Joe D’Aleo and Don Easterbrook have produced a new paper for SPPI. This graph of US Mean temperature versus the AMO and PDO ocean cycles is prominently featured:
I particularly liked the regression forecast fit:
They have this caveat:
Note this data plot started in 1905 because the PDO was only available from 1900. The divergence 2000 and after was either (1) greenhouse warming finally kicking in or (2) an issue with the new USHCN version 2 data.
Hmm. I’m betting USHCNv2.
Abstract:
Perlwitz etal (2009) used computer model suites to contend that the 2008 North American cooling was naturally induced as a result of the continent’s sensitivity to widespread cooling of the tropical (La Nina) and northeastern Pacific sea surface temperatures.
But they concluded from their models that warming is likely to resume in coming years and that climate is unlikely to embark upon a prolonged period of cooling. We here show how their models fail to recognize the multidecadal behavior of sea surface temperatures in the Pacific Basin, which determines the frequency of El Ninos and La Ninas and suggests that the cooling will likely continue for several decades. We show how this will be reinforced with multidecadal shift in the Atlantic.
Here’s the paper you can download:
UPDATE: The goodness of fit, seems almost too good. There may be a reason. I’m reminded in comments of this article by statistician William Briggs – (thanks Mosh)
Do not smooth times series, you hockey puck!
Where he points out:
Now I’m going to tell you the great truth of time series analysis. Ready? Unless the data is measured with error, you never, ever, for no reason, under no threat, SMOOTH the series! And if for some bizarre reason you do smooth it, you absolutely on pain of death do NOT use the smoothed series as input for other analyses! If the data is measured with error, you might attempt to model it (which means smooth it) in an attempt to estimate the measurement error, but even in these rare cases you have to have an outside (the learned word is “exogenous”) estimate of that error, that is, one not based on your current data.
If, in a moment of insanity, you do smooth time series data and you do use it as input to other analyses, you dramatically increase the probability of fooling yourself! This is because smoothing induces spurious signals—signals that look real to other analytical methods. No matter what you will be too certain of your final results! Mann et al. first dramatically smoothed their series, then analyzed them separately. Regardless of whether their thesis is true—whether there really is a dramatic increase in temperature lately—it is guaranteed that they are now too certain of their conclusion.
Perhaps Mr. Briggs can have a look and expound in comments. I only have the output, not the method. But let’s find out and determine how good the “fit” truly is. – Anthony
UPDATE: Statistician Matt Briggs responds in depth here. He says:
I want to stress that if D&E did not smooth their data, the correlation would not have been as high; but as high as it would have been, it would still have been expected. All that smoothing has done here is artificially inflated the confidence D&E have in their results. It does not change the fact that AMO + PDO is well correlated with air temperature.
Great. The relationship couldn’t be clearer!
Everyone who is exposed to that awful 10:10 video should be given this to read as an antidote.
However, surely if the relationship is there in the series the analyses really need to be done with yearly data? A high correlation coefficient between two sets of highly averaged data is meaningless in this case.
Stephen Mosher:
Very good catch, Stephen.
Leif Svalgaard:
Not without reading e.g. Lucia’s blog first to learn how to appropriately treat the correlation you’ve introduced to your series by smoothing. It can be treated correctly, if you’re careful though.
REPLY: I’ve invited statman Matt Briggs over to find out – Anthony
From what I can see, between 1905 and 1975, temp seems to be preceding the AMO+PDO. It could be a slight aberration in the data, the smoothing, both or adding AMO+PDO together. Better check it, or we’ll have The Goreacle saying that it’s complicated, but whenever the temperature goes up, the oscillations go up..
Hockey Schtick says:
September 30, 2010 at 8:18 pm (Edit)
Add in the ‘sunspot integral’ for an even better R^2 = .96
http://hockeyschtick.blogspot.com/2010/01/climate-modeling-ocean-oscillations.html
Nice one. I found similar relationships:
http://tallbloke.wordpress.com/2010/01/05/my-simple-solar-planetary-energy-model/
http://tallbloke.wordpress.com/2010/07/21/nailing-the-solar-activity-global-temperature-divergence-lie/
The points being made on this thread about smoothing are valid for the comparison of two series for which a one to one relationship is sought. However climate is complex, as we know, and many factors act to make the short term data very noisy. Without sufficiently good models to enable the removal of short term influences, smoothing is the one way to try to get a handle on possible principle underlying causative factors.
Holding up a R^2 figure for smoothed series correlations and saying
“Look, this proves it!”
is clearly the wrong way to use the data, but this does not mean we should toss out any and all smoothed series comparisons. They have a use and value, just not a useful numerical ‘truth’ value.
I was under the impression that 1934 was the hottest year in the US record.
Appart from the big smoothing before correlation problems already mentioned, in the graph presented above I see temperatures leading AMO/PDO until 1975. Causality, if it exists, doesn’t seem to go in the direction that the author pretends that it goes.
tallbloke says:
October 1, 2010 at 12:25 am
this does not mean we should toss out any and all smoothed series comparisons. They have a use and value, just not a useful numerical ‘truth’ value.
Of course they have value, just don’t calculate r-squared from them. If you want to suppress noise, then use averages. E.g. with a 100 data points, calculate the 5 average values of successive blocks of 20 points. Do this for both time series, then calculate the regression and the statistical significance of a line through those 5 data point pairs.
First, the contiguous U.S. land surface temperature anomalies do follow the SST anomalies of the U.S. Coastal Waters:
http://s5.tinypic.com/209r3t2.jpg
The graph is from this post:
http://bobtisdale.blogspot.com/2009/03/sst-anomalies-of-us-coastal-waters.html
But….
D’Aleo and Easterbrook write, “Not surprisingly, El Ninos occur more frequently during the PDO warm phase and La Ninas during the PDO cold phase. It maybe that ocean circulation shifts drive it for decades favoring El Ninos which leads to a PDO warm phase or La Ninas and a PDO cold phase (the proverbial chicken and egg), but the 60 year cyclical nature of this cycle is well established (figure 6).”
They have cause and effect reversed. The PDO is positive when the frequency and magnitude of El Niño events exceed those of La Niña events and vice versa. Refer to the Introduction to the PDO:
http://bobtisdale.blogspot.com/2010/09/introduction-to-enso-amo-and-pdo-part-3.html
Also, the “60 year cyclical nature of this cycle” [PDO] does not exist prior to the instrument temperature record. That is, there are no 60-year cycles in any of the paleoclimatological reconstructions of the PDO. Refer to:
http://bobtisdale.blogspot.com/2010/03/is-there-60-year-pacific-decadal.html
D’Aleo and Easterbrook write, “The AMO (Atlantic sea surface temperatures standardized) is the average anomaly standardized from 0 to 70N. The AMO has a period of 60 years maximum to maximum and minimum to minimum.”
The AMO is not simply standardized Atlantic sea surface temperatures. The AMO is detrended North Atlantic SST anomalies. They have elected to standardize it. Second, like the PDO, paleoclimatological reconstructions of the AMO also shows that the AMO does not have a constant 60-year cycle:
http://i47.tinypic.com/ekkhuc.png
Refer to:
http://bobtisdale.blogspot.com/2009/12/atlantic-multidecadal-oscillation-index.html
D’Aleo and Easterbrook write, “Although the two indices (PDO and AMO are derived in different ways…”. Then they go on to write, “I normalized the two indices to make them more comparable and added the two.”
I’ll repeat comments about this from an earlier thread here at WUWT: Unfortunately, the PDO and AMO are not similar datasets and cannot be added or averaged. The AMO is created by detrending North Atlantic SST anomalies, while the PDO is the product of a principal component analysis North Pacific SST anomalies, north of 20N. Basically, the PDO represents the pattern of the North Pacific SST anomalies that are similar to those created by El Niño and La Niña events. If one were to detrend the SST anomalies of the North Pacific, north of 20N, and compare it to the PDO, the two curves (smoothed with a 121-month filter) appear to be inversely related:
http://i52.tinypic.com/fvi92b.jpg
I’ll have to update the discussion of this in the Introduction to the PDO post:
http://bobtisdale.blogspot.com/2010/09/introduction-to-enso-amo-and-pdo-part-3.html
The D’Aleo and Easterbrook caveat reads, “Note this data plot started in 1905 because the PDO was only available from 1900. The divergence 2000 and after was either (1) greenhouse warming finally kicking in or (2) an issue with the new USHCN version 2 data.”
There are longer-term PDO datasets available from the NCDC. Here’s one based on their ERSST.v3b data:
ftp://eclipse.ncdc.noaa.gov/pub/ersstv3b/pdo/pdo.1854.latest.situ.v3b.ts
And here’s one based on their ERSST.v2 data:
ftp://ftp.ncdc.noaa.gov/pub/data/ersst-v2/pdo.1854.latest.ts
D’Aleo and Easterbrook write, “We have shown how these two ocean oscillations drive climate shifts. The PDO leads the way and its effect is later amplified by the AMO.”
The PDO does not lead the AMO. Detrended SST anomalies of the North Atlantic and North Pacific (north of 20N) can run in and out of phase. In fact, the rise in the detrended SST anomalies of the North Atlantic (the AMO) led the detrended SST anomalies of the North Pacific by about 15 years in the first half of the 20th century:
http://i56.tinypic.com/t9zhua.jpg
And that graph is from my Introduction to the PDO post:
http://bobtisdale.blogspot.com/2010/09/introduction-to-enso-amo-and-pdo-part-3.html
Guys, don’t forget that correlation does not equal causation especially with smoothed data. Having said that, it is quite logical that there will be a connection between temps and AMO + PDO. After all, the oceans drive the weather/claimate.
As regards a couple of points raised by Doug Proctor (post at 9:06 hrs on 30th Sept), how the heck can we know that the natural warming from the LIA ceased at 1988? I would like to see the IPCC’s proof of that including a detailed account of the drivers that caused the warming from the LIA and how these diminished over time such that by 1988 – they no longer produced any forcing of temperature. Second, as regards the contention that as from 1988 the warming is all AGW all I can say is that if that is so, it didn’t last very long given that Phil Jones has acknowledged that there has been no warming of statistical effect for the past 15 years, ie., since 1995. That therefore leaves just 7 years of AGW.
Surely the procedure for correlating smoothed data sets is well known (Papoulis, Bendat & Peirsol). The number of degrees of freedom of the data are reduced according to the serial correlation in the signals.
However, I agree with the general point that smoothing data for its own sake is dangerous. It is well known to introduce spurious trends. If one has data that may contain errors, and one smooths the data, there is an assumption about the spectral characteristics of the errors that may be completely wrong.
I agree with Bob Tisdale. Mixing apples and pears. PDO is independent of what is going on in the Atlantic.
http://www.vukcevic.talktalk.net/PDOa.htm
http://www.vukcevic.talktalk.net/CETnd.htm
As noted earlier, the D’Aleo and Easterbrook combined AMO+PDO graph is erroneous because they’re adding datasets that are created differently. They should have used either detrending of SST anomalies like the AMO or PC analysis of detrended SST anomalies like the PDO.
Detrended Northern Hemisphere SST anomalies (the method used to create the AMO )…
http://i53.tinypic.com/5wjkee.jpg
…wouldn’t work because the USHCN data has a trend.
And had they performed a Principle Component analysis of detrended Northern Hemisphere SST anomalies (the method used to create the PDO) they would have discovered that the curve of the 1st PC of detrended Northern Hemisphere SST anomalies bears little likeness to their AMO+PDO curve:
http://i55.tinypic.com/2mxh6o7.jpg
I did a similar thing a few weeks back on my website, except I used the AMO only and also used the CRU (Yes, CRU global data) and the correlation was also very good for the entire global mean.
The main article on that one is here:
http://theinconvenientskeptic.com/2010/09/the-all-natural-cause-of-global-warming/
The most recent article is also lots of fun. It shows how poorly the original idea behind global warming from Arrhenius fits the actual temperatures over the past 80 years.
Too bad we can’t get a meteorologist, a geologist, a solar physicist, a polymath, a statistician, and an oceanographer to sit down with a jug of Cuervo and knock out a climate paper that we could all enjoy.
“They have this caveat:
Note this data plot started in 1905 because the PDO was only available from 1900.”
PDO data is available from 1854
ftp://eclipse.ncdc.noaa.gov/pub/ersstv3b/pdo/pdo.1854.latest.situ.v3b.ts
AMO from 1856
http://www.esrl.noaa.gov/psd/data/correlation/amon.us.long.data
PDO does not match AMO
John Marshall says:
October 1, 2010 at 12:56 am
________________________________
WUWT has featured posts on adjustments to official records that have reduced temperatures during the 1930s peak relative to recent temperatures. It would be interesting to see the first graph of this post, with “unadjusted” values.
Even if they are exact, and I think they are, those calculations are too complex for the simple minds that constitute the majority of the population (and our governments). They prefer CO2 explanation.
Smoothing is no longer the data.
You have broken the eggs to make an omellete. To create a snapshot.
Smoothing is algorithmic, not statement, smoothing is approximation.
Smoothing is not statement and algorthimic process is not statement. It is approximation and can only be used as long as the fit to curve works.
Algorithm relies on data not the other way round.
The problem with using the smoothed series’ is autocorrelation… the other question I have is did they train the regression model to see if it could predict the temperatures during their reconstruction. If you just run a multiple regression over the whole period you are not actually predicted anything because you only have a training period…
Jim D says:
“This kind of thing always overlooks the acceleration of the CO2 effect. The CO2 increase from 1940-1970 was about 25 ppm. 2000-2030: about 90 ppm expected.”
1940-1970 and 2000-2030. Has we lost 30 years?
For Laymen.
The best interopolative algorithims refine on modern data, better data measurement the data itself.
But not smoothing, weighting measured to recent data and collection method.
Assigned on fit.
Yer I know back in yer box Jack.
Me I got chucked out of Maths in project work, not the maths the interpolation side. I was working for a dickhead. I took a pass conceded lost 3 honors for 3 passes and traded for the degree.
I hold a degree in modeling mostly.
Data is paramount in modelling. Data measurement will always improve. Has to, as science and tech improves.
Slipstick to computer is analogous to most science.
Oh cool. From my Guide to WUWT (see button in top right), my first “WUWT Classic” is 2008 Jan 28: Warming Trend: PDO And Solar Correlate Better Than CO2:
It’s nice to see an update, it’s even nice to some criticism about the techniques. That can only serve to improve the study (note – improvements may falsify the study, pesky thing about improvements). Now I have to sit down and apply all of this to Bob Tisdales comments and papers. Curses.
Let me add my own criticism. Figure 3 shows the PDO and a polynomial fit to the data. Polynomial fits are fine for some applications but are utterly worthless as projections on either end – if Joe plotted the polynomial for five years on either side of the fitted data it would be obvious. The fit within a few years of either end is nearly as misleading. Look at the left side. While the unfitted data is oscillating around about 0.2 (doesn’t seem to warrant the label “cool”), only a polynomial fit could be that bad, going from +0.8 to -0.2 to 0.0.
Anyone who looks at that figure and wants to see where we’re heading is welcome to do so – just don’t follow the polynomial!
Other notes:
Figure 9 shows a PDO reconstruction to 1660. It does show as cyclic, but it would take some handwaving to call it a 60 year cycle. D’Aleo and Easterbrook note “This does not mean that the PDO physically controls ENSO, but rather that the resulting climate patterns interact with each other,” which echoes Bob Tisdale’s comments above at http://wattsupwiththat.com/2010/09/30/amopdo-temperature-variation-one-graph-says-it-all/#comment-496283
So, there’s still lots to learn. My conclusion that “things were about to turn interesting” has held up.
Jim D says:
September 30, 2010 at 10:15 pm
This kind of thing always overlooks the acceleration of the CO2 effect.
The CO2 increase from 1940-1970 was about 25 ppm.
2000-2030: about 90 ppm expected.
I expect the CO2 increase rate will overwhelm this cyclical effect, since it has almost quadrupled since the last cooling cycle (if indeed the cooling was all oceans and no aerosols, which is also somewhat dubious), and nobody has accounted for that in these articles that talk about the 60-year cycle.
The effect of CO2 is logarithmic not linear. So in theory twice as much CO2 is required to be added to the atmosphere to provide the same radiative forcing effect of the CO2 already present. There are some claims made that CO2’s forcing is already effectively saturated.
But if the temperature continues stable, or starts to rise at all over the next few years (if the data can be trusted) then the sceptical position is not strengthened by this, but weakened (though not demolished since there are other imponderables!) Be careful what you wish for . . . . .