Anthony Watts, the story behind the story

Anthony Watts has posted up the “backstory” behind his correction of the dodgy stats on US climate change. This explains the mysterious message last week when he closed down the site, pending a  major event.

I’m a bit burnt out, so this is a just a few notes to quench some speculations about Steve McIntyre’s role and to help everyone understand what this week has been like.

This is the major event, for a summary and explanation of significance check out Jo Nova.

In a nutshell: PRESS RELEASE – U.S. Temperature trends show a spurious doubling due to NOAA station siting problems and post measurement adjustments.

This entry was posted in Uncategorized. Bookmark the permalink.

793 Responses to Anthony Watts, the story behind the story

  1. SteveC

    I’m always so impressed by cohenite’s well reasoned eloquence in a scientific debate.

  2. blogstrop

    I’m always impressed by SteveC’s imperviousness to any counter argument by anyone at all. And the fact that he bounces back day after day, unwanted, unloved, and dedicated to filling these threads with frothy gainsay. If he’s not doing it for money or political advancement I’d be very surprised.

  3. SteveC

    I’ve done the analysis for HadCRUT and got very similar results to you, actually with a much better r-squared. Where did you get the monthly CET anomaly? Did you have to calculate it yourself? I can find the monthly CET monthly temperatures at, but of course that gives quite a wide variation. I can calculate the anomaly myslef but sounds like a lot of work if it’s already online somewhere.

  4. SteveC

    blogstrop I’m looking forward to a decent counter argument from you any day now.

  5. Infidel Tiger

    Give it up SteveC. Caring about climate change is like caring about your Cabbage Patch Doll or Hula Hoop. It’s a long forgotten fad and we’ve all moved on to more important issues.

  6. SteveC

    How dare you denigrate the hula hoop!

  7. Infidel Tiger

    ‘Tis no Rubik’s Cube!

  8. Bruce

    SteveC – I’m not sure where I got the CET values, but I recall they were annual average temperatures and they were as a graphic table not csv. But it was the full set. I OCR’ed them and cross checked. Took quite a while. At the time I couldn’t find the official dataset at Met Office (this was about 3 years ago). Later I did, but I haven’t gone back and cross validated.

    OK, looking at them the Monthly HadCET mean.txt one looks like the one I used. I used the ‘year’ column. I’ll pull it into my spreadsheet and let you know shortly if it checksums (have to go over to laptop).

  9. Bruce

    Yep, that’s the one. Checksums without error, though the dataset version I used only went up to 2005.

    To get anomaly values its just the year values minus the 1961-1990 average, which was the standard when I did the analysis.

  10. blogstrop

    SteveC, you can get all the counter arguments you need without me adding to them, but be honest and say that you’re a political operative and the propaganda objective trumps the science for you.
    Whenever you or SfB run into a problem coming from actual scientists you either smokescreen it or like Anna Rose go the man.
    The whole CO2 reduction regime is a scam, and you know it.

  11. cohenite

    Why would you want r^2 for HadCrut; what are you comparing the proportion of variance to; steveC’s IQ?

  12. Bruce

    Cohenite – Give SteveC a break. He’s replicating a bit of graphing I did which shows solar cycle length is linked with HadCRUT. The r^2 is for that. I’ll remind you that Dr Muller said “It was purely greenhouse gases.” If a large part is solar dynamo related then Dr Muller is in error, which in part what this thread is about.

  13. SteveC

    Bruce, you might be interested in the charts I created based on your chart. I plotted previous solar cycle length against avg temperature anomaly for HadCRUT Global Temp and HadCET for 1854 to now and 1755 to now (for HadCET only).
    I’ve include the regression line and r-squared in each chart.
    I did hadcet by month, as that matches the solar cycles better (as the cycle lengths are by month). I also included hadcet by year to compare with yours, those results are comparable to yours, though not exactly the same, I guess it depends which years you included with which cycle.

    A few things stand out to me.
    There appears to be a good correlation with Global Hadcrut and CET (after 1855) and the length of the previous Solar Cycle. However this breaks down as we get more recent. In particular you will see in the monthly hadcrut that cycles 21, 22 and 23 are all well above the regression line. Once you add cycle 24, the goodness of fit breaks down (r-squared drops from .48 to .10).

    The hadCET prior to 1855 doesn’t work so well, as you suggested probably should be cleaned up for known cooling events (volcanoes etc).

    One option would be to adjust the anomalies to account for the known warming trend over the last 300 years, which would possibly remove the bias towards older cycles being below estimate and later cycles above estimate.

    However my main conclusion would be that solar cycle length does predict long term temperature well. Except that since about solar cycle 22 (1986) that prediction is no longer working. So whilst Solar cycle appears to be a factor, in the last 30 or so years, something else is swamping the expected effect of the solar cycle and causing warming not predicted by the solar cycle length.

    Cycle 24 (the current cycle) is particularly problematic. The length of Cycle 23 was 12.6 years. based on the regression line for hadcet annual up to 2008, we woould expect an average anomaly for the current cycle of -0.55. What have seen so far (3.5 years into the cycle) is +0.4. So to get back to the expected -0.55 by the end of the current we would need to see an average drop of about -1.0 from now on (depending on the length of the cycle).
    That seems pretty unlikely to me.

  14. JamesK

    I salute your work SteveC.

  15. SteveC

    CoheniteWhat is the determining variable?
    I assume you mean what is the independent variable?
    Length of Solar Cycle. This was based on the Butler and Johnston paper referred to by Bruce above.

  16. SteveC

    cohenite, the regression equation is on the charts. cycleavg means the average temperature anomaly for the period of the solar cycle. prevcyclelength is the length (in years) of the previous solar cycle. The theory (of Butler and Johnston) was that the temperature is influnced by the length of the previos solar cycle

  17. Bruce

    SteveC – Looks good to me. I see you used the wiki SCL values, whereas I used the minimum to minimum lengths from the DMI dataset.

    Here is the thought for you. If the HadCRUT3v monthly 1843-2008 slope is correct then from 1900 to 2000 you have a +0.44 C component due to this relationship. (This is the difference in trendline level at your SC14 and SC23 points). Here’s what AR4 says:

    The updated 100-year trend (1906–2005) of 0.74°C ± 0.18°C is larger than the 100-year warming trend at the time of the TAR (1901–2000) of 0.6°C ± 0.2°C due to additional warm years.

    So you can see this pSCL-temperature relationship can potentially explain more than half of the centuric temperature rise.

    As you point out the solar cycles before 1850 knock the stuffing out of this relationship for CET (as well as making life difficult for Napoleon in Russia). If you want to try correcting for the volcanic eruptions like Laki and Grimsvotn here is what I think is the official curve. I haven’t bothered as I’m not enough of an enthusiast. I’d be nice to line up a chronological list of Icelandic volcanic eruptions against the CET but I haven’t seen a good list.

    Anyway – nice work.

  18. Bruce

    Except that since about solar cycle 22 (1986) that prediction is no longer working.

    This is where some multivariate analysis should really be brought in. As I said earlier, I mapped the pSCL component from B&J 1996 with the 60 year cycle onto CET and found the residual fits pretty well with a 2XCO2 of about 0.7 C as found by Lindzen & Choi 2011.

    So in my view pCO2 does have an effect and so does the 60(ish) year cycle. Both of those would have added a significant component to SC23 and to the upcoming SC24 number. So I think it is premature to say the prediction is not working. No single variable is going to be the all in all.

    Ideally this sort of analysis should be being done by climate scientists not chemists like me. The challenge for them though if it is right, how do they keep their budgets?

  19. Gab

    Anyway – nice work.

    Yes, including the attending comment. Such a marked contrast to how you normally write, steveC.

  20. SteveC

    Hi Bruce, I can’t read the left axis on that wavelet chart. What is it measuring?
    I’ll adjust the anomalies for the +.44 hundred year trend. I would expect that to improve the goodness of fit. Yes I used the Wiki cycle lengths because I had already typed them in! I will do it again with the “official” list, but they are pretty close so would be surprised if there’s much difference.

  21. Bruce

    SteveC – Sorry, I thought you’d looked at some of the other 60 year cycle links upthread. The wavelet chart is a power spectrum determination which shows the various wavelengths of the Fourier transform if I recall correctly. Of HadCRUTv3. Not sure what the y axis is, by the x axis shows the wavelengths. See also here. This one shows HadCRUT minus the quadratic trend equation. See the cycle? About 0.14 C amplitude, ie 0.28 C trough to peak. The whole post is here.

    Knight et al 2005 (also with Dr Mann) shows a similar power spectrum for the thermohaline circulation (Fig 2) and a graph of the AMO with the 60 year cycle pretty obvious (Fig 1). The THC doesn’t have a 60ish year peak, but has apparent resonance peaks at ~30 and ~90 years. The paper shows a link with AMO but skirts around HadCRUT.

    You see the 60-ish year cycle strongly in the AMO, HadCRUT and less strongly in the PDO and ENSO. More links below due to limit of 3.

    I don’t really know which is the official solar cycle length list, I just assumed that the Danish Meteorology Institute would trump Wikipedia. But Wiki surprises sometimes.

  22. Bruce

    The 60ish year cycle in HadCRUT, PDO and ENSO. There are other forms of representing this data, these I chose as they highlight the longer wavelength cycle (or pseudocycle).

    You can decide yourself whether this signal is real or not, but if it is it represents about 0.25 C of the temperature rise between 1900 and 2000 and also 1970-2000 because of where the troughs and peaks happen to be (see the HadCRUT curve fit graph).

  23. Louis Hissink


    Ian Levy, a professional geologist, did a statistical analysis of the Australian temperature data and noted a 33 and 66 year cyclicity in the data. It was confirmed by variogram analysis, and hence is real. However just what it means physically remains moot.

  24. cohenite

    Ok, Steve, that is a reasonable analysis; I did not look before because I had assumed that you were merely trolling.

    You have accounted for some of the lag in the solar effect on temperature as represented by the previous solar cycle as the predictor for temperature in the next cycle; however that ignores a more basic lag which can be shown by integrating the input, in this case the solar intensity, which is distinct from the solar cycle.

    By integrating that over a period the average of the input becames the base for a temperature response; in effect the solar input is either accumulating above the average for the period, in which case the temperature response is increasing with the rate of increase determined by whether the solar input is rising or descending above the average, with the reverse happening when the solar input is below the average and the solar input is not accumulating.

    This thesis is explored in a recent paper by Stockwell, short version, long version.

    This theory has a remarkable correlation between solar and temperature and accounts for the seeming lack of correlation over the modern period.

    In Stockwell’s thesis the role of GHG would be limited to the non-accumulating mid to upper atmosphere, where, given the lack of robust warming and a THS, such effect is minimal, if at all.

  25. SteveC

    Thanks Bruce, I didn’t see the 60 year cycle links upthread, there’s 700 odd posts in this thread you know! Re the wiki solar cycles, the wiki actually links to the “official” danish list, but doesn’t match it. I posted a question that effect on the wiki talk page.

  26. SteveC

    Cohenite, thanks for the stockwell link. Are you aware if that’s been published anywhere that would generate published responses. I usually look for citations on google scholar but i can’t see any.

  27. Gab

    It’s just amazing stevec how well you write now on these matters. Why didn’t you write like that before?

  28. cohenite

    Steve; there is interest mounting about the Stockwell paper in the US; he will be giving a talk to it in October.

    He is content for it to sit at vixra for the time being.

  29. SteveC: did I already link to this? I can’t remember. Anyway, in case you haven’t noticed, a Skeptical Science post on the issue of solar cycle lengths and temperature is here (and the comments following are well worthwhile too.)

    A list of papers relevant on the topic was at AGW Observer in 2010.

    I am somewhat amused that you appear to replicate the point made at Skeptical Science by your own graphing, and readers here like James K and cohenite go “well done”.

    As for Stockwell’s and cohenite’s idea: I’m afraid it all reeks of desperation borne of the philosophy I pointed out above:

    ‘the end point of any question about global warming is always going to be this – “maybe it’s there, maybe it’s not, but in any event it will never be CO2?.’

    It’s too dense for me to unravel, but I would bet money that it is an example of what some blogs like to call “mathturbation”.

  30. cohenite: interest from who? The Heartland Institute?

  31. cohenite

    You’re such a dickhead stevefromfuckingbrisbane.

    The value of Steve’s graphing is that it puts the lie, as Bruce notes, to Muller’s claim of NO solar imput into temperature.

    I’m sorry you don’t understand Stockwell’s thesis, which is both elegant and physically real; I did trial run it by explaining it to my dog before presenting it here and she understood; next time I’ll write it to the level of a cockroach so you understand.

  32. Bruce of Newcastle

    SteveFB – I have not read the SkS post, as given their philosophical position I know what they will probably say. You pretty much confirmed this with your comment about Solheim et al 2012 upthread. Perhaps I am being churlish, but when they regularly delete sensible comments and after they treated Prof Pielke so crassly I prefer not to give them clicks.

    The discussion I was engaged with SteveC upon was about the actual climate data unfiltered through the brains of people of either side of the argument.

    Relevant to this discussion Dr Muller said:

    In the end, a really good match came to the greenhouse gases. That came as a surprise to me, because I’d really thought that the ocean currents and the variability of the sun would dominate, but it wasn’t the case. It was purely greenhouse gases.

    My point is on this empirical data something like 0.44 C of the temperature rise in HadCRUT in the last century can be explained by the solar dynamo through a mechanism yet to be fully explored (for which a hypothesis exists) and another 0.25 C due to an oceanic cyclical phenomenon.

    This shows Dr Muller is in error, or at very least needs to transparently and urgently address these two variables, which are not included in the climate models.

  33. Bruce

    SteveC – One thing you might like to try quickly is to graph the HadCRUT monthlies minus the quadratic regression fit. I tried briefly in Excel yesterday but couldn’t see a way to do it properly. You’ve got access to a stats package from the look of it, it might be easier for you.

  34. SteveC

    Thanks for the AGW observer link sfb, more papers there than I will ever have time to read!!

  35. steve from brisbane

    You are welcome, SteveC.

    Once again, one can see that solar influences, even ones where the mechanism is not well understood, have been really, really extensively considered by scientists in the context of climate.

    I think the AGW skeptic blogs such as WUWT, Jonova etc give the casual reader the impression that these alternatives have not been properly considered.

  36. JamesK

    ‘weally, weally extensively considered’, liar?


  37. steve from brisbane

    I seem to have upset cohenite, who has not clarified who his buddy Stockwell is talking to in the US about his new “thesis.”

  38. steve from brisbane

    JamesK is suffering an outbreak of Liar! Disease again. Maybe it’s a neurological condition.

  39. cohenite

    Really sffb, what do you care who Stockwell is talking to; you don’t understand his paper, yet you dismiss it and rabbit on about Heartland.

    Anyway, AGW is bullshit, solar and water determine the climate and CO2 grows plants and occupies the space between the ears of AGW believers.

    For those who are interested Nir Shaviv has a 3 parter on the solar influence on climate beginning here.

    Shaviv recently published this paper on the respective contributions of GHGs and solar to climate and found that while the solar contribution was much larger than AGW theory predicted AGW was the larger contributor.

    The point here is that firstly Muller is wrong and 2ndly the science is far from settled.

  40. Really, cohenite, why the big secret about who he’s talking to?

    The Shaviv abstract concludes:

    We also show that a non-thermal solar component is necessarily present, indicating that the total solar contribution to the 20th century global warming, of ?Tsolar = 0.27 ± 0.07 °C, is much larger than can be expected from variation in the total solar irradiance alone. However, we also find that the largest contribution to the 20th century warming comes from anthropogenic sources, with ?Tman = 0.42 ± 0.11 °C.

    Curious how that sits with his earlier post from 2006 which you refer us to which says:

    there is no real direct evidence which can be used to incriminate anthropogenic greenhouse gases as the being the main factor responsible for the observed global warming. The reason these gases were blamed are primarily because (1) we expect them to warm and indeed the global temperature increased, and (2) there is no other mechanism which can explain the warming.

    Although this reasoning seems logical, it turns out that (1) We don’t even know the sign of the anthropogenic climate driving (because of the unknown indirect aerosol effects), and (2) There is an alternative mechanism which can explain a large part of the warming.

    It also doesn’t sit well with your comment:

    AGW is bullshit, solar and water determine the climate

  41. Bruce of Newcastle

    largest contribution to the 20th century warming comes from anthropogenic sources, with ?Tman = 0.42 ± 0.11 °C.

    (my emphasis)

    SFB – Same problem of the omitted variable. Remove the 0.25 C derived from oceanic cyclic sources between the dates of 1900-2000 and that comes down to 0.17 C, which is consistent with a 2XCO2 of 0.6 C

    0.17 x log 2/(log 390 – log 315) = 0.55 C

    Pretty close to Lindzen & Choi’s median 0.7 C range 0.5-1.3 C (99% conf), Spencer & Braswell 2010 of 0.6 C and that which I derive from Gleckler et al 2012 of 0.4 C, which is an underestimate because of the delayed response of SST’s.

    So Steve you have just accidentally confirmed the low sensitivity hypothesis. Numbers for 2XCO2 seems to be converging to around 0.6 C per doubling. Do the sums on theoretical hydrocarbon availability. Cannot possibly be dangerous unless we start importing hydrocarbons from Titan, and probably not even then.

  42. Bruce

    Actually have to lower that 2XCO2, I put in the CO2 numbers for the Gleckler et al timeframe (1960-2010).

    CO2 in 1900 was about 286 ppmV by extrapolation of Mauna Loa
    CO2 in 2000 was 369 ppmV

    So calc is 0.17 x log 2/(log 369 – log 286) = 0.46 C

    Mistake from me, sorry, but it is clearly nowhere near the AR4 estimate of 2.0-4.5 C.

Comments are closed.