« Previous | Main | Next »

A breakthrough in long range forecasting?

Paul Hudson | 16:10 UK time, Monday, 22 March 2010

I am indebted to Dr Jarl Ahlbeck, from Abo Akademi University, Finland, who contacted me about his fascinating new piece of research relating to this winters severe cold across much of Europe, and a possible link to the very low solar activity we have been experiencing.

I am aware that there is a hugely varied readership of my blog; those who are very well informed about weather and climate, and others that have an interest in the subject but would struggle with some of the details contained in scientific papers. I have thus asked the author to summarize the main points of the research, and will include a link to the paper for those that feel brave enough to look into it themselves.

Dr Ahlbeck writes:

"Historically, low solar activity periods like the Dalton and Maunder Minima have been connected to cold winters in Europe. It seems very possible that the low solar activity forced areas of low pressures into a southern route or caused a negative Arctic Oscillation, AO, which in turn allowed cold air from the North Pole to flow across Europe. But can we obtain from real measurements that low solar activity really is able to do that?

I found that the mechanism is statistically significant, but it is not very simple to prove. There is no direct statistical relationship saying that low solar activity always should cause a negative Arctic Oscillation (which caused cold air to push further south than normal). But if we consider a second natural parameter, the strength and direction of the stratospheric wind in the Tropics (the Quasi-Biennial Oscillation index, QBO) I found a very interesting result: During periods of low solar activity (few or no sunspots) an easterly QBO causes a negative AO, but a westerly QBO causes a positive AO.

However, during low solar activity the easterly QBO causes a considerably stronger negative AO than the westerly QBO is able to cause a positive AO. Furthermore, easterly QBO is more common than westerly QBO during the Nordic Hemisphere winter.

The conclusion of my work is clear. If the sun goes into a new Dalton and Maunder minimum, we can therefore expect extremely cold winters in North America, Europe and Russia - which is exactly what was experienced during both the Maunder minimum (1600's) and the Dalton minimum (early 1800's)."


In essence what this research shows is that there is a link between the level of solar activity, the stratosphere, and the weather patterns that we experience, and gives more weight to the idea that solar effects may influence our weather (and hence climate) more than is currently accepted or understood.

There are Intriguing possibilities from a long range forecasting point of view. As the QBO is periodic, it's relatively straightforward to forecast for one to two years ahead. We can also usually determine where we are likely to be in the solar cycle for the following season and so the findings of this research could mean that severe winters like the one we have just experienced may be easier to forecast, months in advance.

Professor Stephen Mobbs at the National Centre for Atmospheric Science, commenting on the research told me:

"The stratosphere is very different to the part of the atmosphere we live in. There is a sharp divide at 8-15 km altitude between the troposphere where the weather occurs and the very stable and quiet stratosphere above. Our familiar weather is dominated by turbulent winds, clouds and transport of heat by convection plus phase changes of water from vapour to liquid to ice and vice versa.

By contrast, the stratosphere has smooth winds, virtually no clouds and only small (but very important) amounts of water vapour. Here, radiation dominates the transport of heat. Nevertheless, there is a growing body of evidence for some downward effect of the stratosphere on the troposphere and its weather systems. A curious phenomenon is the Quasi-Biennial Oscillation (QBO) in which the stratospheric winds change from easterlies to westerlies and vice versa, returning to their original state every 25-29 months. In spite of the name, the QBO has nothing to do with the length of the year - it is driven by atmospheric waves propagating up from the lower atmosphere. These waves are caused by mountains, land-sea heat contrasts, tropical convection and weather systems.

The QBO has proved difficult to reproduce in climate models and it is only in recent years that models such as the Met Office climate model have succeeded in doing this. There is now growing evidence that the QBO affects things like hurricane seasons, the monsoon and El Niño.

Since the heat transport in the stratosphere is dominated by radiation, it is quite plausible that solar radiation fluctuations could affect the stratospheric winds and hence the waves which drive the QBO."

You can read Dr Ahlbeck's research paper in PDF format by clicking here [214KB PDF]


or register to comment.

  • 1. At 7:50pm on 22 Mar 2010, PingoSan wrote:

    A fascinating blog here Paul, and certainly confirms the sceptic's viewpoint that solar changes have not been considered enough by the mainstream carbon-obsessed climate researchers. It's nice to be able to read a climate study without the researcher feeling compelled to mention AGW.

    This winter has seen widespread snowfall to lower latitudes across the northern hemisphere. The albedo effects of the snow cover in reflecting sunlight back out to space will be so much greater given that the strength of the sunlight is greater at these lower latitudes. This will have led to a much larger heat loss from the climate system this winter due to the solar-induced AO changes. Over several years or decades we can then see how this could lead to winters getting successively colder if the sun remains in its current slumber.

    Complain about this comment

  • 2. At 9:27pm on 22 Mar 2010, Tom wrote:

    Thank you Mr. Hudson, a very interesting item, still a new angle on solar activity having an influence on climate, the sun has been very quiescent recently, will we experience a 'maunder minimum'? One rather hopes not.

    Complain about this comment

  • 3. At 10:33pm on 22 Mar 2010, ScudLewis wrote:

    Great post Paul - very interesting. Makes a change to read something that doesn't dismiss the relationship between solar cycles and cold weather events.

    Any thoughts on the Iceland eruptions? If it sets off a bigger eruption - what say we will have more to worry about than a solar lull?

    Been reading up on the Laki eruption of 1783 - http://en.wikipedia.org/wiki/Laki

    1784 was top 10 coldest CET years on record!!

    Fingers crossed nothing bad happens - but the scientists have been quoted as saying every time Eyjafjallajokull erupts, Katla has also erupted.

    What effects could a Katla eruption cause?

    Complain about this comment

  • 4. At 00:57am on 23 Mar 2010, SheffTim wrote:

    I've also come across this explanation for why such large high pressure areas (-AO) form over the Arctic in the first place e.g. as in 2009-2010.
    It seems to link in with Dr Ahlbeck's thinking about the westerly QBO, but has more to do with oceanic influences.

    "What caused the cold outbreak was a stratospheric warming event during late Nov/early Dec. This caused the polar vortex to slow...split in two...and reverse direction...creating an anti-cyclonic (clock-wise) circulation aloft. The reversal took ~3 weeks to propagate to the surface...creating HIGH pressure over the pole....which in turn created favourable conditions for arctic outbreaks and high-latitude blocking...such as the one currently observed.
    These reversal events occur preferentially during years (such as this one) where an east wind is observed in the tropical stratosphere...the quasi-biennial oscillation - QBO...and solar activity (sunspots) is low.

    Above normal snowfall in eastern Eurasia this fall played a significant role in initiating the stratospheric warming event."
    Judah Cohen of Atmospheric and Environmental Research.

    See also:
    Improved Skill of Northern Hemisphere Winter Surface Temperature Predictions Based on Land–Atmosphere Fall Anomalies.
    Cohen & Fletcher Journal of Climate. 2006.
    It's a PDF but Google the title.

    There's also an article on it at Science Central.
    "Cohen says land features like the vast Siberian snow fields have a bigger impact on North American winters than previously thought. The increased cold and reflected heat of heavy autumn snows in Siberia affect a less well-known pattern called the Arctic Oscillation, the circulation of wind around the North Pole, which pushes high pressure and cold southward."

    The above (Above normal snowfall in eastern Eurasia) may have some linkage with this about the Indian Ocean Dipole Mode:

    " Results reveal that the peak positive phase of the dipole during autumn could suppress the following summer monsoon activity over East Asia three seasons later, in particular over the Korea-Japan sector, South China and the adjacent West Pacific region. Composite and correlation analysis suggests that the autumn positive phase of the dipole could induce heavy snow over Eastern Eurasia, north of the Korea-Japan (EENKJ) peninsula, during the following winter and spring seasons."

    Delayed Influence Of The Indian Ocean Dipole Mode On The East Asia-West Pacific Monsoon: Possible Mechanism.
    Kripalani & Chaudhari. International Journal of Climatology. 2009.

    So, a peak positive phase of the Indian Ocean Dipole could trigger heavy snows over Eastern Eurasia, north of the Korea-Japan peninsula that in turn initiate the stratospheric warming event that in turn triggers high-latitude blocking as we've seen this winter.
    This might interest people too, on a different subject.
    Positive Weather Solutions (Not the Met Office) are now predicting a very hot summer this year.
    "It now predicts that average temperatures this June, July and August will beat those of 1976, the hottest summer ever recorded with a sweltering average of 17.8C (64F)."

    Complain about this comment

  • 5. At 10:15am on 23 Mar 2010, Jarl Ahlbeck wrote:

    Due to large random (stochastic) variations between different winters, the QBO-SUN-AO relation is only one tool of many possible other tools if one wants to make long-range forecasts of forthcoming winters. The sun may also influence the troposphere by the cosmic ray-cloud connection. Please visit www.factsandarts.com, scroll down a little, and you will find more interesting reading on this fascinating subject! - Jarl Ahlbeck

    Complain about this comment

  • 6. At 11:01am on 23 Mar 2010, Boleslas_Broda wrote:

    Though one respects the scientific pursuit of knowledge, especially in so fascinating a topic, a sceptical mind distrusts what seems science by coincidence.

    There are only a mere twenty AO cycles (and one supposes a like smaller-than-statistically-very-significant number of QBO's) in the 60 years of temperature records; and, the temperature sample is from one single favourable location.

    How does the proposed 'mechanism' of low solar activity generating low pressure regions work, mechanically? Without "how" it's tempting to say the mechanism is more of a correlation than actual hypothesis.

    For all of this, an intriguing and exciting study, and one hopes to see it greatly expanded.

    It leads to huge questions about sensitivity in the stratosphere to GHG levels (both in the troposphere below and in the stratosphere itself) and their impact on this mechanism.

    Is the mechanism strengthened as effective GHG levels in the stratosphere drop (as elsewhere has been suggested is happening due to the troposphere becoming more efficient at retaining heat as GHG levels other than water vapour increase) by depletion of stratospheric water vapour, or is it weakened, or unaffected?

    Will the opposite effect happen once non-water-vapour-GHG levels in the stratosphere rise enough to overcome the depletion of water vapour levels?

    In other words, is this a positive or negative feedback to the greenhouse effect?

    And why does it pick on Europe?

    Isn't North America a similar enough land mass that the mechanism ought draw Arctic air as strongly as the Old World?

    Sincere thanks, Dr. Ahlbeck for providing so much to consider.

    Complain about this comment

  • 7. At 11:14am on 23 Mar 2010, Lazarus wrote:

    PingoSan wrote:

    "It's nice to be able to read a climate study without the researcher feeling compelled to mention AGW."

    But you had to do it in the very first post DOH!

    Complain about this comment

  • 8. At 5:06pm on 23 Mar 2010, astatine wrote:

    I can't help thinking that Paul is overly optimistic about the ramifications of Ahlbeck's article, which itself lacks some perspective in its discussion of the statistical correlation between two apparently independent meteorological
    indicators (the quasi-biennial oscillation, QBO; the sunspot number, SUN)
    and the arctic oscillation (AO).

    The article claims a P<0.05 significance, meaning that there is a less than 5% likelihood of the observed data being explained by the null hypothesis (i.e. that
    the AO is uncorrelated with the QBO and the SUN). So at first glance the analysis appears to lend some support to the alternative hypothesis that a combination of these factors correlates with the AO to a statistically significant extent.

    The first point to note is that it is entirely possible for there to be a statistically significant correlation that is of a trivial magnitude. That
    is, it would be entirely possible for these two variables to reliably correlate with the AO in the way described by the author, but for other factors *not* considered in the analysis to make a much greater contribution. It would have been helpful for the author to include the R2 value, which gives an indication of what
    proportion of the observed variance may be ascribed to the factors studied.

    In other words, the P-value reported in the study tells us that the QBO and the
    SUN are correlated with the AO in a statistically significant manner, but not whether that correlation is actually important in the grand scheme of things or just a small component of the overall variability. (NB. whilst it is clear that the SUN is independent of terrestrial weather, it is not obvious that correlation between the QBO and the AO necessarily implies that the QBO *drives* the AO; from a purely statistical point of view, in the absence of a theoretical mechanism, the reverse would be an equally valid interpretation of the data; alternatively, *both* may be driven by some other meteorological phenomenon).

    Anyway, in relation to Paul's speculation regarding the possible use of the reported correlation to predict future winter weather patterns, he should be aware that once statistical significance has been established (i.e. P<0.05) it is actually the value of R2 that generally determines the predictive abilities of that correlation. Ahlbeck presumably has this information, so it would have been nice to have found it in the paper. Paul presumably claims to understand the papers that he posts here, so it would have been nice for him to have confirmed this point with the author before speculating wildly about a new era in long-range weather prediction.

    Secondly, the author provides the F-values for inclusion of the QBO and the
    SUN coefficients into his model. These are F=7.68 for the coefficient of QBO
    and F=4.33 for the coefficient of QBO*SUN. These correspond to t-values of
    t=2.77 for the QBO coefficient and t=2.08 for the QBO*SUN coefficient. I believe
    I am correct in saying, however, that with only 60 data points, the threshold for significance at the P<0.05 level corresponds to t>2.00. This means that whilst
    the correlation with QBO is clearly statistically significant, the correlation
    with QBO*SUN only just squeaks across the line. That's not to say that the correlation is not there, but it is worth stressing that it is right on the cusp
    of credibility. I'm quite willing to believe that I'm grasping the wrong end of the statistical stick, as this is something that crops up relatively rarely in my own work. But the paper would definitely have benefitted from much more discussion on this point, either (a) to avoid people like me misunderstanding the significance test, if that's indeed what has happened; or (b) to ensure that readers are properly made aware of how tenuous the reported correlation with the SUN really is. I can't help thinking that a referee would have asked for a similar clarification in the course of peer review, so I can only surmise that this paper has not been through that process. I would be interested to know if Ahlbeck has plans to submit the paper for publication in a reputable journal anytime soon. Oddly enough, his contribution to the comments (#5) recognising that many other factors may well be important to the AO (and indeed that stochastic effects readily mask any simple correlations) is much more measured than anything posted above the line.

    Complain about this comment

  • 9. At 6:04pm on 23 Mar 2010, Hudsonfan wrote:

    Well I'm absolutely stunned that solar activity has an effect on our climate! Again, lets stop going up blind alleys and start accepting that we are in the hands of Mother Nature. Don't fight it, simply adapt and accept we cannot change it. It's whats being going on for millions of years after all!

    Complain about this comment

  • 10. At 7:05pm on 23 Mar 2010, Jarl Ahlbeck wrote:

    To Astatine: Please start by reading the paper by Karin Labitzke, link in the end of my .pdf report. She came to the almost the same conclusion. What I did was using a standard multiple regression analysis program, input freely available data from NOAA. All data can be checked from the sources I gave. The partial SUN*QBO F-value 4.33 for (1;57) degrees of freedom is statistically significant because the critical partial F-value for p=0.05 is considerably lower (have not the table or the complete printout here at home, but anybody can check the critical number from any book). The multiple correlation coefficient is not essential for the statistical significance but it is also available in the complete printout. If you want, I can mail you the datafile in ASCII-format so you can run the regression analysis yourself. I have been teaching mathematical statistics for 20 years so I really know how to do a regression analysis. And it was the computer who said that SUN*QBO is statistically significant, not me!

    Complain about this comment

  • 11. At 08:31am on 24 Mar 2010, Boleslas_Broda wrote:

    I began digging graves considerably longer than 20 years ago, and yet I still sometimes find myself losing my footing to a soft spot that escaped my notice while making a hole, which can be a danger and is always a little humbling.

    My computer has taken to telling me that I can have sparkling white teeth using a secret discovered by a mother, and my abs can be ripped in only six days, and yet I do not find these claims interesting.

    I've seen famous shopping cart statistical certainties presented showing that families who buy beer also buy diapers, and yet it remains a mystery to me if I am to interpret the need for diapers as leading to beer drinking, or consuming beer leading to a need for diapers, or some positive family feedback effect.

    Given the complexities and challenges, I'd be glad of an astatine to look for the soft spots in my work, and would prefer the educated and experienced eye of someone who disagreed with me over anything a computer tells me or I find in a table in a textbook.

    So, in this stratospheric shopping cart, how is one to interpret the cause, the effect, the feedbacks, and the direct mechanical explanation of how cause leads to effect?

    Right now, with the statistics presented, absent this explanation, it is plausible from the paper to interpret that if one set a very large hot bonfire in Turku, Finland on New Year's Eve, one could generate sunspots.

    Complain about this comment

  • 12. At 09:47am on 24 Mar 2010, John Marshall wrote:

    It seems that this new report from Dr. Ahlbeck points back to solar input into the climate models. To paraphrase Mr. Clinton, 'its the sun stupid'
    Perhaps Dr. Ahlbeck and Dr. Svensmark (The Chilling Stars) should get together.

    Complain about this comment

  • 13. At 10:17am on 24 Mar 2010, minuend wrote:

    The Sun warms the planet ?????????

    Low solar activity cools the planet ??????????

    Now these are heretical statements against the green goddess Gaia.

    We know it is people who are to blame, they have sinned against the planet and must be punished.

    The eco-faithful should close their eyes, stick their fingers in their ears and shout, "Lah, lah, lah, ........ lah", until we deal with these denialist scum.

    Complain about this comment

  • 14. At 12:17pm on 24 Mar 2010, astatine wrote:

    More detailed thoughts later, if I can find time, but just a quick illustrative example of the extent to which the results in the paper are not necessarily predictive.

    Let's ask the following simple question: would the correlation presented in the
    paper have successfully predicted this year's exceptionally negative AO?

    Taking the data for 2010 from the paper's appendix, and feeding it into Eq.2,
    we get a "predicted" AO for the most recent winter of -1.16, which is to be
    compared with the "observed" AO of -3,42. So, although the correlation was at
    least correct in suggesting a fairly strong negative AO, it was some way off
    regarding the magnitude. AO values lower than the 2010 "predicted" value of -1.16 occur about once every five years on average through the data presented in the paper, whereas the "actual" AO was the lowest in the dataset by a margin of 0.8.

    Just how badly off would the prediction for the most recent winter have been?
    Well, given the low sunspot numbers through that period, the QBO*SUN contribution
    to Eq.2 is trivially small (approximately 0.03) and so we can estimate the confidence limit of the prediction quite easily just from Ahlbeck's quoted standard error in the coefficient for QBO. Using this number (0.022) and the listed value for the QBO index (-15.000) we can estimate the standard error in the predicted AO value as being somewhere around 0.33. Since the 95% confidence limit coincides with roughly twice the standard error, we can say that Eq.2 would have predicted an AO lying between -0.50 and -2.00 with a 95% confidence. It would have predicted an AO lying between -0.17 and -2.33 with more than 99% confidence. The actual AO lies nearly *seven* standard deviations away from the predicted value!

    Perhaps I'm being a bit unfair, focussing on just one year? OK, but the claim
    does seem to be that the correlation will be especially helpful in predicting
    strikingly negative values of the AO (and hence cold winters) so let us focus
    on all eight years in the data appendix for which the actual AO fell below -1.50.
    The results are:-

    1960 Predicted AO: -0.30 Actual AO: -1.58
    1964 Predicted AO: -1.13 Actual AO: -1.91
    1966 Predicted AO: -1.30 Actual AO: -1.50
    1969 Predicted AO: -0.33 Actual AO: -2.29
    1970 Predicted AO: -0.28 Actual AO: -1.87
    1977 Predicted AO: -0.98 Actual AO: -2.62
    1986 Predicted AO: +0.24 Actual AO: -1.81
    2010 Predicted AO: -1.16 Actual AO: -3.42

    So, over these years with abnormally low AO values, the predictions based on the correlation presented in the paper average out at -0.66, while the observed AO values for those same years average out at -2.13. In only *one* of these eight years (1966) would Eq.2 have come even close to predicting the observed AO value.

    I don't especially doubt that the correlation exists, at least to within the P-values quoted in the paper, but as to whether it can ever be predictive of abnormally low AO values, I remain far from convinced. It seems to me that other factors (possibly including chaotic contributions) must obviously be dominant over the weak QBO and QBO*SUN correlations.

    Complain about this comment

  • 15. At 3:57pm on 24 Mar 2010, Jarl Ahlbeck wrote:

    Agree with ASTATINE (14). The simple regression model, although statistically significant, is far from a perfect predictor for the next winter (providing that we know QBO and SUN, both are very easy to forecast). But it may still be better than recent MetOffice AO-predictions. However, the equation (should be updated and improved every year) may work as one part of a prediction algorithm together with the complicated but rather unreliable AO-predictors used today. We may also discover other plausible correlations. If we get a new long solar minimum (say 10-100 years without sunspots) I am quite convinced that negative AO will be more rule than exception and therefore the low solar activity will cause the domination of cold winters in North America, Europe and Russia. An old phrase: "more research is needed". Please visit www.factsandarts.com for additional reading about this and other interesting subjects!

    Complain about this comment

  • 16. At 5:25pm on 24 Mar 2010, Brent Hargreaves wrote:

    2. At 9:27pm on 22 Mar 2010, Tom wrote:
    " ...will we experience a 'maunder minimum'? One rather hopes not."

    Yes, that's a very sensible thing to hope, Tom. On the other hand, a century or so of glacial weather might be a price worth paying if it shuts up the Global Warmers. We could call it the Gore Minimum.

    Complain about this comment

  • 17. At 5:41pm on 24 Mar 2010, PM wrote:

    I'm assuming this paper has not been published yet. Apologies if I am incorrect.

    As a pointer, all figures should have the correct units showing, Kelvin for temperature etc, otherwise they are meaningless.

    Complain about this comment

  • 18. At 7:33pm on 24 Mar 2010, Jarl Ahlbeck wrote:

    No, the paper is not a refereed report yet, only an internet publication originally made for www.factsandarts.com. It has to be enlarged (more references a.s.o.) in order to be acceptable for a fancy climate journal. The procedure will also take many months. The feedback I have got so far on this blog and by e-mail will be very useful, txs! Nowadays there a lot of good non-refereed scientific reports on the net, but also many bad works. Among refereed climate reports you can also find horrible junk science or as we say in Finland "nollatutkimus" (zero research). I'll work on the subject, but it will take some time because I get no funding whatsoever from any climate research projects or from the industry. Actually I am a little proud of being a 100% free scientist.

    Complain about this comment

  • 19. At 12:32pm on 25 Mar 2010, Jarl Ahlbeck wrote:

    The reason why the model despite large random noise is statistically significant is its ability to correctly predict that a negative AO (and thus a cold winter) will be the result if the QBO is negative AND the solar activity is low. Such winters have been for example 1955,1963,1966,1977,1987,1994,1996,2006 and 2010. The analysis shows that because the QBO is more often negative than positive during the winters, a future long period of low solar activity will almost certainly cause the domination of cold winters.

    Complain about this comment

  • 20. At 2:15pm on 25 Mar 2010, astatine wrote:

    15/18/19.Jarl Ahlbeck,

    I'm glad to hear that you intend to submit your work for peer-reviewed publication, which in my view is the most appropriate way to begin disseminating scientific research. The web is a great tool for sharing the results of our work with the wider public, but I do feel that it is better to do so *after* those results have gone through the most basic quality control mechanism (i.e. peer review).

    And I agree that peer review does not guarantee that "junk science" is never published, but it does improve the "signal-to-noise" ratio tremendously. Plus, once published in a relevant peer-reviewed journal there's a good chance that other experts in the field will read/discuss/refute/extend the work; with the best will in the world, publishing on a blog is never going to lead to the same level of critical analysis. I don't view your contribution here as "junk science", by the way, at least as regards the data analysis itself. But there are real issues relating to the interpretation and presentation that lead to two potentially serious problems:-

    1. People who understand the rudiments of statistics get frustrated, because although you provide information about the statistical significance of your correlation, you don't draw attention to (or quantify) the lack of predictive power attached to it. You point (in your most recent comment) to a statistically significant trend towards negative AO (I assume you imply "more negative than the mean") under conditions of negative QBO and low SUN. But that only tells us that a protracted period of low solar activity should correlate with a series of winters for which the *mean* AO is more negative than usual. Nothing at all about whether that mean is *far* below the long-term average or a *little* below the long-term average. And certainly the variation around that mean is likely to ensure that we still get a mix of warm and cold winters. In all likelihood, it means that we would see just slightly more colder ones than warmer ones. Hardly "the domination of cold winters" you mention. All these issues could be fixed for an expert audience by providing more details of the statistical analysis (e.g. R2 values, etc).

    2. People that don't understand the rudiments of statistics get confused, and start writing silly things like "the findings of this research could mean that severe winters like the one we have just experienced may be easier to forecast" and "its the sun stupid". For that kind of inexpert (or inattentive) audience, it is important that your conclusions should actively draw attention to the model's limitations. You've accepted yourself that the predictive power of the correlation is limited to statements about average behaviour over many years, so don't you find it a bit worrying that it is being misinterpreted as suggesting a possible route to the forecasting of individual extreme events? I think the root problem is a lack of pre-emptive expectation-management in the paper itself, compounded by an unduly gung-ho final paragraph in your layman's summary as reproduced by Paul.

    In my view, there are things that could be changed in the paper that would help avoid either of these problems arising in the first place. If this manuscript ever crossed my desk as a reviewer (it wouldn't; I'm qualified in neither meteorology nor climatology) then I'd certainly flag these to the editor as essential changes.

    I sympathise about the time and effort that goes into the preparation of papers for peer-reviewed publication, and the even greater time and effort that often goes into accommodating the demands of referees. But as responsible scientists, that's just part of our job, isn't it?

    Complain about this comment

  • 21. At 4:17pm on 25 Mar 2010, astatine wrote:

    OK, so in a spirit of constructive criticism, here are a few comments on how the paper should be improved for publication. These are the points that I would make if the manuscript were passed to me for peer review. Apart from being useful to the author (I hope) it may be interesting to the general readership of this blog to see an example of how peer review really looks (since it has come in for a lot of unwarranted criticism lately). The following would not be an atypical referee report from me (in my own field, of course) in terms of length, depth and tone.

    I completey accept that it is a trifle unfair of me to do this, since the author has confirmed that he was *not* expecting to submit the paper in its present form for peer review. It is entirely possible that the issues I raise would have been fixed by him anyway before that point. But I think it might be genuinely useful to highlight the differences between what is essentially a journalistic article (i.e. the paper as it stands) and a peer-reviewed research article (i.e. the paper as it would have to be revised to get published).

    This is clearly not a *normal* way to referee a paper, and I do keenly realise that it is also uninvited. So I hope the author will accept that I mean this both respectfully and constructively. I just think it may shed some light on why some of us insist that the media should base its reporting of science primarily (possibly exclusively) on the peer-reviewed literature. It occurs to me that most non-scientists will have no experience of the level of robust constructive criticism that is usually involved.

    --- Report Begins ---

    This paper reports on a statistical analysis of the winter-period arctic oscillation (AO), concluding that a statistically significant correlation exists with the quasi-biennial oscillation (QBO) and the sunspot number (SUN). Specifically, it is claimed that the significant correlation links AO with QBO and with the product QBO*SUN. So far as I can tell, the actual analysis looks sound. I do, however, have several important criticisms regarding the presentation, interpretation and conclusions.

    1. The paper appears to lack sufficient reference to the existing literature on this subject. (Ahlbeck notes the same thing himself in his comment at 18).

    2. The inclusion of the Turku temperature record is distracting and unnecessary. The main point of the paper is to demonstrate a correlation of QBO and SUN with AO. The fact that AO correlates well with temperature in northern latitudes is, I believe, already well-established. Thus, correlating the Turku data with either AO, or with QBO and SUN, adds nothing that is not already known. In fact, it risks confusion, since it focusses attention on a local phenomenon, when the paper primarily addresses a phenomenon that affects a large region (i.e. AO). The author therefore should remove specific comparison with the Turku temperature data. The fact that AO correlates with temperature in northern latitudes can be conveyed more succintly and appropriately by referencing the relevant literature.

    3. Figures 3, 4 and 5 show Gaussian and Log-Gaussian fits to the AO, QBO and SUN variables. It is not made clear whether these fits are intended only as illustrative "guides-to-the-eye" or whether they feature somehow in the subsequent data analysis. The fits for QBO are plainly very poor, so this point does need to be clarified.

    4. The author refers to behaviour of the AO being "very typical for 1-lag and 2-lag random walk mechanisms" but provides neither a reference nor an explanation. The point is very far from obvious to me. Furthermore, in the conclusions section, the author claims that the correlation he finds "explains why the AO seems to behave according to a random walk mechanism". In fact, he has not even satisfactorily *demonstrated* (merely asserted) that the AO exhibits random walk characteristics, let alone *explained* it. These appear to be the only two references to "random walk" in the manuscript, so again the importance of this point (if indeed it *is* important) is not made clear.

    5. The statistical analysis leads to a best-fit model involving the overall mean, the coefficient of QBO, and the coefficient of QBO*SUN. The F-statistics for the latter two coefficients seem to confirm statistical significance. I do, however, have some questions and comments:-

    5.a. Should there not be a quoted standard deviation for the mean? I may be misunderstanding this point, but naively I would expect there should be. I'm
    quite happy to have my naivety corrected on this point.

    5.b. The statistical significance of the QBO coefficient is probably quite robust, but that of the QBO*SUN coefficient looks to be only slightly beyond the 95% confidence limit. It would be worth stressing that this latter correlation is at the edge of what can safely be deduced from this data.

    5.c. The quoted confidence level for the model (i.e. Eq.2) establishes that the null hypothesis corresponding to "no dependence on QBO or QBO*SUN" may be rejected.
    I would be interested to see the results of an F-test conducted to determine whether the model including QBO and QBO*SUN gives a better fit that one including only QBO. That would seem to be necessary to confirm that SUN needs to be included at all.

    6. It is entirely possible for a correlation to be statistically significant, but to have very limited predictive power, due to variations correlated with variables not considered in the model (or indeed random variations). For that reason, it would be very useful for the author to include details of the R2 value associated with his fit. This would allow the reader to assess what fraction of the observed variability in AO is captured by the reported correlation with QBO and QBO*SUN. If R2 is as low as I suspect, then the author will primarily have demonstrated the *lack* of a predictive relationship between these variables and AO.

    7. Figure 6, showing selected regression lines from the model, has the potential to be highly misleading, for two reasons:-

    7.a. The plot shows the regression lines for QBO=0 and for two extreme cases, namely QBO=-20 and QBO=+11. These latter two correspond literally to the two extreme values recorded for this variable over the entire period covered by the dataset. For the vast majority of years, the QBO value will be well within these extremes, so choosing to plot these highly atypical values risks exaggerating the effect of SUN on the AO. It might be better to plot for QBO values that correspond to two standard deviations from the mean QBO value, which arguably would be a much better representation of the normal variability in this index.

    7.b. The plot should certainly include 95% confidence limits for all of the regression lines. These will be very wide, and will show clearly that the 95% populations for different QBO values overlap considerably over the majority of the SUN range. At present, the diagram makes it look like different QBO values would be predictive of abnormally high or low AO values over quite a wide range of SUN. In fact, inclusion of 95% confidence limits would clarify that the populations only fail to overlap at extremely low values of SUN.

    8. In the conclusion, the author states that his analysis "shows that the influence of solar activity together with stratospheric mechanisms acting on the AO is statistically significant". This statement is factually incorrect. In fact, the author has shown (although see 5.a, 5.b and 5.c above) only that the *correlation* between these variables is statistically significant. It is *possible* that QBO and SUN influence AO, but there are alternatives. Let us agree that QBO and AO do not influence SUN, but we could still find that: (a) QBO and AO are both influenced by some unconsidered variable, espcially during times of low SUN; (b) AO influences QBO, especially during times of low SUN; or (c) QBO and QBO*SUN correlate with AO over the period investigated through coincidence, with no causative link whatsoever.
    The statement as it stand should not only be removed, but a carefully worded statement outlining also the alternative possibilities should be inserted, giving equal weight to each.

    9. In view of the above, the author's comment that an extended period of low solar activity could signal a period of very cold northern winters is a highly speculative statement, based upon just one possible interpretation of his results (see 8 above). This should be removed, or at least balanced with a clear caveat that contradictory conclusions might equally be drawn. Arguably, the immediate relevance of this point is also highly debatable, in the absence of any clear indication that either a Dalton or a Maunder minimum may be on the horizon.

    In summary, this paper includes some interesting work, but details of the presentation are omitted that would allow the reader to properly evaluate its importance. As it stands, the interpretation carries the strong implication that the reported correlation may have predictive qualities, and the omitted details serve to obscure the fact that this is not the case. If the points raised above are properly incorporated into a revised manuscript, then the work might be publishable. It is debatable at that stage whether the remaining conclusions would be of sufficient interest for publication in a general journal, but publication in a highly specialised journal would perhaps be warranted.

    Recommendation: Reconsider after Major Revision

    --- Report Ends ---

    At this point, the editor would pass the report (and usually an additional report from another referee) to the author for comment. The author can then make changes in the light of the referee comments, or can rebut them in a detailed reply to the editor. If the editor can make a decision at that stage, then the paper is either accepted or rejected. If there are clearly differences of opinion between author and referee, then the author's rebuttal will be passed to the referees, along with a revised version of the manuscript, and a second round of refereeing begins. Two or three rounds of refereeing is pretty normal. Usually, it is clear after a few rounds that the paper has either been adapted to the reasonable satisfaction of the referees, or that there are irreconcilable differences of opinion. In the latter case, the editor may either take a decision based on their own experience and knowledge, or they may consult a third referee before doing so.

    For the very highest impact journals (Science, Nature, etc) editors will often reject (but not accept) papers *before* formal refereeing takes place, on the grounds of insufficient general interest (i.e. better suited to a more specialised journal). This happens to all of us, and is not evidence of persecution. It has happened to me more times than I care to remember. I do not feel persecuted.

    This whole process takes time. Certainly weeks; often months. Papers go through multiple revisions. Sometimes they end up published in lesser journals than the authors had hoped for. All of this can be demoralising, and it is certainly a lot of effort. But it is what scientists do. All else is journalism.

    Complain about this comment

  • 22. At 09:51am on 26 Mar 2010, Jarl Ahlbeck wrote:

    Txs "astatine" for the good comments! I will add a little:
    - The Karin Labitzke (pre-refereed) report 2005 + update showed a very clear influence of the solar activity together with the QBO on NH stratospheric warming and hence on the AO-index. My regression equation confirmed her finding. Furthermore, my pattern of the influence of solar activity was exactly the same as for Karin Labitzke: increasing AO with solar activity for QBO-east, and decreasing for QBO-west. However, Labitzke obtained a stronger influence of solar activity for QBO-west, but I found a stronger influence for QBO-east. But we probably did not use exactly the same database, my data may be more reliable because they are directly from the updated high-quality NOAA website. Her division of data into two datasets, one for QBO-west, and one for QBO-east did not take into account the probability distribution: QBO-east is more common. Therefore I presented the probability distributions in my report. I think that the dataset should not be divided as the problem should be solved by mathematical modeling directly.
    - Usually 95% confidence (have tested many times by Monte Carlo simulation) is enough for separating "real" signals from noise. I my case, adding the solar activity improved the model significantly compared to a run with QBO only. Therefore I think we now have strong evidence (both Labitzke and my run) to believe that solar activity significantly influences AO and hence the winter climate.
    - Random walk (autocorrelated) signals are typical in process control applications when stochastic noise is disturbed by two or more harmonic oscillations. The autocorrelation function for AO shows both 1-lag and 2-lag correlations. But explaining this fact by adding more references and autocorrelation functions was not necessary in this stage.
    - The IPCC use 90% probability as proof of this and that. According to my experience, 90% means a risk for incorrectly explaining random variations by physical variables. This is due to the fact that the devations are not Gaussian which in turn means that the 90% probability is not a real probability either, it is just a critera.
    - But if the influence of solar activity on winter climate can be explained by more than 95% probability, I think we have reason to believe that the influence of solar activity is real.
    - Having said that, I am aware of the need for more research on the subject.
    - I will write no more comments.

    Complain about this comment

  • 23. At 5:08pm on 26 Mar 2010, bandythebane wrote:

    Congratulations Paul.

    Not only was your article good in itself, but it has encouraged a much more informative set of comments - the last two from Astatine and Jarl Ahlbeck in particular.

    Until about a week ago I didn't even know what a "random walk" was, but thanks to Andrew Montford who explains it very well in discussing "hockey sticks" I am now informed and can (almost) read these posts with understanding.

    Again well done. If you keep going on like this, even McIntyre will soon have to go back to brush up his statistics!

    Complain about this comment

  • 24. At 10:50pm on 03 Apr 2010, bandythebane wrote:

    I didn't realise your post was that good Paul. None of the WWF types can think of anything more to say..

    Complain about this comment

  • 25. At 9:23pm on 04 Apr 2010, Thomas wrote:

    " I am not an educated man just an observer.Climate change due to mans activities - some. Check out Prof Brian Cox that the orbits of the planets in the solar system can change over time as can the angle the earth spins relative to the sun. Loads of earthquakes recently? More to come as gravity pulls the earth and the cracks start to show. Climate change is a symptom we all know that - is the angle of the earth to the Sun changing. Is gravity heating up the inside of the earth as the earths orbit ecomes more elliptical. I honestly dont know. But I do wonder

    Complain about this comment

View these comments in RSS


Sign in

BBC navigation

BBC © 2014 The BBC is not responsible for the content of external sites. Read more.

This page is best viewed in an up-to-date web browser with style sheets (CSS) enabled. While you will be able to view the content of this page in your current browser, you will not be able to get the full visual experience. Please consider upgrading your browser software or enabling style sheets (CSS) if you are able to do so.