Why Climate Feedbacks Cannot be Regional

March 30th, 2013 by Roy W. Spencer, Ph. D.

Whenever I see reference to the regional nature of climate feedbacks, I cringe.

I will admit that what happens on a regional basis determines net global climate feedbacks, but feedbacks cannot be evaluated regionally. Feedbacks only make sense in the global average.

First, a summary of what climate feedbacks are, by definition: In response to a surface temperature change, other changes in the climate system (clouds, etc.) can either magnify (positive feedback) or reduce (negative feedback) the original temperature change. The single largest feedback is negative: the increase in infrared energy lost to space as temperature increases. This so-called “Planck effect” is what stabilizes the climate system against runaway change.

Cloud feedbacks are generally considered to be the most uncertain, and could be positive or negative (I believe they are negative). The contribution to water vapor feedback by the atmospheric boundary layer is almost certainly positive, but the free-tropospheric contribution to water vapor feedback is much more uncertain, since it depends upon microphysical processes within precipitation systems which are the source of free tropospheric air.

(And for those who object to the use of “feedback” in a climate context, sorry. Until a better term comes along which better reflects the recursive nature of the forcing-response process, we are stuck with it.)

So, why can’t feedbacks be evaluated regionally? Because a change in one region will, in general, affect other regions, through changes in atmospheric vertical circulation systems.

For example, if the Pacific warm pool was to warm, we might expect increases in clouds and precipitation there. But those changes are the result of increased rising air over the warm pool, and that extra rising air must — through mass continuity — be exactly matched by increased sinking air away from the warm pool…possibly thousands of miles away.

In fact, since in the tropics the areal extent of (weakly) sinking air is so much greater than that of the (strongly) rising air, the feedback response to a warming of the warm pool can be dominated by what happens thousands of miles away from the warm pool. This is why the original “thermostat hypothesis” of Ramanathan and Collins (1991) was widely criticized as too simplistic.

Feedbacks can only be evaluated over entire vertical circulation systems, and since these systems are interconnected around the world without clear boundaries, feedbacks really only make sense in the global average.

Now, it might well be that the feedback response is different for different kinds of forcing, or it might be that the net feedback varies over time as the climate system evolves in response to a forcing. My only point is that it doesn’t really make sense to talk about “regional feedbacks”, unless you know that the regional change has not affected vertical circulation systems that extend outside the region of interest. Sure, you can compute a number for the change in the regional radiative budget in response to a temperature change, but it would be incorrect to call that number the “feedback response” to the temperature change.


38 Responses to “Why Climate Feedbacks Cannot be Regional”

Toggle Trackbacks

  1. Thanks, Dr. Spencer, very interesting article.

    • Johan Couder says:

      I second that. In a few simple words Dr. Spencer always manages to express some profound insights.

  2. A. Physicist says:

    I disagree with your characterization of feedbacks. Inherent in their definition is the notion of a timescale, and you must be precise here if you wish to distinguish between regional or global feedbacks. This seems especially important for clouds, for which regional differences (especially land vs. ocean) completely define their properties through aerosol indirect effects. Since cloud timescales are quite short, these will indeed result in different regional warming over continental or ocean regions. Besides, the only evidence you need to see that feedbacks _are_ inherently regional on transient timescales is the contrast between observed warming over land versus ocean.

  3. jack mosevich says:

    Does Venus’s atmosphere represent an extreme case of feedback vis. CO2 , Nitrogen, clouds, sulphuric acid? There must be a huge positive feedback to keep it so hot.

    • Nullius in Verba says:

      No. Venus is hot because the average altitude of infrared emission to outer space is so high. Sunlight is absorbed by the cloud tops and re-radiated as infra-red, which keeps the cloud tops at about the same temperature as Earth. But the clouds are 50-80 km above the surface, and since temperature changes about 8 C/km due to the pressure change, the surface is much hotter. On Earth emission occurs from only 5 km up, so it’s a lot cooler.

      Feedback refers to the response to a *change* in temperature. It’s what happens when the temperature changes some other variable (like clouds or ice cover) that itself has an effect on temperature. On Venus, changing the temperature wouldn’t change the cloudiness (always 100%) or the constitution of the atmosphere (there being no oceans to act as a source for different gases) so it isn’t obvious that there are any feedbacks. If temperature affected the heights of the clouds that might work. I don’t know. But it’s not needed to explain the high temperatures.

  4. Massimo PORZIO says:

    @Dikran Marsupial

    Uhmmm… I don’t know what happened to my previous message, maybe I used such kind of “bad characters” in it, because It has a couple of instead of the references I did to your statements, and it is missing the final part.

    Anyways, I repeat it here below:

    “you need physics, not statistics.”
    Yes, I agree. My one was a rhetorical question, It’s a long time I ask for physics, which is a true science. Statistic instead could be science, but it should be taken with the velvet gloves. Too many times statistic has been used to prove anything and nothing about the some thing.

    “That CO2 is a greenhouse gas is widely accepted by pretty much everybody, as Dr Spencer says (on his global warming 101):

  5. Gordon Robertson says:

    Roy…you explained once that positive feedback in climate science is a not-so-negative, negative feedback. That is not the same as a positive feedback in physics, which requires amplification.

    There is no way to take a feedback signal, other than in thought experiments, and use it to amplify anything. Positive feedbacks are not amplifiers, they require an external power source to do the amplifying. Climate scientists need to focus on exactly what causes any increase in heat and get away from inferring that feedbacks cause it. Feedbacks are used in climate science as a wrapper around complex processes which are not clearly understood.

    A perfect example of classical PF is a public address system where an input signal from a microphone is amplified with an amplifier that gets it’s power externally, from a wall socket, a generator, or a battery. Amplification in electronics is the control of a larger current source by a smaller input current or voltage.

    No signal is amplified per se, rather the smaller input signal is transferred to the output via a transistor (or other transducer) that has the ability to control a larger current from an external power source by means of the smaller input signal. The output signal is not the same current as the input signal, it is a reflection of it, often inverted. By putting a load in the output circuit, the load experiences an amplified version of the input signal, but it does not experience the input signal current or voltage.

    The positive feedback signal that results in a feedback squeal occurs because the feedback signal is in phase with the input signal and makes it incrementally larger.

    That alone cannot cause amplification, it must go through the amplifier, and when the combined signal is amplified, the output signal is larger. A sample of that larger signal gets back to the microphone acoustically, off walls, floors, and ceilings, and increases the input signal, when in phase, to an even higher level. Eventually, the output signal runs away and produces the squeal.

    (Since you are a musician, it might interest you to know that equalizers depress feedback squeals by attenuating the frequencies at which feedback tend to occur in a narrow band. All rooms have at least one resonant frequency at which feedbacks occur. By notching those frequencies, the feedback can be controlled, or eliminated)

    Could we do that in the atmosphere? :-)

    That process is not possible in the atmosphere. Clausius developed the 2nd law to prevent such perpetual motion, which was allowed in some instances by the 1st law, which is only about energy in general. Carnot hypothesized there were no losses in heat engines but he was wrong, hence the 2nd law.

    What many people don’t get is that the 2nd law is about heat, not energy per se. People are confusing infrared energy with heat, and they are not related. IR is electromagnetic energy whereas heat is a measure of the average energy in atoms and molecules. IR contains no heat, just as light contains no colour, the colour being added by the human eye as it responds to various frequencies in the light spectrum.

    I can buy your arguments vis a vis feedbacks as redefined by climate science, provided they are variations of a negative feedback system. Systems in the atmosphere may approach unity amplification but they will never exceed it.

    The anthropogenic theory has done just that, however. They have hypothesized a positive feedback from rare gases like ACO2 that can add to the surface heat provided by solar energy. Not only does that contravene the 2nd law, it represents perpetual motion. You cannot take heat from a surface warmed by solar energy, that subsequently warms GHGs in the atmosphere, and recycle that heat to raise the temperature of the surface.

    To do that, an external source of heat would be required plus a mechanism to do the amplification. All clouds can do is participate in a negative feedback system by manipulating the distribution of heat. They cannot amplify heat beyond unity.

    Remember that heat is a measure of the average energy of air molecules, and a measure of the degree to which the molecules become excited. How are you going to amplify that average energy without adding an external form of energy to increase the heat?

    Furthermore, if you have GHGs at a cooler temperature than the surface, how will infrared from that cooler source increase the average energy of surface heat? The 2nd law says it can’t. In fact, Clausius said it in words, in his treatise on mechanical energy. He stated clearly that infrared energy can be exchanged between bodies of different temperatures but that heat can only be transferred from the warmer body to the cooler body, under normal means.

  6. Gordon Robertson says:

    Massimo…”Too many times statistic has been used to prove anything and nothing about the some thing”.

    A perfect example is the UAH satellite temperature record graph on Roy’s site. Alarmists use the linear trend from 1979 – 2013 as proof that warming has occurred over the period, yet John Christy points out there has been little or no warming during that entire period.

    Roy has added a running average which tells a far different story than the linear trend. By visual inspection, using the area under the curve of the running average, John Christy’s comments become apparent. The first 19 years of the record fall largely in a negative temperature anomaly region and even NOAA admit on their site that negative anomalies indicate temperatures below a certain average.

    In this case, the average is the 1981 – 2010 average, and it is apparent that for 2/3ds of the record, temperatures were below that baseline. Even after the 1998 El Nino, which seems to have left a 0.2 C residual, post 1998, the trend has been fairly flat, hovering around that 0.2 C.

    Therefore, the linear trend tells us nothing, even though alarmists cling to it as proof that the satellite record is in line with the surface record and models.

    What I would like to see is a graphical record of absolute temperatures rather than anomalies.

  7. Massimo PORZIO says:

    Sorry Gordon,
    my full post was for an another thread. I had some problem posting it and inadvertently I posted this here.
    Anyways I agree with you.

    But Dr.Spencer always said that those polynomial trends were just a joke.

  8. Massimo PORZIO says:

    My “But Dr.Spencer always said that those polynomial trends were just a joke.” is just for all those who complained when he placed the polynomial trend on the graph and those who complained because he removed it :)

  9. RW says:

    Roy,

    Can you be more specific as to what you mean by ‘regional’?

  10. Gordon Robertson says:

    Massimo…just to clarify, I was not talking about the polynomials. There is a red curve running through the graph that is the running average. I find it far more meaningful than a linear trend that bridges a negative and positive anomaly region with 2/3ds of it in the negative region.

    What I got from the polynomials is that the global average tends back toward zero, not to a hockey stick, a la Mann. If you plot global warming on an absolute scale, it looks like a wavering line near zero C.

    See Figure 5 at this link:

    http://www.ianschumacher.com/global_warming.html

    I think Roy’s polynomials suggested that.

    If Mann had used the sat temps for 1990, he’d have gotten an Irish shillelagh rather than a hockey stick, although the shaft would have been quite warped by the LIA and MWP.

    http://en.wikipedia.org/wiki/Shillelagh_%28club%29

    • Massimo PORZIO says:

      Hi Gortdon,
      yes, I know what Mann did “playing” with the different methods to perform the running averages at the ends of his “dendroclimatology” data.
      I had to deal with the very same problem when few years ago I designed an optical spectrum analyzer with some degrees of smoothing functions to filter away the photo multiplier noise in the UV measurement region.
      In my opinion, no one should use filtered (averaged) data to predict where the curve goes outside of its limits.
      Averaging works well for filtering incoherent noise superimposed to the signal. But if you don’t want to “invent” the curve, you must know that the nature of the noise is not really coherent with the signal. Even the Aqua AMSU Ch5 experience demonstrated one more time that if you have an unknown superimposed signal, there is no way to completely remove it from the measurement.

      “I think Roy

  11. Massimo PORZIO says:

    Grrr… I lost part of the previous post again!
    It has been truncated where I copied a part of the Gordon post I was referring to.
    Now is time to have dinner here in Italy, I’ll return later.

  12. Massimo PORZIO says:

    Ok, back from dinner, I hope this message will be correct without truncation.

    Gordon,
    about your “I think Roy’s polynomials suggested that”

    I’m really critic about the use of polynomials to predict the tendency of a data curve outside its start/end limits.
    Some years ago, I implemented a polynomial regression function to compute the normalization curve of the sensor in a RF power meter design. What I learnt from that experience is that the polynomial regression is very reliable between the start/end source data limits, but it is absolutely unreliable outside them, even if the regression is done to very high orders (my function was designed to show the results for a polynomial of order from the 1st to the 21st).

    Note also, that I used it for the frequency normalization of a sensor which behavior wasn’t so chaotic such as the data handled in the climate arena. So, in my opinion, maybe that between the measured points of a so chaotic data array, such as the temperature record, even an high order polynomial could fail to approximate the reality between those measured points.
    Anyways, in my personal opinion, polynomial regressions of higher orders are not better than the linear regression in predicting where a curve goes.
    That is, they fail at all.

    About the “Mann’s hockey stick” it wasn’t only the way he used to handle the running average at the end edge that give that shape. As far I know, he also applied the infamous “trick” of mixing “dendro proxies” with thermometers temperature data overlapping them in an almost arbitrary way.

    I repeat myself again, statistic is a science which should be used with the velvet gloves.

    • Gordon Robertson says:

      Massimo…I agree with you about statistics. I don’t relate the noise in electronic equipment to the noise in climate issues. In electronics, we know the source of the noise for the most part. It is essentially impossible to remove shot noise from signals since it is due to the electrons that represent the signal colliding naturally with other electrons and atomic nucleii. All we can do is minimize it using various cancellation techniques.

      In climate science, one man’s noise is another man’s signal. Rahmstorf et al tried to remove the El Nino signal from the temperature record, calling it noise. I don’t regard EN influenced temperatures as noise if EN is a prime driver.

      Rahmstorf is such an alarmist that he has a religious faith in radiative processes as being the sole drivers of atmospheric temperatures. He thinks that removing the EN signal, which is essentially impossible to do, will leave a pure anthropogenic signal. Even Trenberth complained that the El Nino signal was hiding the anthropogenic signal.

      I regard both of them as deniers. The sat record has shown a low sensitivity between anthropogenic CO2 and atmospheric temperatures, over 33 years, if one exists at all.

      I did not regard Roy’s polynomial as anything more than a curiosity. I found it kind of neat how the polynomial seemed to replicate what was happening. I think that’s how he intended it.

      • Massimo PORZIO says:

        Ok, I get what you mean.
        I apologize, but sometime I misunderstand the English language and maybe I say exactly what you say but I don’t realize it until I reread the post.

  13. Paul_K says:

    Dr Spencer,
    I disagree quite profoundly with your argument here.

    I think that with a very small shift in your mental model, you might find it perfectly reasonable and possibly essential to talk about regional feedbacks.

    Assume an areal discretisation of the Earth. The surface average temperature (SAT) is a linear combination of the discretized surface temperature field. The local feedbacks at any point within that field may be approximated as either a simple function of the local point temperature or for some feedbacks as a function of local plus surrounding regional temperatures. Additionally, the variation in temperatures will change the meridonial fluxes over time – both atmospheric and oceanic. So any regional model which seeks locally to balance net radiative flux also has to include sensible heat flux into and out of that local discretized regional element. This latter becomes the ultimate control on the local temperature change with time.

    Your main objection appears to be

    So, why can’t feedbacks be evaluated regionally? Because a change in one region will, in general, affect other regions, through changes in atmospheric vertical circulation systems.

    With the above conceptual model, I hope you agree that your main objection is overcome. It is perfectly reasonable to examine regional feedbacks, I believe. I would go further and say that it is essential to examine regional feedbacks. There is currently a clear mismatch between analyses based on a zero dimension global average linear feedback vs the curvilinear net flux response with change in SAT observed in almost all of the AOGCMs. For the AOGCM results tested, the climate sensitivity required to match the model results over the instrumental period is about half the ultimate sensitivity reported for those models. There is strong evidence to suggest that this curvilinear behavior derives directly from the regional variation in feedbacks and response times observed in the models. A continued reliance on global average data masks this huge source of variation in sensitivity between models as well as between models and reality.

    • ferd berple says:

      We know that that variance of the sum = sum of the variances + 2 * sum of the covariance.

      I take this to mean that we can calculated the global variance from regional variance, but only if we know the covariance between the regions.

      Or, conversely, that if we calculate both the global and regional variance, then this would allow us to calculate the covariance between the regions.

      However, without knowing the co-variance, the regional variance cannot be used to calculate the global variance.

      In other words, regional variance is equal to global variance only in the case where there is no covariance between the regions. Which is not true for climate.

      http://en.wikipedia.org/wiki/Variance#Properties.2C_formal

  14. Nabil Swedan says:

    “For example, if the Pacific warm pool was to warm, we might expect increases in clouds and precipitation there. But those changes are the result of increased rising air over the warm pool, and that extra rising air must — through mass continuity — be exactly matched by increased sinking air away from the warm pool…possibly thousands of miles away.”

    Dr. Spencer, does this mean that at the global level feed backs are equalized?

  15. Slabadang says:

    For once I think your wrong Roy!

    Everyone has been fooled to try to find a canstant f

  16. Slabadang says:

    For once I think your wrong Roy!

    Everyone has been fooled to try to find a canstant f

  17. Svend Ferdinandsen says:

    “So, why can

  18. Svend Ferdinandsen says:

    “So, why can

  19. Svend Ferdinandsen says:

    So, why can

  20. Svend Ferdinandsen says:

    I hope it works now. The former trials were cut to only thre words.

    I would say that regional feedbacks must exist, like for instance the albedo from snow and ice in the polar regions.
    It is anyway an explanation you often hear.
    I am not sure if the devil is hidden in the word evaluate, which is not in the headline.

    • ferd berple says:

      regional feedbacks exists. however, without a measure of covariance they tell us nothing about global feedbacks. Dr Spencer’s argument is mathematically correct in this sense.

      since it is likely that determining regional covariance is a (much) harder problem than determining global feedback, using regional feedback to determine global feedback is likely a computational dead end.

  21. George White says:

    By considering clouds and ice as feedback mechanisms, you have to acknowledge that local feedback matters (at least hemisphere specific) since the average cloud coverage, surface ice and surface temperatures for the 2 hemispheres are quite different (the Northern hemisphere has 64% average clouds and 15% ice, while the Southern hemisphere has 69% average clouds and 10% ice, per ISCCP weather satellite data, see http://www.palisad.com/co2/ca.png and http://www.palisad.com/co2/st.png and http://www.palisad.com/co2/ia.png ). Since there is little, if any, **net** energy exchange between the hemispheres, they can be modeled as independent entities and the results summed. As such, they have independent net feedback (f), open loop (Go) and closed loop (Gc) gain terms where per Bode, 1/Go = 1/Gc + f. Consensus climate science erroneously assumes that Go is 1 and since Gc is greater than 1 (about 1.61), 37.5% net positive feedback is required to fit the data given the error.

    Interestingly enough, the Gc terms for the 2 hemispheres are approximately the same and if you didn’t know that Go was not unity and was different per hemisphere, you might conclude that the feedback terms for the 2 hemispheres were the same.

    Note that the above temperature plot shows a flat temperature profile over the last 3 decades. The raw ISCCP data has a serious calibration error around 2002 resulting from the assumption of continuous polar orbiter coverage which I have corrected. The raw global temperature data is here http://www.palisad.com/co2/sat/st_raw.png

  22. Thanks for your blog loaded with so quite a few information. Stopping by your blog helped me to have what I was looking for.

  23. RW says:

    I would also add that despite significant topographical differences, each hemisphere still exhibits the same fundamental feedback signature as the other:

    http://www.palisad.com/co2/plots/wbg/sh/gain.png
    http://www.palisad.com/co2/plots/wbg/nh/gain.png

    How can this be reconciled if you’re claiming there is enough energy flow between the hemispheres to ‘contaminate’ each as measure of global feedback if summed together? I don’t quite understand that.

  24. RW says:

    George,

    You mention the calibration error in the satellite data from around 2001 that you’ve have referred to prior here:

    http://www.palisad.com/co2/eb/sat.html

    and specifically:

    http://www.palisad.com/co2/eb/temp.gif

    Can you elaborate further on the specifics of this error? I mentioned to Roy that just by eye it looks like the error is in his UAH Global Average temperature plot.

  25. RW says:

    Roy,

    One other thing, if you cannot accept separate but summed hemispheric responses as a measure of global net feedback, how can you accept Dick Lindzen’s extrapolation of tropical feedback to the whole globe? That seems inconsistent to me, or maybe you don’t agree with Dick on this?

    • RW says:

      Also,

      Isn’t part of Dick’s rationale for doing this that energy primarily flows from the tropics to the poles and not the other way around?

  26. George White says:

    Yes, the closed loop gains of the 2 hemispheres are approximately the same, however; both the open loop gain and feedback fraction are different.

    My suspicion is that the behavior of clouds equalizes the gain by varying feedback and that there are other factors which dictate what the gain must be.

    From the temperature dependence of clouds,

    http://www.palisad.com/co2/gf/st_ca.png

    There is an inflection point at 0C, where the effects of incremental clouds changes because clouds are no longer more reflective than the surface, even as water vapor increases monotonically linear. There is another inflection point at around 300C, where increasing temperature starts to dramatically increase GHG absorption owing to exponentially increasing water content.

    http://www.palisad.com/co2/gf/st_wc.png

    The adaptive response of cloud cover consequential to changing characteristics of the system is a clear indication that clouds are the control variable in a feedback system controlling the planets energy balance. Note how this differs from the consensus view of a feedback model controlling the surface temperature.

    Its easy to see how the additional constraint is radiative balance when you consider that the radiation emitted by the planet is the cloud fraction weighted sum of cold cloud emissions and warm surface emissions. Only one fraction of cloud cover will work since the effect of clouds on albedo and on power the returned to the surface vary at different rates, relative to the surface temperatures.

  27. George White says:

    RW,

    The method used by Rossow to cross calibrate the satellites relied on polar orbiters to establish the baseline calibration for the geosynchronous satellite imagery and was incapable of properly handling the case of discontinuous polar orbiter coverage. Most of the time there are 2 polar orbiters and when one drops out or is replaced, the new one can be cross calibrated against the other.

    In September 2001, NOAA-14 was the only polar orbiter and in October 2001, it was replacer with NOAA-16 and there was nothing to calibrate NOAA-16 against, moreover; starting with NOAA-15, they upgraded the sensors resulting in significantly different response characteristics making the calibration error somewhat non linear with an average 3C error.

    At the time, I believe this was reported as the result of no planes flying after 9/11 and fewer aerosols, but that will not explain the almost instantaneous 3C temperature rise which did not go away after planes started flying again. BTW, this is the primary reason why the ISCCP data has a bad rap and can not be used for detecting trends (at least not easily …).

    • RW says:

      That’s interesting. To me it looks more like about +2C. None the less, at least by eye, it looks like there is a shift up in temperature at that same time in Roy’s UAH. As I interpret Roy, he already thinks UAH accounts for any such sensor drifts, etc.

  28. Doug Cotton (DJC) says:

    Roy and others

    Real world data does in fact indicate that regional climate (in a typical city area, for example) is statistically correlated with levels of precipitation.

    I have carried out a study which proves this point. Furthermore, the regions with greater rainfall had statistically significant lower daily maximum and minimum temperatures.

    In all of your article there is no actual study of temperature records for various regions. In contrast, there is such a study in the Appendix of my “Planetary Core and Surface Temperatures.”

    I understand where you are coming from, Roy, when you write “Because a change in one region will, in general, affect other regions, through changes in atmospheric vertical circulation systems.”

    However, it is the mean thermal gradient in the atmosphere above a region which affects its mean temperature over a reasonable period – usually a month or two is sufficient. We know that water vapour reduces this gradient, even though we probably disagree regarding the process – you saying it is the release of latent heat, and I saying it is mostly due to inter-molecular radiation – as explained in the above paper. A less steep gradient dictates a lower supported surface temperature.