Guest Post by Willis Eschenbach
On another thread, a poster got me thinking about the common practice of using the El Nino 3.4 Index to remove some of the variability from the historical global average surface temperature record. The theory, as I have heard it propounded, is that the temperature of the Earth is “signal”, whereas the El Nino cycles are natural swings and as such are just “noise”. So if you remove the El Nino swings from the temperature, the theory goes, then we can see more of the underlying temperature signal by removing the noise.
Figure 1. Various “Nino regions” used in the study of the El Nino / La Nina phenomenon. Each area has its own index, with one of the most commonly used being the Nino 3.4 Index. SOURCE. See also the NOAA page
The more I thought about the practice of subtracting the Nino 3.4 variations from the global average temperature anomalies, the more questions came up for me. I don’t have the answers, hence this post. The first question that came up is, how do we decide that the Nino 3.4 Index represents noise instead of signal?
The Nino 3.4 region covers about 2.4% of the planet’s surface, a bit bigger than the USA. So … why isn’t the temperature of the USA “noise”? Or perhaps, is the temperature of the US “noise” but no one ever checked? And how would you check? What mathematical procedure would allow us to discriminate? What test would we use to say well, Nino 3.4 is noise so we can safely subtract its effects from the global temperature signal, but, for example Nino 1+2 is not noise, it’s part of the signal?
My next question about the situation revolves around the fact that the Nino 3.4 Index is merely a linear transform of the sea surface temperature of the Nino 3.4 area. So what we are doing is taking a linear transformation of the surface temperature anomaly in one part of the world, and subtracting it from the global average surface temperature anomaly.
As a result the question is, is this a legitimate operation? Subtracting a linear transform of something from the whole of which it is a part? Like, say, taking the average temperature variations in the whole US including Texas, but then subtracting out some linear transform of the temperature variation in Texas? What is the meaning of that procedure, subtracting something from itself? And if we are going to subtract a transform of say the Nino 3.4 temperature from the global average, should we include the Nino 3.4 temperature to begin with when we calculate the global average, or not?
Next question is, is this a legitimate operation in a system with a thermostat? Like for example, taking the variations in my body temperature, but subtracting out some linear transform of the temperature variations in my foot? What does that procedure give us, what does the result mean?
Next question. If we’re going to remove the transform of the El Nino Index from the global average temperature record, then should we remove the other indices as well? Should we remove the AMO (Atlantic Multidecadal Oscillation) Index? The PDO (Pacific Decadal Oscillation) Index? The Madden-Julian Oscillation Index? Some combination of them? All of them?
Final question. From my perspective, the El Nino/La Nina oscillation actively regulates heat loss, and thus is part of the planetary temperature regulation system. It regulates the heat loss by way of both the ocean and the atmosphere. Let me give a functional explanation of how it works. The explanation is slightly but not significantly simplified.
During La Nina conditions, in the upper part of Figure 2 below, the warm blanket of water normally covering the Pacific has been blown to the west by the strong eastern trade winds. From there, that mass of warm Pacific surface water splits and moves north and south along the coasts of Asia and Australia towards the Poles. The mass of water is radiating and losing heat as it travels. Functionally, the El Nino/La Nina alteration serves as a huge, slow-cycling, thermally regulated Pacific-wide pump. The La Nina pump stroke moves warm Pacific surface water poleward to lose its heat through conduction, radiation, and evaporation.
Figure 2. La Nina and El Nino conditions. North and South America are the brown areas in the upper right. Australia is at the lower left. Black arrows in the atmosphere show the direction of atmospheric circulation. White arrows show surface ocean currents SOURCE: NOAA El Nino Theme Page
In addition to moving warm Pacific water poleward, the removal of the warm Pacific tropical surface waters exposes the atmosphere to huge amounts of cooler sub-surface Pacific water. This lowers the air temperature over that whole area of the tropical Pacific. Soon, however, the surface of the Pacific starts to warm again. One effect of this is that it slows down the eastern trade winds. As a result of reduced winds and reduced clouds, the warming of the surface of the Pacific continues. In addition, some of the warm surface water in the Western Pacific moves back out east. Soon, with the sun beating down on an ocean with reduced clouds, it warms up all across the Eastern Pacific. This leads to neutral conditions, which can last a while.
However, if the tropical Pacific surface temperature warms enough, then El Nino conditions develop. After the El Nino conditions come into being, at some point as the surface of the Pacific continues to warm, and the El Nino thunderstorms drive the surface air upwards, the eastern trade winds start to strengthen. Soon the eastern trade winds start pushing the warm tropical surface waters and their associated thunderstorms and clouds to the west across the Pacific and eventually poleward again. This is the power stroke of the pump, when the trade winds strip the warm surface waters off and push them westwards. In this process, the full La Nina conditions come into existence. Finally, the La Nina conditions eventually peter out to a neutral condition once again.
Note that this system is triggered by temperature. If the temperature doesn’t build up across the surface of the eastern Pacific for some reason, then things stay neutral, neither El Nino or La Nina. In that case, the El Nino doesn’t form, and so the eastern trade winds don’t build up to pump the warm water across the Pacific and towards the poles.
But when the surface waters of the Pacific do heat up beyond a certain point, El Nino conditions arise, the eastern trade winds strengthen and pump the warm tropical surface water, first across the Pacific and then to the poles. It also exposes the atmosphere to a large area of cooler subsurface water.
Note the effect of this amazing temperature regulating heat pump. It functions to prevent any long-term buildup of heat in the waters of the surface Pacific. If the water in the surface of the Pacific stays cooler, the heat pump doesn’t kick in. But as soon as a certain amount of heat builds up in the surface Pacific waters, the El Nino/La Nina alteration occurs, pumping the surface water west to be flushed out toward the poles. The layer of warm surface water that was blown west is then replaced by cooler water from the subsurface, cooling the entire tropical Pacific.
This mechanism, this El Nino/La Nina pump skimming off the hot Pacific water and pumping it to the poles, prevents long-term Pacific heat buildup and thus actively keeps the planet from both overheating and excessive cooling. It is one of the many interacting thermoregulating mechanisms that keep the earth from either overheating or becoming too cool.
So … this brings up the final question regarding the theme of this post.
Since the variations in the Nino 3.4 index are indicative of the functioning of one of the Earth’s major thermoregulating mechanisms, namely the giant El Nino/La Nina pump that magically materializes to move warm tropical Pacific water to the poles whenever the planet gets too hot and sweaty … then under what possible construction could the Nino 3.4 Index variations be called “noise”?
Like I said … lots of questions, I don’t have the answers, all courteous contributions welcomed.
Regards to all,
w.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
I’d add that, for me, the big question mark over the forcings theory is its utility. Basically, does it produce useful predictions.
from wikipedia,
Although it is generally believed that one theory replacing another means that the previous has been falsified, in reality the reason why one theory is accepted and another abandoned is because the former is considered to have more utility than the latter. In other words, the previous theory is not falsified, but rather, the new theory is simply preferred by scientists as it helps solve certain puzzles that the previous theory was not capable of addressing. The criterion of falsifiability is therefore limited. This has been illustrated in the work of Thomas Kuhn, The Structure of Scientific Revolution.
Philip Bradley:
At January 18, 2013 at 2:01 pm you say
Let me answer that for you, it doesn’t.
Or, to be precise, to date all its predictions have been falsified as being wrong.
Richard
“rgbatduke says:
January 18, 2013 at 5:20 am
“Words fail me.”
I did indeed, laugh out loud.
And actually, I was expecting a slightly longer answer. Maybe 50 minutes worth.
I hope I haven’t completely missed your point.
I’m not sure. The point was that one can’t trivially remove the effect of a single input of a complex nonlinear multivariate function with an equally complex “memory” with a wide range of timescales. My secondary point that if one attempted to do so empirically, the immediate result would be the complete elimination of greenhouse-caused warming as the only apparent lasting “signal” in the satellite record is a burst of warming caused by the 1997-1998 El Nino, not a steady warming associated with smoothly increasing GHG concentration. In fact, over the 33 year satellite data, there is almost no discernible correspondence between the latter and the global temperature(s), e.g. UAH LTT.
That doesn’t mean that there isn’t one, but it does mean that the noise is much larger than this “signal”, and any attempt to remove the signal to extract the noise presumes a knowledge of the noise and signal that a) nobody has; and b) to the extent that it is input in the form of assumptions, begs all questions and proves nothing.
rgb
richardscourtney says:
January 18, 2013 at 2:23 pm
Let me rephrase that into 2 statements.
The big question mark over the forcings theory is its utility. Basically, does it produce useful predictions.
The big question mark over the forcings theory is its utility. Basically, can it produce useful predictions.
I’d agree with you on the first statement, but I’m unsure about the second. I think it could produce useful predictions but major uncertainties need to be addressed; aerosols, BC, maybe GCRs, indirect solar forcings, etc.
richardscourtney says:
Philip Bradley:
My post addressed to you at January 18, 2013 at 2:23 pm answered your post at January 18, 2013 at 2:01 pm which asked if the forcings theory produces useful predictions.
My response said
At January 18, 2013 at 4:12 pm you have replied and your concluding paragraph in that reply says
OK. If you dispute that “to date all its predictions have been falsified as being wrong” then please state one that has been right.
And remember, it only takes one exception to falsify a hypothesis.
Richard
Richard,
One of my issues with the forcings theory is, its not clear how it can be falsified. That it has to date produced no correct predictions doesn’t falsify it, because the magnitude of the forcings used (in the models) are (likely) wrong and there likely are yet to be discovered or accepted forcings.
And see my quote above from wikipedia concerning utility.
Willis Eschenbach, Bob Tisdale, richardscourtney,
Thank you for good posts. Bob Tisdale has made a good case that ENSO is a part of a high-dimensional, non-linear, oscillatory, dissipative heat transfer system. Most of the posts here clarify how that affects some of the statistical analyses (treating ENSO as “noise” or a “covariate”.)
Philip Bradley: Articulate and lucid as always, but my understanding of the forcings theory (and the predictions out of the climate models) is that the internal complexity of the Earth’s climate doesn’t matter.
That only applies to the equilibrium state (or possibly the “steady state”, though the word “equilibrium” is what is used.) What happens in between now and then depends on the internal complexity of the climate system. Standard texts make clear that they are deriving the change in the equilibrium temperature, without clarifying what, if anything, in the Earth climate system is represented by the calculated result. It’s some sort of leap of faith that if the calculated hypothetical equilibrium temperature increases then the global mean has to increase by that much. But when, and what happens in the mean time, are unspecified, or guessed.
rgbatduke January 17, 2013 at 9:29 am
First, I fully agree with you when you point out that “… Somewhere in the world, as I type this, not one but hundreds of millions of people are cooking a sparse day’s meal on animal dung or a small charcoal fire…. We have the technology, we have the wealth, to utterly eliminate global poverty within a few decades. …. And we will never succeed in doing so at the same time we make energy more expensive and discourage its use. The poverty in question is energy poverty…” It is an energy poverty created by pseudoscientific advocates of global warming to fight a non-existent AGW. But then you go on and still say that “….AGW is by no means disproven by the last 15 to 18 years of arguably flat temperatures… Temperature change cannot either prove or disprove the (C)AGW hypothesis….” That is utter nonsense. It can and it does disprove AGW as I will show. First, you have to understand the global temperature curve distributed by the climate science establishment. It is falsified to recruit people into believing that global warming is real. As you mention, there has not been any warming lately. But do you have any idea of what happened before that? If you have been looking at temperature curves put out by NOAA or by the recent draft climate report by NCADAC you will have noticed a large red triangle that includes prominently a “late twentieth century warming” in the eighties and nineties. That is supposedly anthropogenic warming, the very warming that Hansen testified about to the Senate in 1988. That testimony itself was a travesty and my book describes its background. Fortunately this temperature period overlaps satellite temperature measurements and according to the satellites there was no warming at all for 18 years. But Hansen claimed to the Senate that anthropogenic global warming had started and showed computer predictions of future temperature rise up to 2019. His predictions have been proven spectacularly wrong. Six months after he spoke that El Nino was finished and a La Nina that followed had already reduced global temperature by 0.4 degrees Celsius. And now back to the temperature record. There has been no warming whatsoever this century. Satellites show no warming from 1979 to 1997. What happened next was the super El Nino of 1997/98. It was an outlier, an extremely rare and powerful warm peak you might get once in a century. It brought so much warm water across the ocean that it caused a step warming. In four years global temperature rose by a third of a degree Celsius and then stopped. This was the only warming within the last 33 years. It is quite impossible for greenhouse warming to cause any kind of step warming. It, and not any imaginary greenhouse warming, is the cause of the very warm first decade of this century. And with it, we can say that there has been no greenhouse warming at all for the last 33 years. Now lets backtrack to the beginning of the twentieth century. It turns out that for the first ten years of the century we had cooling. Then, in 1910, warming suddenly started, kept going until 1940 and then stopped. There was no sudden increase of atmospheric carbon dioxide in 1910 when the warming started. This immediately rules out greenhouse warming as a cause by the laws of physics. Bjørn Lomborg thinks this warming is of solar origin and I agree with him. The warming ended abruptly with the arrival of the severe World War II cold spell. This is another indication that it cannot be greenhouse warming which cannot be turned on or off like this. With that forty percent of the twentieth century is expelled from the greenhouse. After the war there was no warming in the fifties, sixties and seventies until 1976. That is when the Great Pacific Climate Shift took place that supposedly raised global temperature by 0.2 degrees Celsius. Another step warming which cannot be claimed by any greenhouse effect. It was over by 1980 which overlaps the beginning of the satellite era we already covered. All in all, I cannot see any period within the last 100 years that can legitimately be called a greenhouse warming period. How can you say that temperature change can neither prove nor disprove the existence of AGW? If there has been no greenhouse warming in recorded history are you still going to wait for it to happen? I purposely left out the explanation of why there has been no greenhouse warming. The explanation is that greenhouse theory is just plain wrong as Ferenc Miskolczi has shown. QED
According to Bob Tisdale, the most important thing I can recall is that it can be shown that the steps in warming over the 20th century can be shown to be as a result of the ENSO process. Namely, the overcharging from some La Ninas. The La Ninas can remain long enough to reduce cloud cover such that more solar irradiation warms the cloudless pacific equatorial region. This heat is stored in the oceans to eventually release more heat to the global. It shows that the warming trends were natural and caused by the sun being allowed to shine past the clouds and into the pacific ocean.
Calling ENSO a zero sums game is fool-hardy if you do not know the mechanisms involved.
Typo in my last post proves should be process.
[Fixed. -w.]
I think some could benefit from seeing the actual Nino 3.4 Index over time.
The trend from 1871 to where Jan 2013 looks to end up is -0.000033C per month or 0.004C per decade. It varies by +/-3.000C so a trend of 0.004C/decade is pretty well lost in the measurement error.
http://s7.postimage.org/5zecegoaz/Nino_3_4_Jan13.png
It goes up and it goes down and it exhibits Zero trend.
And also the AMO index back to 1854 as well the Raw data [red line] it comes from (the AMO is the detrended version of this raw SST data). The AMO charts on the internet are the smoothed version of this which is technically not as useful as the straight monthly data.
http://s8.postimage.org/8mv5m0jnp/AMO_Index_and_Raw_Dec12.png
It can and it does disprove AGW as I will show. First, you have to understand the global temperature curve distributed by the climate science establishment. It is falsified to recruit people into believing that global warming is real. As you mention, there has not been any warming lately. But do you have any idea of what happened before that?
We’ll have to agree to disagree on this. For one thing, the physics of the GHE is fairly clear, and predicts roughly a degree to a degree and a half absolute warming with a doubling from 300 to 600 ppm, where we have already reached 400 ppm. The physics could be wrong, but the evidence for the GHE itself is IMO absolutely clear (visible in TOA and BOA IR spectra) and most scientifically competent CAGW skeptics don’t really argue about this.
What they argue about is the feedbacks. The whole argument, really, is about the feedbacks. If the feedbacks are positive, this CO_2 induced warming is further amplified. If the feedbacks are negative, it is reduced. If they are net-neutral, we are looking at the 1 to 1.5 C of anthropogenic, CO_2 linked global warming that even confirmed and vocal skeptics like Christopher Monckton acknowledge are likely (but hardly catastrophic).
At this point, one can argue that there is sufficient evidence to reject large “climate sensitivity” (another terminology for the feedback), because the temperature curves that have not been futzed around with — the 33 year satellite record(s) — are deviating from the large sensitivity predictions by enough to reject the outlying values, and indeed the IPCC AR reports have systematically cut down their predictions of total warming so that in the current one they are calling for a mere total doubling of the expected CO_2-only warming. This is safely impossible to falsify with the reliable, unadjustable modern era data, but an unbiased fit to the 33 year data suggest net-neutral feedback as the most likely univariate extrapolation.
However, all of the arguments I made for complexity that make it impossible to untangle ENSO from global mean temperature make it equally impossible to untangle CO_2 forcing from global mean temperature. So the fact of the matter is we just don’t know. Limited AGW is likely, CAGW is possible but relatively unlikely at this point, and the data is consistent with neutral or even weak negative feedback. There is too much we don’t understand to claim more, and IMO it weakens the skeptical argument to make egregious claims of “disproof” of AGW just as much as the CAGW claims of “proof” of catastrophe weaken their claims.
And a proper scientist has to be somewhat dispassionate and remain open minded, regardless of their personal biases on the political and economic issues contingent on various possible scientific outcomes. So when I say “I don’t know”, I really mean “and I don’t think anybody does”. Is that so unreasonable?
rgb
Bill Illis says:
January 19, 2013 at 4:29 am
Thanks, Bill. I, on the other hand, think some could benefit from seeing the source of the data used to create your graph of the actual Nino 3.4 Index over time. It drives me nuts to go see a graph and find it interesting, only to discover that the poster gives me absolutely no idea where the they got the mystery data.
Citations, citations, citations! Provide them or get roundly ignored!
w.
rgbatduke: For one thing, the physics of the GHE is fairly clear, and predicts roughly a degree to a degree and a half absolute warming with a doubling from 300 to 600 ppm, where we have already reached 400 ppm.
The simplified physical models predict an increase in the equilibrium temperature. How is it known that the predicted increase in the equilibrium temperature relates to anything in the actual climate system? The “climate” is merely (merely?) the aggregate of all of its component parts and processes, and something like the “mean Earth surface temperature” refers to a quantity that no part ever has, except small areas for for brief intervals of time. The climate system has great heat and mass flows that are not constant, and the effects of CO2 changes on those mass and heat flows are mostly not known. How is it known that all of those known (and even more unknown) unknowns are irrelevant, and that the calculated change in the equilibrium temperature accurately represents some change on Earth?
Also, I notice that some writers use “equilibrium” and “steady-state” almost interchangeably. With steady-state, there are heat flows into and out of every volume element of the climate system, but no temperature change — a change in that is what AGW predicts (though even there, at best there might be a new “stationary distribution” of measurements, with higher mean temperatures in most places.) But in the derivations, such as Pierrehumbert’s derivation of the Clausius-Clapayron law, what is assumed is clearly “equilibrium”. Is there some climate dictionary or style manual which clearly explains when it is ok to use “equilibrium” and “steady-state”?
Willis: If you’re looking for a long-term trend and you believe that you understand the source of some of the noise in the long-term signal, it certainly makes sense to try to subtract some of this noise. Everyone does this almost without thinking when they remove the annual signal and work with temperature anomalies. There are practical, but not theoretical, reasons why you can’t do the same thing with the PDO or AMO. The process of subtracting the noise contributed by any of these processes introduces uncertainty while removing noise. If proper statistical analysis says that you multiply the ENSO or PDO index by a factor of 0.28 +/- 0.19 (95% confidence interval) to estimate the “noise” contributed by these processes, then including them in your analysis doesn’t help much. If the factor is 0.28 +/- 0.9, you can remove some noise without introducing too much uncertainty. The problem with using longer cycles like the PDO or the AMO is that we probably don’t have enough data to accurately estimate how reliably they are correlated with noise in the long term trend. You probably need to do some sort of Monte Carlo calculation to see how much you improve you estimate of the long-term trend by subtracting estimated noise.
Frank says:
January 19, 2013 at 11:41 am
Yes, but it begs the question to assume that you know what is noise and what is not. I asked, how do you know that e.g. the Nino 3.4 index is “noise” while say the temperature of the US is signal, not “noise”?
Me, I prefer to think about these questions, which is the reason for this post. In addition, removing strictly cyclical signals with known clearly defined periods is a whole lot different that removing something like the Nino 3.4 index. For the former you can just average by month, but the Nino 3.4 removal involves linear regression, so your example is quite different than what we’re discussing here.
You claim the “practical reason” we can’t remove the PDO or the AMO is lack of data … I don’t see why. Yes, the pseudo-cycles are longer, but we still are doing a linear regression of month-to-month changes in the index regarding the global temperature, and we still have the same number of monthly samples for PDO or AMO as we do for the Nino 3.4 … so we could easily remove them as well.
w.
The simplified physical models predict an increase in the equilibrium temperature. How is it known that the predicted increase in the equilibrium temperature relates to anything in the actual climate system?
(and a bunch more stuff). Agreed. The Earth is an open highly multivariate dynamical nonlinear non-Markovian chaotic driven system, and statements like “1 to 1.5 degrees of warming” are themselves consequently moderately suspect. As I pointed out to Phil Bradley, the way one must interpret them is with the understood “in a separable Markov approximation”. That is, one makes the twin assumptions that one can speak of what happens if one increases only the CO_2, all other things being/remaining equal. This is more or less making certain assumptions about the partial derivatives of e.g. “global average temperature” with respect to all sorts of stuff that are almost certainly untrue and are very likely not even true enough to be a reasonable short term approximation (the separability).
It is also assumes that the past history of the climate system is more or less irrelevant to its future evolution, that no phenomenon with a time scale longer than a year or two is relevant to whether the global average temperature goes up or down (the Markov approximation). Again, this is almost certainly absurdly wrong, as the entire topic of this post is about removing the “noise” associated with one of the many multidecadal exceptions and I don’t think anybody comes close to understanding things like the ocean and its multiple timescales and effect on everything from CO_2 to the storage and delayed release of heat.
We are clearly still learning things about the climate, things that appear to be a lot more important than they were assumed to be in the earliest all-things-equal, Markov approximation models. Soot, for example, “suddenly” appears to be quite important. Soot is a nearly ubiquitous urban-industrial by-product. Recently evidence has emerged that the sun’s state may affect upper atmosphere chemistry much more powerfully than it was previously believed. The climate effects of those changes (if any) are still poorly understood, especially given that we barely understand what’s going on at the simple electrochemistry level, let alone at the level of complexity of the climate.
There are a number of puzzles that are clearly evident in the historical climate record over the last half billion years. Warm spells, ice ages, sometimes interleaved with “razor sharp” edges, where it is warm and then suddenly it decides to get cold until it equally suddenly warms up again. There is no reasonable way to explain some of the events with any simple picture involving continental drift, axial tilt, or modulation of atmospheric CO_2 — indeed, the Ordovician-Silurian ice age took place with CO_2 levels during the ice age that were ten times the current level.
Attempts to explain things like this, so far, smack of science fiction — things that sound like science but that are unsupported by any evidence (and likely unsupportable by evidence this long after the fact). And yet we can hardly claim to understand the climate system as long as these events not only lie unexplained, but are unexplainable in the one-size-fits-all contemporary Markovian separable model.
rgb
rgbatduke says:
January 19, 2013 at 8:34 am
First, as always, my great thanks to you for all of your contributions. I always read them with interest.
In this case, let me say again, it’s not about feedback. Analyzing it in terms of feedbacks is useless in a situation with a governor that is capable of overshoot.
The problem is that the governor applies either negative or positive feedback as needed at that moment to keep the governed parameter within bounds. Over a long period, the average of these could be positive, negative, or zero. To show how meaningless this feedback number is, consider the situation where the net feedback is zero. Does this mean that nothing is happening? Not in the slightest. The system is constantly going up and down, but the application via the governor of the appropriate feedback keeps it within a narrow range despite the feedback averaging zero.
In other words, you can’t analyze a governed system by seeing if the net feedback is positive, negative, or zero.
w.
Frank says: “If you’re looking for a long-term trend and you believe that you understand the source of some of the noise in the long-term signal, it certainly makes sense to try to subtract some of this noise. Everyone does this almost without thinking when they remove the annual signal and work with temperature anomalies. There are practical, but not theoretical, reasons why you can’t do the same thing with the PDO or AMO.”
First of all: An ENSO index such as NINO3.4 sea surface temperature anomalies are in fact a measure of the sea surface temperature of the NINO3.4 region. The Multivariate ENSO index is a modified NINO3 region sea surface temperature ENSO index. The AMO data from the NOAA ESRL is a detrended North Atlantic sea surface temperature anomaly dataset. On the other hand, the PDO is an abstract form of the sea surface temperature anomalies of the North Pacific, north of 20N, and the PDO is actually inversely related to the sea surface temperature anomalies there. That’s why you can’t remove the PDO from the surface temperature record.
With respect to ENSO, when anyone attempts to remove ENSO from the instrument temperature record, they’re removing the signal primarily from the East Pacific Ocean and other parts of the globe where temperature responds linearly to the ENSO index. That is, using a scaled and lagged ENSO index to remove the ENSO signal from the global surface temperature record will only capture the portions of ENSO in the global temperature record that respond proportionally to El Niño AND La Niña events as represented by the ENSO index. And as we can see, the East Pacific for the most part responds proportionally to El Niño and La Niña events.
http://bobtisdale.files.wordpress.com/2012/12/east-pac-vs-scaled-nino3-4-ssta.png
The East Pacific hasn’t warmed in 31 years so there’s no reason to detrend that dataset for the comparison to the scaled NINO3.4 data. However, the sea surface temperatures for the Rest-of-the-World (that’s not the East Pacific) has warmed so we have to detrend it for a comparison to the NINO3.4 data:
http://bobtisdale.files.wordpress.com/2012/12/figure-10-detrended-row-vs-nino3-4.png
As we can see, the Rest-of-the-World data warms during the El Niño events of 1986/87/88 and 1997/98 but it does not cool proportionally during the La Niña events that trailed them. Because the Rest-of-the-World sea surface temperatures do not cool proportionally during La Niña events, you cannot remove the effects of ENSO using an ENSO index—El Niño and La Niña do have proportional impacts there.
The reason for this is that there can be a huge volume of warm water that’s leftover from an El Niño. That was one of the basic messages of Willis’s post. Part of that leftover warm water (a Rossby wave) is captured in the following animation of sea level anomalies from JPL. The animation starts at the peak of the 1997/98 El Niño. Quite plainly, you can see a huge volume of warm water (it’s below the surface and north of the NINO3.4 region, so it’s never noticed by the ENSO index) being returned to the western tropical Pacific at the end of the 1997/98 El Niño. Watch what happens when that phenomenon called a slow-moving Rossby wave reaches Indonesia. It’s like a secondary El Niño event taking place in the western tropical Pacific, but it’s happening during the La Niña. All of that leftover warm water counteracts the effects of the La Niña throughout the globe. It causes the divergences during the La Niñas that follow the major El Niño events that were illustrated above. The YouTube edition of the full animation from JPL is here.
As a result of the leftover warm water, the sea surface temperature anomalies of the Rest-of-the-World appear to shift upwards in response to the strong El Niño events:
http://bobtisdale.files.wordpress.com/2012/12/figure-8-row-a.png
In fact, without the strong El Niño events, the Rest-of-the-World sea surface temperature anomalies would not have warmed since 1984:
http://bobtisdale.files.wordpress.com/2012/12/figure-9-row-b.png
In the following video, you can actually watch an upward shift occur in a portion of the Rest-of-the-World data. The East Indian and West Pacific Ocean subset was where I first discovered the upward shifts caused by those El Niño events. The animation of sea surface temperature anomalies also has an infilling graph to its right that compares the East Indian-West Pacific data to scaled NINO3.4 data.
http://bobtisdale.files.wordpress.com/2012/02/east-indian-west-pacific-97-thru-012.gif
Part of the upward shift is caused by the warm water returned to the western tropical Pacific by the Rossby wave. And part of it caused by the warm surface waters being blown back to the west when the trade winds resume after the El Niño.
Regards
Willis Eschenbach says:
January 19, 2013 at 9:19 am
Citations!
——–
Nino 3.4 Index from the original Trenberth,97 paper is here going back to 1871 ending in 2007 – the first reconstruction considered reliable.
http://www.cgd.ucar.edu/cas/catalog/climind/TNI_N34/index.html#Sec5
The data is being updated by the CPC here (most commonly used one now) 1982 to current.
http://www.cpc.ncep.noaa.gov/data/indices/sstoi.indices
Weekly Nino 3.4 values on the same basis here.
http://www.cpc.ncep.noaa.gov/data/indices/wksst8110.for
You can also get the Nino 3.4 values from various different SST datasets at the Climate Explorer – monthly climate indices section. They vary a little from the Trenberth version but not much. I’ve noticed the Climate Explorer data is starting to have mistakes now so one has to double-check downloads from here now.
http://climexp.knmi.nl/selectindex.cgi?id=someone@somewhere
——–
The AMO can be obtained from the NOAA ESRL division which is the maintainer of the accepted official number.
http://www.esrl.noaa.gov/psd/data/timeseries/AMO/
The easy link to the monthly data is here:
http://www.esrl.noaa.gov/psd/data/correlation/amon.us.long.data
And the Raw SST data is here. (with the average climatology linked right after).
http://www.esrl.noaa.gov/psd/data/correlation/amon.us.long.mean.data
http://www.esrl.noaa.gov/psd/data/correlation/amon.climo.data
Bill Illis, thanks for the double handful of citations to everything under the sun.
Unfortunately, you didn’t answer my question. I had asked:
You gave me a whole raft of citations, but as far as I can see, you never did indicate which one was the source of the data that you actually used …
w.
Bob Tisdale says:
January 19, 2013 at 3:32 pm
As I said above, there is absolutely no reason we can’t remove either the AMO or the PDO from the temperature record. The fact that the PDO is inversely related to the temperature is not a roadblock to doing that in any sense. You just do the regression and subtract it from the temperature data, what’s the problem?
Now, I don’t think you should do that (unless you’re studying the PDO and not the temperature), but that doesn’t mean you “can’t remove the PDO from the surface temperature record” as you state. Of course you can … it just may not give you anything meaningul.
w.