Been a bit of a fuss and since I haven’t looked at UAH data for quite some time I decided to take a quick look.
Correction 10th Nov: I used the gridded data before it had been updated, Oct 2012 is missing. This has no material effect. Currently vortex server at uha is not responding.
I didn’t have current UAH data here. Turned out to be the usual pain of having to handle yet another version (goes into a database here in a common gridded format). TLS is unchanged, TMT and TLT have moved to version 5.5, more difficult than the “silent” dataset changes which do not get version bump but at least it is obvious there is change.
I don’t trust the dataset, keeps on changing. RSS is no more useful. And what shows up next time?
Seems to me that doing your best, at least trying and doing things, is right.
Figure includes 10 year low pass but to get up the nose of folks is end corrected. (about -60dB @ 10y) and much longer than the data.
The figure also shows two simple models, which are data driven (I just showed the data to software set to default apart from need to know and limiting size). This too involves a unique work. In this case it is a Fourier model but do you comprehend when I say it is non-discrete? Can’t do this with DFT or DFFT.
Experience tells me pushing past 4 terms on this data is unwise.
Such a model is time free (time and discrete output are arbitrary Nyquist excepting) hence it can sometimes extrapolate. Result here is moot, who knows, I don’t.
Seems to overlay reasonably… my data is computed from the published gridded version.
That should leave room for the unwary to have a go at me.
Now what does w/h mean? Withhold. This was arbitrary, guessed a sensible figure and used that. To end 2009.
(I’m very busy on other things, taking a break, so I probably won’t respond much, oddly perhaps I have not read the WUWT post or thread so maybe I am misunderstanding: seems Spencer was doing stats stuff without much warning yet he has done that often enough to make comment superfluous on a casual, I don’t like splines)
Tim
Nice work. Simple patterns, huh? A two-up-one-down of 6.5 years, with the second “up” greater than the first because the overall trend is up. But perhaps the two-up is actually one, with a suppression/oscillation: are we seeing active negative feedback?
Prior to 1978 global temps were down, which means that the curve on the left goes down, not mathematically correct but the actual trend if that is what you are looking to see.
In part from my work on Central UK temps and sunshine hours, I see 2010 as the rollover point. As such, temps are going to go down a la Archibald. So I see 2013 as a lower peak, at 0.3C (smoothed), leading to a “down” of -0.5C in 2016, and a progressive decrease to +0.5C in 2017.5. Actually, I think lower by 0.5C, but I haven’t got my stuff in front of me because I am in a bar in Vancouver, BC right now.
We are on the cusp of very interesting times vis-a-vis CAGW theories. By 2015 I think we will know if Gore is a goat or a hero. (I vote for goat. Suzuki will be one, too. He already has the appropriate beard, so maybe that is his opinion, too.)
I love science: it appeals to the risk-taker in me, as we get to make predictions and then the Universe rolls the dice.
Good work Tim.
It’s a nice ENSO model. I hope you don’t mind, but I’ve taken the liberty of scribbling on your plot to illustrate the relationship between ENSO and the solar cycles.
http://tallbloke.files.wordpress.com/2012/11/uah-tsi.png
I think this also shows that Pinatubo coincided with a natural downswing in the ENSO cycle, which indicates that volcanic forcing is overestimated along with sensitivity studies based on the event.
By the way, Roy Spencer told us on his blog a couple of months ago that an update was on its way.
The graph is interesting…. The temp plateau of the 21st cty comes out well…..
Objection: The great max temp spike of 2013…….derived by curve fitting…..
…there lies the dog in the grave…..I do not believe in a 2013 temp spike, this
would go against the literature [not allowed to be mentioned on this site]…
Let’s dig out this forecast in 6 months time and compare then, whether the 2013
up-spike is only wishfull thinking or whether the spike really has materialized….
…next summer is the time, hopefully, Rog remembers…….JS
“The literature” Joachim refers to is his own work, peer reviewed here and at WUWT recently. It can be mentioned and the talkshop thread can be linked, but I won’t allow links to external commercial sites. I already told Joachim that, so I think his comment , implying that his work is being ‘suppressed’, is in bad faith.
If he looks carefully at Tim’s plot, he will see that the peak of the current oscillation is forecast for the end of this year and January next year, not in summer 2013. Considering the way LT temperature has reached a plateau in the last couple of months, and that the SST’s LT lags started to fall this month, I would say it is possible that LT will fall from here. If so, then Tim’s model slightly overestimates the amplitude and duration of the recent positive cycle phase, but only by a couple of months, and by less than 0.05C.
I would venture that this may be due to the non-linear effect of the recent strong downturn in solar activity which the model doesn’t capture fully, due to the way it derives best fit combinations of cycles from a data range covering a longer period.
I used to think that the UAH data was the best measure of global temperatures because it had a large geographical spread with lots of ‘measurement’ points. However, as Tim points out, there have been so many different version of the data that I no longer have confidence in it.
If you compare the previous months old data to the new v5.5 you will see that temps prior to ~1997 have been cooled slightly, but warmed post this date (1997 is like the pivot of a see-saw). If you load the following two graphs and toggle between your browser tabs you can easily see this change…
http://www.drroyspencer.com/wp-content/uploads/UAH_LT_current.gif
http://www.drroyspencer.com/wp-content/uploads/UAH_LT_1979_thru_Oct_2012_v5.5.png
A few more ‘tweaks’ like this over the next few years will ensure the UAH data is fully ‘on message’ and will make it useless as far as testing Tim’s model for future temperature change goes.
I don’t understand ENSO. A reasonable sane dataset seems to be at
http://www.jisao.washington.edu/data/globalsstenso/
HIndcasting a model is trivial. (type in the decimal date)
I’m unconvinced there is either much of a connection in the satellite era, or that any timing holds.
Hi Tim,
I think you’d get a better looking fit if you had kept the amplitude of the model curve the same as you had it in the headline post.
By the way, I love the ‘show source’ trick.