aspect is really the only chance for survival of the barycentric idea in any form. The system in question would be a weakly forced nonlinear oscillator. Problem is, it is the other sort, the strongly forced nonlinear oscillators (e.g. human heartbeat) which give more-or less constant frequencies amenable to wave analysis techniques. Weak forcing means possibly very complex relation between forcing and emergent frequencies, possibly making proof of cause-effect linkage impossible, at least by spectral analysis techniques discussed up to now. So an unprovable refuge, like CAGW, GMO harmful effects etc. in an ever more remote future.
Amatør1 – Thanks for your remarkably lucid and concise explanations(!) Very interesting about the orbital angular momentum argument; the spin-orbit coupling was indeed what I was thinking of. I had a mental image of the sun being slung around hither and to, “sloshing” as it were, but didn’t make the connection to describe my mental image in terms of changes in angular momentum.
It’s still a little hard for me to visualize there not being some sort of sloshing, when the center of the sun’s orbit keeps changing like that, but I don’t remotely have the math chops to calculate it out. (I was never any good at calculus in the first place, and the last time I did anything with it was 35-40 years ago, in college/grad school 🙂 You make a very good point, though, that the sun is basically just in free-fall, like any other orbiting body.
So, by definition, all that’s left are tides, which I assume are proportional to the vector sum of the gravitational pull of all the planets at that point. I wonder how much the tidal forces distort the shape of the sun? I suspect a relatively insignificant amount, compared to its overall dimensions. The apparent correlations between sunspot numbers or other measures of solar activity and global temperatures are nonetheless very intriguing; we just don’t know what the underlying mechanism is that gives rise to solar variability.
Thanks again for your exceptionally clear-headed explanations, they’re much appreciated!
@Steven Mosher: You describe how model development should work. Unfortunately, the practice is different. Two years ago I pointed out an incorrect treatment of water evaporation/condensation in CAM5 – you commented on it 6/29/2013. It is still there.
Anthony, I think a reasonable first step here is to look at the moon. You know, just pondering this a bit more. It has a huge gravitational effect on the earth, moving water, heaving the earth crust etc. Would it not make more sense to propose/test/rule out lunar gravitational effects as a model? Then expand a model into ares of decreased sensitivity and lower signal?
I live with the tide here out east. It effects everything. What I do not know is if it has been the subject of study in relation to climate. (Why not? Everything else has.)
IMO that is where Scafetta should move his modelling. Walk before running.
As far a him concealing his data? I don’t know anything about that since I am not close to the subject and the data exchange but that seems to have really put a burr under your saddle. I would say this, the idea that something a fundamental as gravity has something to do with the movement of the oceans and the effect on CO2/CH4/O2 production is not way out science. I think most reasonable people would agree with that as a general concept. A better man than me will have to do the research.
It seems to me that the paper is a bit of contrivance, a directed laser beam contrivance by somebody with an axe to grind. Particularly since we now have a commenter named “Holm”. The paper stands on its own, notwithstanding, and on its own it is more of the image of a body blow to an image of barycentrism.
Reblogged this on Tallbloke's Talkshop and commented:
.
My my, the rhetoric levels are running high. Sceptics to be stabbed with icicles, or at least labelled as ‘Deniers’ b y captains of industry and leading politicians, unproven deep ocean warming to be guessed in ‘Hiroshimas per second’ and now ‘Death Blows’ to work-in-progress theories.
Svere Holm.
Why write this paper? Of all the things to study, why this?