In Their Own Words: The IPCC on Climate Feedbacks
by Roy W. Spencer, Ph. D.
Despite the fact that the magnitude of anthropogenic global warming depends mostly upon the strengths of feedbacks in the climate system, there is no known way to actually measure those feedbacks from observational data.
The IPCC has admitted as much on p. 640 of the IPCC AR4 report, at the end of section 8.6, which is entitled “Climate Sensitivity and Feedbacks”:
“A number of diagnostic tests have been proposed…but few of them have been applied to a majority of the models currently in use. Moreover, it is not yet clear which tests are critical for constraining future projections (of warming). Consequently, a set of model metrics that might be used to narrow the range of plausible climate change feedbacks and climate sensitivity has yet to be developed.”
This is a rather amazing admission. Of course, since these statements are lost in a sea of favorable (but likely superfluous) comparisons between the models and various aspects of today’s climate system, one gets the impression that the 99% of the IPCC’s statements that are supportive of the climate models far outweighs the 1% that might cast doubt.
But the central importance of feedbacks to projections of future climate makes them by far more important to policy debates than all of the ways in which model behavior might resemble the current climate system. So, why has it been so difficult to measure feedbacks in the climate system? This question is not answered in the IPCC reports because, as far as I can tell, no one has bothered to dig into the reasons.
Rather unexpectedly, I have been asked to present our research results on this subject at a special session on feedbacks at the Fall AGU meeting in San Francisco in mid-December. In that short 15 minute presentation, I hope to bring some clarity to an issue that has remained muddied for too long.
To review, the feedback measurement we are after can be defined as the amount of global average radiative change caused by a temperature change. The main reason for the difficulty in diagnosing the true feedbacks operating in the climate system is that the above definition of feedback is NOT the same as what we can actually measure from satellites, which is the amount of radiative change accompanied by a temperature change.
The distinction is that in the real world, causation in the opposite direction as feedback also exists in the measurements. Thus, a change in measured radiative flux results from some unknown combination of (1) temperature causing radiative changes (feedback), and (2) unforced natural radiative changes causing a temperature change (internal forcing).
The internal forcing does not merely add contaminating noise to the diagnosis of feedback – it causes a bias in the direction of positive feedback (high climate sensitivity). This bias exists primarily because forcing and net feedback (including the direct increase of IR radiation with temperature) always have opposite signs, so a misinterpretation of the sum of the two as feedback alone causes a bias.
For instance, for the global average climate system, a decrease in outgoing radiation causes an increase in global average temperature, whereas an increase in temperature must always do the opposite: cause an increase in outgoing radiation. As a result, the presence of forcing mutes the signature of net feedback. Similarly, the presence of feedback mutes the signature of forcing.
The effect of this partial cancellation is to result in diagnosed net feedbacks being smaller than what is actually occurring in nature, unless any forcing present is first removed from the data before estimating feedbacks. Unfortunately, we do not know which portion of radiative variability is forcing versus feedback, and so researchers have simply ignored the issue (if they were even aware of it) and assumed that what they have been measuring is feedback alone. As a result, the climate system creates the illusion of being more sensitive than it really is.
One implication of this is that it is not a sufficient test of the feedbacks in climate models to simply compare temperature changes to radiation changes. “This is because the same relationship between temperature and radiation can be caused by either STRONG forcing accompanied by a large feedback parameter (which would be low climate sensitivity), or by WEAK forcing accompanied by a small feedback parameter (which would be high climate sensitivity).”
Only in the case of radiative forcing being either zero or constant in time – situations that never happen in the real world – can feedback be accurately estimated with current methods.
Our continuing analysis of satellite and climate model data has yet to yield a good solution to this problem. Unforced cloud changes in the climate system not only give the illusion of positive feedback, they might also offer a potential explanation for past warming (and cooling). [I believe these to be mostly chaotic in origin, but it also opens the door to more obscure (and controversial) mechanisms such as the modulation of cloud cover by cosmic ray activity.]
But without accurate long-term measurements of global cloud cover changes, we might never know to what extent global warming is simply a manifestation of natural climate variability, or whether cloud feedbacks are positive or negative. And without direct evidence, the IPCC can conveniently point to carbon dioxide change as the culprit. But this explanation seems rather anthropocentric to me, since it is easier for humans to keep track of global carbon dioxide changes than cloud changes.
Also, the IPCC can conveniently (and truthfully) claim that the behavior of their models is broadly “consistent with” the observed behavior of the real climate system. Unfortunately, this is then misinterpreted by the public, politicians, and policymakers as a claim that the amount of warming those models produce (a direct result of feedback) has been tested, which is not true.
As the IPCC has admitted, no one has yet figured out how to perform such a test. And until such a test is devised, the warming estimates produced by the IPCC’s twenty-something climate models are little more than educated guesses. It verges on scientific malpractice that politicians and the media continue to portray the models as accurate in this regard, without any objections from the scientists who should know better.
“Only positive feedback has the capacity to amplify the input signal in closed loop.”
Oops .. sorry, didn’t mean to say that bit! Negative feedback can amplify in closed loop, as many everyday examples of servos will happily demonstrate. (I think my mind was still on passive systems at that point).
Anna V (10:21:44):
Albedo at any level doesn’t change the total irradiance, only the fraction that is thermalized. Methinks John S. is deliberately keeping things simple in order to focus attention on the fact that an independent power source is required for any genuine feedback, to which the usual transfer functions apply. Otherwise, only recirculation is obtained. Many confuse the latter with the former.
Jordan (16:02:25):
Passive systems can recirculate (with or without dissipation) a fraction of the output, thereby stealing power from the output. Active systems sense the output, but don’t disturb it in any way. What is fed back in the latter is a replica of the output signal, which requires its own source of power. That makes for categorical differences in system behavior.
Juraj V. (10:47:58) :
Re Measurements of the Radiative Surface Forcing of Climate
Fig 1 shows measured winter spectra of greenhouse gases. How come there is no sign of water vapor emission line? If absorbtion spectrum = emission spectrum, the thick line on the left between 5-8 microns, assigned to CO2 is actually H2O line.
There’re not many H2O lines from 2000-600 cm^-1 (not microns!) and in the winter in Ontario on the day in question there wasn’t much H2O in the atmosphere. The band on the left is the ~15 micron CO2 band, 600-750 cm^-1.
Fig 2 shows measured summer spectrum, but again H2O is not assigned to any of the spectral lines.
http://z.hubpages.com/u/1293007_f520.jpg
That’s fig 3, the additional lines (the ‘grass’) cf fig 1 are the H2O lines.
The flux measurements presented in Table 2 provide important experimental verification of the driving radiation that is responsible for global warming.
First, the Table 2 does not contain water vapor flux (by far the biggest contributor to downward LW flux).
Tables 3a & 3b.
sky
“Passive systems can recirculate (with or without dissipation) a fraction of the output”
Circulation without loss sounds a little too hypothetical. I’d suggest it’s best to assume dissipation for practical systems.
I wonder if your reference to “circulating” energy is to energy storage. As I mentioned above, passive systems can be underdamped.
A basic example would be a pendulum. Or a mass hanging at the end of a spring. Both of these systems will oscillate if they are given a single shove. The energy will “circulate” between kinetic and potential energy at different phases of each cycle.
But we can all agree the initial input cannot be amplified without another source of energy. The oscillations cannot be greater than an amplitude defined by the limited energy provided by the initial shove. Without another shove, the oscillations will diminish over time due to losses.
AGH! Point taken. I’m so used to working with systems with loop gain > unity that I forget this.
“AGH! Point taken. I’m so used to working with systems with loop gain > unity that I forget this.”
Positive feedback tends to have very limited utility in engineering, and this little exception tends to be ignored and forgotten. I once got caught out on the same point in discussion ClimateAudit. AGH!
sky: “Passive systems [steal] power from the output.”
Power is an energy flow from a higher potential source to lower potential sink. The concept of a sink “stealing power” is novel, but probably of no practical use.
“Active systems sense the output, but don’t disturb it in any way.”
I disagree. Any “signal” involves a flow of energy. George E Smith will be able to tell you about the possible attenuation of electronic signals, and the importance of “impedance matching” at the input.
“That makes for categorical differences in system behavior.”
Again I disagee. If you accept the definition of passive and active systems I gave above, it will save you a lot of confusion.
Again, in reply to sky
“the fact that an independent power source is required for any genuine feedback, to which the usual transfer functions apply.”
Afraid that’s complete nonsense. Feedback is in no way conditional on an independent power source.
Imagine two tanks of water with a connecting pipe and valve. The level in one tank is below the other. When you open the valve. water flows into the lower tank, but the flow reduces as the levels in the two tanks equalise. That’s a passive feedback system to which “the usual transfer functions apply”. The only energy supplied to the system is in the initial conditions (and a trivial input to open the valve).
“Many confuse the latter with the former.”
Confusion probably goes back to the failure of climatology/earth sciences to adopt the standard concepts of feedback found in mathematical and engineering textbooks. The result appears to be no end of confusion over what different modes of feedback can do, stability conditions, and conservation of energy (which is the underpinning principle of systems analysis).
Jordan (14:14:32):
I fear that In your preoccupation with mechanical systems subject to the force of gravity and friction, you miss the entire point about the difference between feedback and recirculation (storage, if you wish) of system output in cases where there is no conversion of energy from kinetic to potential. Feedback in such cases requires a transducer at the output node and an external power source to replicate the output. That’s why the node is called a “pick-up point” in standard engineering terms. The output signal power is not affected by the transducer. In recirculation, the node is a branch-point, at which only a fraction of the available power goes out of the system. The complementary fraction is returned to the input for reuse. A common thermal example of recirculation is a boiler/radiator system. There’s a huge difference between the heat it produces and what would be obtained if the radiated heat were replicated and added to that from the boiler. That is more akin to the bogus feedback issue at hand than mechanical examples wherein the external (sic!) force of gravity acts as the power source.
“”” Slartibartfast (10:22:57) :
In classical feedback systems; the forward “gain block” (A) contains a power source that is the energy source that makes it all happen.
Good point. I’m not sure how universal that is, but in servo control, the command is the input, but the power system feeding the servomotor does the lion’s share of the work. This is also true for wrapping feedback loops around amplifiers.
I applaud Mr. Sowell’s take on the problem, as well as Dr. Latour’s. I’ve been pointing out that the climate model has been had a number of control-systems misnomers applied to it for some time now, but the climatologists still insist on borrowing heavily from the systems engineering lexicon without paying much attention to what the various terms actually mean.
“Positive feedback” in this context would mean an exponential increase in temperature, until something went nonlinear to the point where positive feedback was no longer occurring. “””
Why do you assume that positive feedback means an “exponential” increase in temperature.
In a sense, “positive feedback” means only that the feedback generates an additional signal at the input, which contains at least a component that has the effect of increasing the total input signal, and hence by inference, increases the output. But that is a far cry from asserting that the output must grow without limit; or at least until some non linearity or limit process stops the increase.
The term “exponential increase” is bandied about a bit too freely.
When water vapor captures LWIR emissions from the surface, thereby warming the atmosphere, and also warming the surface, as a result of the additional solar energy that arrives during that delay in the exodus of the LWIR; that is certainly a positive feedback effect. Water vapor can also have a negative feedback effect, in that it also can absorb insoming solar energy; thereby also warming the atmosphere; but in this case reducing the solar energy that reaches the ground; thereby resulting in a cooling. That is a negative feedback effect; even though the atmosphere was warmed (more) by the additional solar energy absorbed by the increased water vapor.
But neither effect is a runaway condition; that leads to warming without limit; well until the “power supply rails are reached.
However the phase change that occurs with water and not with other GHGs in earth atmosphere, kicks in a definite cooling of the surface (over climate time scales); and that is why the earth atmosphere condition cannot thermally runaway; like an out of control positive feedback loop with excessive gain.
I’m surprised to see that some people feel that warmer surface temperatures should lead to less outgoing LWIR emissions, rather than much more as suggested by the usual blackbody radiation principles.
Probably they are the same people who think that it is the polar regions that are cooling the earth.
No the polar regions are cold because they don’t get enough energy from everywhere else to be at a higher temperature; it is not because they are more efficient radiators. It is the tropical deserts (and UHIs) that are the efficient radiators that are really cooling the planet; because of the Stefan-Boltzmann 4th power law. And when you consider that the CO2 absorption band is on the long wavelength side of the surface LWIR spectrum; then we should consider the peak spectral emittance at the LWIR spectrum peak; because that moves further away from the CO2 band as the emission temperature increases. And in that case the peak spectral emittance increases as the fifth power of the temperature.
So hot things make effective energy radiators; cold ones do not.
So I’m not surprised that that weird looking satellite sees more LWIR from warmer regions.
sky (16:39:38) :
My preoccupation is with conservation of energy.
I have used examples of mechanical systems because it is easier to convey ideas using mechanical examples. It hope they will help others understand my misgivings about climate modelling. (More on that in a moment).
The feedback systems I have given you are not considered “bogus” in the engineering world – a world where feedback concepts are routinely put to good use, allowing you to take them for granted.
For example, my pendulum did not use gravity as its power source. The only energy source was the initial shove. Gavity only served as a mechanism to store energy in one part of the cycle. Same for the water tanks – the only energy input was the initial condition (the difference in the levels).
As George E Smith correctly says, the climate model has had a number of control-systems misnomers applied to it, and climatologists insist on borrowing heavily from the systems engineering lexicon without paying much attention to what the various terms actually mean.
I can assure you that anybody in climatology who wishes to carry out a formal mathematical analysis of feedback systems, the first thing they will do is to adopt the sensible convention of positive and negative feedback used by mathematicians and engineers.
With that in mind, I suspect that nobody has ever had the need to invoke positive feedback to model a passive oscillator (such as a pendulum).
You appear to be suggesting that the climate behaves as a passive oscillator. If so, an immediate consequence would be that the GCMs will then have been incorrectly formulated, if they use positive feedback mechanisms to explain the motion of the climate.
You mention cases where there is no conversion of energy from kinetic to potential. The movement of energy is kinda central to systems analysis, so I struggle to think of an example of a real-world system which doesn’t involve a transfer of energy from one form to another.
I never mentioned a transducer or “replication of output”. Methinks ’tis like a Red Herring.
Maybe I do miss the entire point about the difference between feedback and recirculation (storage, if you wish). On recirculation, I think you are straying into Pascal’s objection (if memory serves me correctly). This states that you cannot return heat from the cold reservoir to the hot reservoir of a practical machine and expect a net gain in output. The amount of energy you need to put in to do so will always exceed the amount you get back. It’s another way of saying that we cannot have perpetual motion.
So to leave you with a couple more examples.
Let’s say you wish to be lifted to a height of 20 metres. Do you climb inside a bucket and lift yourself up 20 metres using the handle? No you cannot. So what makes you think you can send energy back to the input of a passive system to achieve an analogous result?
Or, let’s say you’re watching your pendulum swiniging, and you guess that the air resistance is having an effect on the dynamics. So you put the pendulum into a vacuum chamber and set it swinging by giving it an initial shove. As you empty the chamber, what do you expect to see? Surely you don’t expect the absence of air in the chamber to result in the pendulum increasing the length of its swing! So what makes you think a change in the radiative transfer properties of the atmosphere can *amplify* a change in the temperature on the Earth? (To understand the question, be careful to understand what I mean by “amplify”.)
These are just some more examples of the reasons why I am far convinced about the proposition of positive feedback in a passive climate system.
John S. (09:50:15) :
“The lack of grasp of the fundamentals of complex system behavior is what keeps “climate science” in its primitive state.”
I applaud this statement. Complex (including chaotic and nonlinear) systems certainly do need to be understood better. The work by Tsonis such as “The Southern Oscillation as an Example of a Simple, Ordered Subsystem of a Complex Chaotic System” (1998) is a good start.
I’m not an engineer of physicist. But I get the feeling people are using the term “feedback” in different ways and causing confusion. Feedback in the important sense in a chaotic system is what is also called damping or dissipation or friction, and it is the key ingredient for non-equilibrium pattern formation.
Have a look at this video of spatio-temporal emergent patterns in vibrated corn-starch:
http://chaos.ph.utexas.edu/research/vibrated_cornstarch.htm
There are some cyclical phenomena visible: for instance the merging then separation of the persistent indentations from the blown holes. This event repeats in a cycle. At the end the protrusion of the fingers and the eventual collapse of each finger. This also has a timecourse.
Engineers such as John S state that climate being passive cannot generate its own temporal dynamics – or only as some simple derivative of a forcing frequency. But the cornstarch shows emergent cyclical events with timecourses very decoupled and many orders of magnitude longer than the frequency of the forcing vibration.
Tsonis treats the oceanic decadal oscillations as emergent spatio-temporal patterns in a chaotic-nonlinear system and this is the way forward for researching such systems, plus many other temporal and cyclical phenomena of atmosphere and ocean.
I agree completely that positive feedback in a passive system is out of the question. Given the stringent requirement that no power be diverted from the output signal in any feedback sytem governed by a transfer function with the Laplace-transform representation H(s)/[1 – G(s)H(s)], I don’t think that even negative feedback can be physically implemented without recourse to some external source of power to replicate the output signal. Replication of output in the loop is at the very heart of all genuine feedback.
Control system engineers are by no means shy in using a variety of configurations of sensors/transducers/servomechanisms/op-ams in the feedback loop to accomplish their goals (see, e.g., Fig 5.6 in Nise’s classic text.) Whether you recognize it or not, gravity is the external restoring force that powers the pendulum response. Without it, the pendulum would not swing in response to any forcing.
Finally, my reference to bogus feedback was not directed at your mechanical examples, but to the ungrounded notions prevalent in climate science. To see a system block-diagram with an op-amp in the foward loop–the simplest control system configuration–in a discussion of natural climate is highly perplexing, to say the very least. Op-amps are always independently powered. No one has come up with any credible explanation of what the corresponding natural mechanism would be. Let’s leave it at that.
Jordan (06:22:35):
I completely agree that positive feedback in a passive system is out of the question on energy conservation principles alone. Inasmuch as any feedback system governed by the transfer function H(s)/[1 – H(s)G(s)] relies upon replication of the undisturbed output signal in the feedback loop, I don’t think that even negative feedback is possible without an outside power source. Control system engineers are not shy at all in using independently powered sensors/transducers along with servomechanisms and op amps in the feedback loop to achieve their goals (see, e.g., Fig 5.6 in Nise’s classic text).
Gravity does supply external power in your mechanical example. Without it, the pendulum would not oscillate in response to applied nonoscillatory forcing. Please note, however, that my reference to bogus feedback was not aimed at your mechanical examples, but at the prevalent misuse of the term in climate science. Let’s leave it at that.
Thank’s for your constructive reply sky. Still a couple of point at issue.
Gravity only stores energy in the pendulum. It is the initial shove that gives it some kinetic energy, which then oscillates between kinetic at the lowest point in the swing, to potential energy at each extreme.
No shove, no movement, no oscillation.
“Inasmuch as any feedback system governed by the transfer function H(s)/[1 – H(s)G(s)] … I don’t think that even negative feedback is possible without an outside power source.”
I’ll give you an example of a simple series RC filter with input voltage V1(t) and output voltage V2(t).
The Laplace Transform current through the resistor: R.I(s) = V1(s) – V2(s)
The Laplace Transform of current at the capacitor I(s)= C.s.V2(s)
(on the initial condition, output voltage V2(t=0) = 0)
The negative feedback here is the reducing current (reducing potential across the resistor) as the capacitor charges or discharges to the try to achieve equilibrium at V2 = V1.
We can equate the two equations for I(s):
RC.s.V2(s) = V1(s) – V2(s)
which simplifes to a neat linear transfer function:
V2(s)/v1(s) = 1/(1+RC.s)
This has the form of the transfer function you mention. The change of sign from – to + in the denominator is becuase of the negative feedback.
It short, negative feedback exists without a power source in a passive system. The only source of energy to a passive system is the energy provided by the input. As the components of passive system dissipate energy (unconditionally in practical systems), a passive system will always produce a bounded output for a bounded input, consistent with conservation of energy.
Jordan (09:34:50):
My first comment yesterday (17:28:56) dissappeared from the thread, prompting me to resubmit an abridged version (18:24:38). It now mysteriously reappears! Sorry for the repetition.
I’m even more sorry, however, to see your persistent confusion about the role of internal and external variables in a system. Gravity is not a capacitance effect that “only stores energy in the pendulum.” It is a persistent external force that provides all the restorative power. If you look up the governing nonlinear differential equation in normal form, you’ll find it on the same side as the forcing function, but with opposite sign, in a classic example of negative feedback. Without it, the position of the pendulum would continue to change indefinitely in response to any forcing.
Your notion thata passive RC filter constitutes feedback, because the voltage across the resistor can be expressed in similar algebraic form as a feedback system transfer function, convinces me that fundamental features of system analysis continue to escape you, along with many others. It is the relationship between system output and input signals that the transfer function specifies, not some state of internal components. Passive RC filters have no effect whatsoever upon the input signal. Without such an effect–the sine qua non of feedback control–it makes no sense to speak of system “feedback.” I will not elaborate this any further.
sky
You may recall from high school physics how the potential energy of a body of mass ‘m’, at a height ‘h’ (from some datum) can be expressed as:
Q = m.g.h.
When I used this in our discussion on the above thread, I suppose it was a way of testing whether your knowlege of dynamics has progressed as far as high school physics. But you do insist on coming back here with more confused arguments.
It is not MY notion that a passive RC filter constitutes feedback. I was taught that as an undergraduate when I was studied for a batchelors degree in …. wait for it … control engineering!
When you mentioned the transfer function, I thought it might be worthwhile to spell out the negative feedback in that passive RC circuit. To make it easy for you I even said specifically how the changing output voltage will be in the direction which reduces further changing voltage. But you still come back to argue about it. I would not expect such an argument from a first year undergraduate engineer.
If you cannot take it from me, why not get yourself an undergraduate text book in control engineering. I thought my comments would help, but you seem intent on making an argument whatever I say.
My only request would be to spare the community on this site any more of your confused notions of feedback theory. Because I would feel the need to reply to misinformation.
[snip – dial it down and then resubmit if you wish]
Jordan:
The issue is not what you were “taught,” but what feedback systems actually do. They operate–by definition– not just on the input signal, as with the RC filter, but upon the output signal, as well. That’s why the complete transfer fuction has both H(s) for the forward system function and G(s) for that in the feedback loop. There simply is no feeback of the filtered signal in the classic RC configuration. Unlike with feedback, the input signal left unaltered.
As for elementary physics, take pendulum out to space an give it a shove. You’ll find that in the absence of any other external force, it will not oscillate, but simply revolve around the pivot point indefinitely.
Wanting to enjoy my weekend, I’ll leave you with a final thought: condescension is never a good substitute for clarity of analysis.
sky
I wanted to show you how the net response of the RC circuit is the combination of the input and the output. The transfer function I gave encapsulates those internal interactions.
It should be enough to convince you that negative feedback exists in passive systems. We don’t need to discuss it further here.
Take a pendulum out to space and we have turned one of the parameters down to zero. After a shove, we expect to get circular motion (rather than an arc) as the limit of reducing the gravity parameter to zero. The physical system hasn’t changed and the mathematical model shouldn’t need to be changed.
We could call gravity an external force if it was useful to do so. But it is analogous to calling the electrical field across a capacitor an external force. What would be the point? It seems more parsimonious to get straight to the point – these can be represented as energy storage devices.
I’m sorry if I was condeseding. But undergraduate engineers would have points deducted for suggesting there is no feedback in a simple RC circuit. With that in mind, it doesn’t do WUWT any good to have such ideas being aired here.