Saturday, May 28, 2016

Review of recent global temp with ENSO cpmparison

I have been maintaining a plot I did in February, comparing the temperatures month by month of various indices with those of 1997/98. I thought I should post the plot again here for easier access. There is some sign of recession from the peak in April, and this will probably be reinforced in May, although global SST still seems warm. In fact, I had to modify the plot somewhat, as in April for the first time, some indices (RSS and UAH, troposphere) dropped below their 1998 levels. This was because of an unusual TLT spike in 1998, where April stood out as hottest month. I have made the bars somewhat transparent.

So here is the plot. You can use the arrows at the top to cycle through the various datasets:




Below the jump I'll show the NOAA plot, to show the April result (1.103°C, down from 1.23). I'll also mention Arctic Ice and news on my battles with HTTPS.

Here is the NOAA plot, which is fairly typical of recent land/ocean data:



Arctic Sea Ice continues its rapid melting. As usual, full details are at Neven's.

I've been making progress with HTTPS accomodation. It's a frustrating business. As I said earlier, by scripts etc are held in an Amazon bucket, which means an actual change to the address I use (not just adding a s to http). But then I battled with another annoyance - Firefox wouldn't even let me into pages to test, and seemed to say that the certificate from Blogger was wrong. And Google is the big booster of https use! It turned out that the problem was that I regularly write the URL as www.moyhu... out of habit (I think it was originally required) and that has always worked. But it turns out that the certificate has no www. and that is a problem. So now I have to change all that.

However, the good thing is that it has forced me to create a database of all the files I use and their location, both in local files and on web. My idea of removing post content from the system managed by blogger editor, and replacing with a JS stub, didn't work. However, I used a marvellous old DOS-like system, NirSoft, to automate a lot of the editor handling. Then I can rationalise, which will help. I think everything is almost working, and I'm using a lot of https links internally. Please let me know of any problems. I may switch to https redirection quite soon.


4 comments:

  1. Nick,

    From an Excel Spreadsheet (converted to comma delimited as shown) ...

    Group,1998 - (1997+1999)/2,Average,2010 - (2009+2011)/2,Average
    UAH TTT v6.0b5,0.48,,0.37,-
    UAH TLT v6.0b5,0.54,,0.30,-
    UAH TLT v5.6,0.52,,0.25,-
    RSS TTT v3.3,0.57,,0.33,-
    RSS TLT v3.3,0.49,,0.32,-
    RSS TTT v4.0,0.56,0.52,0.34,0.32
    NOAA,0.17,,0.10,-
    BEST2,0.18,,0.11,-
    HadCRUT,0.21,,0.10,-
    GISS,0.20,,0.10,-
    BEST1,0.21,,0.10,-
    C&W,0.22,0.20,0.11,0.10
    Difference,,0.33,,0.21

    I tool the target year average and subtracted out the average of the previous and following years.

    Satellite average = 0.52 (1998) and 0.32 (2010)
    SAT average = 0.20 (1998) and 0.10 (2010)
    Difference = 0.52 - 0.20 = 0.33 (1998, rounds up by 0.01) and 0.32 - 0.10 = 0.21 (2010, rounds down by 0.01)

    This doesn't tell you anything about the long term trends, it does tell you about the relative amplitude and relative phase (if you fill out the monthly time series).

    I think there is enough information to make a reasonable forecast on the relative differences to expect between the satellite and SAT for 2016.

    Anyways, I think I can see how the satellite would differ from the SAT in your plots shown above.

    ReplyDelete
    Replies
    1. Note to self: UAH TTT v6.0b5 above should be UAH TTP v6.0b5. UAH TTP (temperature tropopause) ~ RSS TTS. It should NOT be included with the TLT and TTT time series. My bad. Sorry for my error.

      Delete
  2. The qualitative comparison of pairs of El Nino's is a bit half-baked, wouldn't you say?

    Just saying that because there is so much information buried in the complete ENSO time-series. Say we take a section of an ENSO index from 1990 to 2013 and train with that interval to the known long-period Chandler wobble periods of 6.5 years, 14 years and 18.6 years, and modulating that with a metastable biennial period. Then get this extrapolated fit:

    http://imageshack.com/a/img921/9613/FAdycI.gif

    Within the training interval, the fit is very aggressive with a correlation coefficient of nearly 0.98, but outside that interval its also not too shabby, dropping to a respectable 0.71 and following a visually obvious deterministic stationary pattern. A few places shown in yellow are points of divergence. Try achieving that kind of correlation with a red noise process.

    I think its only a matter of time before the precise dynamics of ENSO is mapped out. And its likely nowhere near as chaotic as the denier Tsonis has asserted.




    ReplyDelete
    Replies
    1. And then if we try a different training interval for ENSO, the results are equally good:
      http://imageshack.com/a/img922/126/wlwAxO.png

      This model is only a little more complex than doing tidal analysis. We will let the denier Richard Lindzen eat his own words: "it is unlikely that lunar periods could be produced by anything other than the lunar tidal potential." It's difficult enough to get a model to match all the squiggles over a 100+ year interval, but then to have all the parameters match precisely to wobble and tidal periods, that's the icing on the cake. It would be easier to invalidate the model if that was not the case, i.e. if parameters don't match, can't be tide or wobble factors, but with that kind of agreement, the onus is on others to find a better model.

      Delete