Bart: Cloud feedback is negative – ocean response is around 4.88 years

Posted: September 11, 2011 by Rog Tallbloke in atmosphere, climate, Ocean dynamics

I’m not sufficiently up on stats to really understand this, but I’m collating the relevant comments from Climate Audit and Roy Spencers site here, because it looks important, and moderation at CA seems to be impeding the flow of the conversation so it has become disjointed.

Is there anyone with a voice in these matters who could just do a cross spectral estimation of the two series and read off the phase relationship directly?

You cannot diagnose feedback by performing a linear regression on a phase plane plot when the driving input is all over the place and the phase response is nonlinear. As I explained on the last thread, and have written up here.

What I have found is significantly greater evidence that the feedback relationship is, indeed, negative, contrary to Dessler. The phase shift at low frequencies, which determines the sign of the feedback, is very clearly near 180 degrees. Saying an input is 180 degrees from the output is a long winded way of saying that the one is the negative of the other…I’m sitting back and watching people argue about techniques which are entirely unsuited to the problem in the first place. It’s insane.

 I need other knowledgeable people to become aware of the result and determine what the effect of a -9.5 W/m^2/degC feedback would be. I think, though, that the IPCC models require the overall feedback to be positive to get any significant warming, so it appears to me that this result could, potentially, kill CAGW.

There could be a valid criticism that the span of data is too short to have high confidence in the result. But, I would argue that the onus is on them to prove that the overall feedback is positive, because I think this establishes that the running assumption should be that it is likely negative.

.

If you have an input x(t) and an output y(t), the cross spectrum is the expected value of X’(f)*Y(f), where X(f) and Y(f) are the Fourier transforms and “‘” denotes the complex conjugate. Since Y(f) = X(f)*H(f), where H(f) is the transfer function, the cross spectrum is C(f) = P(f)*H(f), where P(f) is the power spectral density of x(t). Since P(f) is a zero phase, positive real function, the phase response of C(f) is identical to that of H(f). If it is near 180 degrees at low frequency, then you’ve got negative feedback. If near zero, it is positive.

The dispersion in the phase plane makes it very difficult to diagnose the sign of the feedback – the nonlinear phase characteristic of H(f) ensures that you will get a messy cloud unless you focus on a particular frequency range and filter everything else out. For sure, the cross spectrum will be polluted with other processes and noise, too, but I expect it should be a pretty good indicator of the dominant process occurring, if there is one.

Caveat: estimation of the cross spectrum is not as easy as multiplying the FFTs of the two data streams. The outcome of that will be highly erratic and of questionable value. A good method is to do that, then inverse transform to get the impulse response. Window that impulse response with a good window (e.g., Hamming or other) at a truncated limit before it visibly starts breaking up, then transform the windowed impulse response function back to get the estimate of C(f). There are, of course, other reasonable methods. A mathematical analysis tool like MATLAB has its own built-in functions for doing it.

Well, here are some preliminary results. I hope I used the right quantities. I pulled this data from the Spencer link. I used column 9 for the HADCRUT3 temperature anomaly, and I used column 5 minus column 8 for the cloud response. The relationship between these variables most assuredly has a negative dc gain.

Here is a plot of the estimated frequency response. If this holds up, I think it’s going to shock some people. The response at high frequency is a jumble, and probably due to independent processes going on. But, the low frequency region is dominated by a fairly well defined 2nd order response with natural frequency of about 0.0725 year^-1 and a damping ratio of about 0.45, which indicates a time constant of about 4.88 years. Yes, years.

The time constant is 4.88 years, meaning the step response is 63% of the way to the final value in that time. Convergence is exponential in the standard, usual way.

The impulse response is:

I am talking about the time it takes for clouds to react to temperature changes. That is, if you increase global temperature by 1 degC (and could hold it there), within 4.88 years, you will be 1-exp(-1) = 63% of the way to creating an opposing 9.5 to 10 W/m^2 reduction in insolation. The unit step response is plotted below

I have posted the transfer function and impulse response estimates here and here, respectively, created using artificially generated data according to the prescription here.

These show it is, indeed, possible to pull out the long term correlation using only 124 monthly data points. It doesn’t always work, but sometimes, it does.

It would be very nice to come up with a more robust deconvolution scheme which works the majority of the time, and reanalyze the data to prove the veracity of the estimated functions beyond doubt, and I encourage any interested parties to look into this.

Here is the MATLAB code used to compute all this:

% Import monthly data
temp = data(:,9);
dR = data(:,5)-data(:,8);
N = length(dR);

% Sample period
T = 1/12; % years

% Pad time series with zeroes to prevent time aliasing of impulse response
Nsamp = 8192;
Npad = Nsamp-N;
X = fft([temp;zeros(Npad,1)]);
Y = fft([dR;zeros(Npad,1)]);

% Compute impulse Response
h = real(ifft(Y./X))/T;

% Window Impulse Response
Nc = Nsamp/2^2;
w = [ones(Nc/2,1);(1 + cos(pi*(0:(Nc/2-1))'/(Nc/2-1)))/2];
w = [w;zeros(Nsamp-Nc,1)];
hw = h.*w;

% Plot smoothed impulse response
c = [1:15 15:-1:1]/(15*16);
figure(1)
hs = flipud(filter(c,1,flipud(h)));
t = (0:(length(hs)-1))*T;
plot(t,hs)
grid
xlim([0 50])
title(‘Cloud-Temperature System Smoothed Impulse Response’)
xlabel(‘time (years)’)
ylabel(‘W/m^2/^oC/year’)

% Compute Frequency Response and plot
H = T*fft(hw);
f = (0:(length(H)-1))’/Nsamp/T;

% Create 2nd order model
s = sqrt(-1)*2*pi*f;
w0 = 2*pi*0.0725;
zeta=0.45;
Hmod = -9.5./((s/w0+2*zeta).*(s/w0)+1);

figure(2)
subplot(211)
loglog(f,abs([H Hmod]),’LineWidth’,2)
grid
xlim([1e-3 5e-1])
title(‘Cloud-Temperature System Magnitude Response’)
ylabel(‘W/m^2/^oC’)
xlabel(‘frequency (years^-^1)’)
legend(‘From Data’,’Model’,’Location’,’SouthWest’)
subplot(212)
semilogx(f,faz([H Hmod]),’LineWidth’,2)
grid
xlim([1e-3 5e-1])
title(‘Cloud-Temperature System Phase Response’)
ylabel(‘deg’)
xlabel(‘frequency (years^-^1)’)

  • Nick Stokes Posted Sep 10, 2011 at 5:32 PM | Permalink

    Bart,
    I translated your code to R, up to the impulse response plot. I get the same result.

    But if I apply a Hanning taper to temp and dR (down to zero at each end of the data window) it is very different. I think that suggests that the finite data length is having a big effect.

    I’ve put up a html page here which has some explanation, the impulse response with and without Hanning, and the R code.

  • Bart
    Posted Sep 10, 2011 at 8:23 PM | Permalink

    Nick – yes, it can change things a little through loss of resolution, but it isn’t really all that significant and, importantly, the feedback is still negative (180 deg phase shift at low frequencies).

    I have other responses which may be of interest to you and which are directed to your queries but, unfortunately, the moderator seems to be taking the weekend off. Your insights are cogent, but I think I feel relatively confident now in asserting that this response is real, and the feedback is, clearly, negative. But, please, continue to investigate.

  • Bart
    Posted Sep 10, 2011 at 8:32 PM | Permalink

    “You have to accept low frequency problems with finite data…”

    Actually, Nick, abrupt transitions are a high frequency phenomenon. Multiplication in the time domain is the same as convolution in the frequency domain, so the tapering tends to lower the resolution of the frequency response estimate. Which is good for the higher frequency portion of the spectrum, but not so good for the low. Yes, a truncated data record is bad for resolution of low frequency stuff. But, providing a taper only really helps the high frequency portion. And, so, your impulse response so estimated has the slight ringing smoothed out.

  • Bart
    Posted Sep 10, 2011 at 8:58 PM | Permalink

    I’m going to try one more time to get through the blocker on this. Here is a discrete time model for generating data with the desired correlation which you can play with. I assume the input is white noise. You should find that, occasionally, it is possible to generate a short time series for which the analysis works, other times, not. Which is why I said: “My preliminary conclusion right now is that the Spencer data just happens to be lucky.”

    a = [1.000000000000000 -1.967462610776618 0.968691947164695];
    b = -[0.617926899846966 0.611409488230977]*1e-2;
    temp=randn(10000,1);
    dR = filter(b,a,temp);
    temp = temp((10000-123):10000);
    dR = dR((10000-123):10000);

    I generate a lot of data before truncating to be sure of eliminating transient start up effects.

    Bart says:

    neill says:
    September 10, 2011 at 9:33 pm

    “Bart, if you would be so kind as to translate these findings for the less mathy/scientific among us, it would be much appreciated.”

    Basically, clouds exert a negative feedback of about -9.5 W/m^2 per degree C of increased temperature (and, vice versa, +9.5 W/m^2 per degree C of decreased temperature). Which means they contribute at least partly to a climate thermostat mechanism such as Willis Eschenbach has proposed here at WUWT from time to time.

     

Comments
  1. Paul Vaughan says:

    “I’m sitting back and watching people argue about techniques which are entirely unsuited to the problem in the first place. It”

    I’ll agree on that! Bart is using some oldschool methods that don’t ask (let alone answer!) the questions of interest (to me) at all.

    I have NO interest whatsoever in the code (& oldschool methods) of others. All I need:

    Links to:
    a) data on a plaintext webpage.
    b) concise description of variables.

  2. tallbloke says:

    It doesn’t matter if techniques are old skool or new skool if they get the job done correctly.

    If you can use .csv files they are here:
    http://www.climateaudit.info/data/spencer and http://www.climateaudit.info/data/dessler

    Bart’s summary of which column’s were used in the analysis are here:

    “Well, here are some preliminary results. I hope I used the right quantities. I pulled this data from the Spencer link. I used column 9 for the HADCRUT3 temperature anomaly, and I used column 5 minus column 8 for the cloud response.”

    I have seen links to CERES data floating around over the last few days. Whether they are plain text I don’t know. Hadcrut 3 data is available as plain text from woodfortrees.org

  3. Paul Vaughan says:

    (a) answer:
    1. http://www.climateaudit.info/data/spencer/flux.csv
    2. http://www.climateaudit.info/data/dessler/dessler_2010.csv

    Now how about (b)?

    And I’ll add:
    c) Which variables from (a) [by name]?

    [Reply] What was that you were saying to Svalgaard about attitudes towards volunteers again?

  4. tallbloke says:

    A question asked at Bishop Hill and my response:

    “Bart’s analysis identifies a 5 year lag between cloud changes and temperature changes, and agrees this is surprising. If true, then how should this be interpreted relative to the inverse proportionality between sensitivity and response time?”

    Between El Nino events, heat builds up in the pacific warm pool, most of the energy is therefore hidden from the surface temperature record. The approximate period of the ENSO cycle is around 5 years.

    The last 3 solar cycles were all around 10 years long. It is noticeable that the big El Nino’s occur just after solar minimum. I have a theory why….

  5. Paul Vaughan says:

    Tallbloke, it’s reasonable to request links to data & metadata. It’s up to volunteers, of course, to decide whether to respond. I’m still hoping (not demanding) that someone (doesn’t matter who) will point to specific variables [by name, to avoid ambiguity] & metadata [to supply context, since data without sufficient context cannot be properly interpreted]. I literally don’t have time to dig through the long, messy WUWT & Climate Audit threads on this subject, but if some volunteer responds to my request, I’ll volunteer my skills in exchange to explore Bart’s claim. Maybe you can persuade Bart himself to drop by. Since there appears to have been a misunderstanding, I apologize. Your hosting & moderation efforts are appreciated.

  6. tallbloke says:

    Hi Paul, thanks for that. I only have the column numbers Bart supplied for now. Sleepy time for me.

  7. Nick Stokes says:

    Paul,
    The data is here. Col 9 is temp, col 5 is all sky radiation (CERES) and col 8 is clear sky. Bart calls [5]-[8] dR, and gets a transfer function relative to temp. That’s where it all comes from.

    I have a page here. But I’m more dubious now. Comments around here.

    BTW, Dessler says (giving good reasons) that dR is the wrong way to get cloud radiance. He uses reanalysis.

  8. Paul Vaughan says:

    Thanks Gentlemen.

    Calculations are easy. That’s no problem. But I’ll have to dig around for more info on context, as this is a new area for me. Not sure if the effort is worth it, but I trust Bill Illis (what finally got my attention on this stuff that falls outside my usual climate interests since this whole branch of climate science [i.e. anything mentioning "W/(m^2)"] smells of VERY bad assumptions).

    One last question:
    Where’s the “Read Me” file that goes with the data? Anyone know? A link to that would be invaluable.

    Best Regards.

  9. Paul Vaughan says:

    [snip]

    [Reply] Paul, I don’t know if Steve McIntyre would be happy to have an unsubstantiated comment like that on the original thread. I’m not permitting it here. See my response to your comment on WUWT here

  10. Steven Mosher says:

    Somebody is new to our process of working together.

    tallbloke when you guys are ready I have hourly data ready and waiting. Incoming solar at 206 locations.. in case you have any theories about Forbush
    ..

    Feb 11, 2011.

    predictions gents?

  11. cementafriend says:

    Not sure if it is relevent here but has anyone thought that you can not just add or average anomalies with clouds because of time lags. In Australia the weather moves from west to east because of prevailing winds. It takes between 2 to 7 days for high or low pressure cells or cold fronts to move across the continent.

  12. Philip says:

    tallbloke @ September 11, 2011 at 5:17 pm

    That’s very interesting – thank you very much for your reply.

    “Between El Nino events, heat builds up in the pacific warm pool, most of the energy is therefore hidden from the surface temperature record. The approximate period of the ENSO cycle is around 5 years.”

    Do you think this therefore corresponds to the S term in SB10 eqn 4?

    “The last 3 solar cycles were all around 10 years long. It is noticeable that the big El Nino’s occur just after solar minimum. I have a theory why …”

    I can see what you mean for ’98 and ’10, but earlier than that…?

    http://www.climate4you.com/Sun.htm#Global%20temperature%20and%20sunspot%20number

  13. Steven Mosher says:

    cement.. Its worse that that

  14. tallbloke says:

    Hi Philip, and welcome. Take a look at this post which has some hastily chopped up graphics at the comment linked below which try to show what I mean about ENSO’s timing in relation to the solar cycle:

    http://tallbloke.wordpress.com/2011/06/12/the-timing-of-el-nino-in-relation-to-the-solar-cycle/#comment-6798

    The continuous record is at the top of the thread.
    My hypothesis is that during the ascendding phase of the solar cycle, the ocean mixed layer goes into ‘heat recieving mode as the sun strengthens and clouds diminish. Then just after the peak the sequestered energy starts to re-emerge, often with a small el nino just after solar peak, followed by the big one near minimum or just after. Other factors come into play, so it doesn’t run like clockwork, but I think there is a general principle at work here.

  15. manicbeancounter says:

    Might I ask a potentially dumb question, to help me put Bart’s calculations in context?

    Bart says
    “I need other knowledgeable people to become aware of the result and determine what the effect of a -9.5 W/m^2/degC feedback would be. I think, though, that the IPCC models require the overall feedback to be positive to get any significant warming, so it appears to me that this result could, potentially, kill CAGW.”

    Given the IPCC estimated that a 40% increase in CO2 from 1750 to 2005 (from about 280 to 390ppm) gave a warming of 3 W/m, does the effect of -9.5 W/m^2/degC
    (a) reduce the global warming effect of a doubling of CO2 (e.g. the Lindzen and Choi 2009 postulated that it would half the impact to 0.5 degC?
    (b) so counteract the impact as to make increasing CO2 levels reduce temperature? The “Cloud-Temperature System Step Response graph seems to suggest this.
    (c) the long-term impact of increases in CO2 on temperatures (>20 years) is about zero?

  16. Brian Hall says:

    manic;
    (c) And based on the non-response of the globe’s temperature to the accelerating CO2 output of the last decade, also for <20 years.

  17. tallbloke says:

    Joe Born says: Posted Sep 12, 2011 at 8:11 AM
    To horn in on the Skywalker-Talbloke colloquy about making Bart understandable, two suggestions.

    First, in supplying the translation, it would be helpful to explain why the response variable is not the whole-sky radiance rather than the cloud radiance. After all, isn’t the target the response of the net insolation of the earth as a whole to the average temperature of the earth as a whole?

    Second, to finesse around an explanation of Bart’s “statistics,” (actually, his use of discrete Fourier transforms, impulse and step responses, etc.), one might just say this:

    “It is uncontroversial that, if one observes the complete response (e.g., electrical current flowing as a function of time) of a linear system (such as an unknown network of resistors and capacitors) to a known stimulus (voltage impressed across that network as a function of time), one can accurately infer from it what that system’s response will be to some other stimulus (e.g., to voltage applied as some different function of time).

    “What Bart is doing is simply applying the universally accepted technique for doing so to determining, from the response (net insolation as a function of time over the past ten years) to the known stimulus (average temperature as a function of time over that ten years) to determine what the system’s response would to a step change increase temperature–and he concludes that the net insolation would fall.

    “There are a couple of complications. First, unlike the resistor-capacitor network, Bart’s system (the climate) is not linear, so application of his technique to this particular problem would not be completely uncontroversial. Still, this first complication is not a subject of much if any discussion, because linear-systems techniques are widely applied in estimating responses to small perturbations in non-linear systems–with the tacit recognition that the results can’t be completely accurate.

    “The other complication–which is the subject of the discussions–is that the technique Bart uses theoretically requires an infinite record of both the known stimulus and its response if one is to use them to compute the response to some other stimulus with complete accuracy. Of course, no one ever has an infinite record, but one can come close enough with a long-enough finite record. The disputants recognize, however, that the ten-year record is not long enough to avoid significant inaccuracies. So their discussion concerns whether it nonetheless admits of confidence in Bart’s basic conclusion: that a temperature increase tends to decrease net insolation: feedback in this system is negative”

    My first comment no doubt betrays appalling ignorance of the subject on my part, but I am hopeful that the second comment will be helpful to a significant subset of us unwashed masses who typically just watch from the sidelines.

  18. tallbloke says:

    If my own hypothesis is correct, insolation into the ocean is much more important than levels of longwave radiative flux in the atmosphere, because there is no mechanism for rapidly entraining significant heat from the atmospheric radiation into the ocean mixed layer, and the ocean drives the atmospheric temperature due to its vastly higher heat capacity.

    This is evidenced by the ~3 month lag of air temperature behind sea surface temperature. It is also evidenced by the relationship between insolation, albedo, and outgoing longwave radiation you have been working around.

    So I think Bart is correctly saying that reflection of incoming solar is the most important cloud effect.

    The determination of a -9.5W/m^2 cloud feedback plus the empirically measured drop in low cloud over the tropics from 1980-1998 pretty much nails the global warming issue in my opinion. Co2 is along for the ride.

  19. lateintheday says:

    Hi tallbloke,
    “So I think Bart is correctly saying that reflection of incoming solar is the most important cloud effect.”
    I’m back with another stupid question I’m afraid. Its often reported that there has been a satellite measured decrease in outgoing radiation specifically at the CO2 absorption wavelength. If I follow correctly, you’re saying (to some extent) that during cloudy conditions, proportionately more energy is reflected back to space without having been through the mill of our lower atmosphere. Is this/can this be related in any way? Perhaps a relationship might be present if one were to assume some biological cause for this reduction in specific outgoing radiation wavelength?

  20. tallbloke says:

    Water vapour spectrum overlaps the co2 spectrum. An increase surface temp leads to more phosynthesis and biomass. This leads to more co2 and less vater vapour through transpiration because plants need less water if they have more co2.

    Cloudiness fraction depends on a range of things. Temperature to some extent, but also GCR’s according to Kirby and Svensmark, marine surface biology creating aerosols to cause cloud condensation nuclei to form, protecting the plankton from UV.

    Outgoing longwave radiation (OLR) is not very accurately measured. Probably only to within an order of magnitude of any alleged change due to co2. I suspect the reports might be relying on models…

    Here’s a graph of OLR from the NCEP reanalysis:

    I think this indicates that heat accumulates in the oceans on multidecadal timescales, due to ongoing high levels of solar activity and reduced cloud cover.

  21. lateintheday says:

    Thanks for that. The link shows that OLR remains historically high despite the long drawn out solar minimum. Is this the ‘indication’ that you refer to, since one might otherwise expect OLR to drop around solar min.

    Also, stumbled over this NCAR/UCAR prediction from 2006 which you may find amusing.

    http://preview.tinyurl.com/6h66otj

  22. tallbloke says:

    LitD: The ongoing high OLR relates to energy being released from the ocean following solar minimum. This is in line with the big 2010 el nino and my hypothesis that the ocean goes into heat release mode when the Sun goes quiet. See the reply I gave to Philip above.

    Dikpati’s SS24 forecast was a spectacular failure. Svalgaard was on the Panel and said it would be much smaller, but got outvoted.

  23. lateintheday says:

    thanks again.

  24. Brian Hall says:

    tallbloke: An increase surface temp leads to more phosynthesis and biomass. This leads to more co2 and less vater vapour ”
    Say what? More photosynthesis absorbs and thus decreases CO2, and transpires more water vapour. Sounds like the bass is ackwards!

  25. tallbloke says:

    It’s complex. There’s been an overall increase in biomass. This means more co2 being absorbed, but also released by rotting down of dead plant material. Plants with more co2 available need less water.

  26. diogenes says:

    mmm… it seems as if Bart has had enough….Nick and the consensus guys have a deal of work to rescue their position, and no amount of complex math will work because it looks as if they do not know the position they are arguing against…..the future looks interesting

  27. diogenes says:

    to be honest, I do not think that Bart is actually saying anything. He is just demonstrating a tool-kit that the climatologists seem to be unaware of. I suspect that the data he is using it on is too smoothed and averaged to give a meaningful result from this analysis.

  28. tallbloke says:

    I’m inclined to agree. For one thing, we have empirical evidence that tropical low cloud cover was reduced between ~1980 and 1998, while surface temperature increased due to less albedo letting more Sun to the surface. Clearly, cloudiness averages across the latitudes are not going to tell us much about how it all works.

  29. diogenes says:

    tallbloke…which gets us back to Paul Vaughan at the start…if you look at Nick Stokes’s maps of monthly temperature anomalies, you can see just how localised the warming/cooling is…it is quite amazing. Any approach to climate that does not take this localisation into account is just a load of round objects. as is the concept of teleconnected trees

  30. Stephen Wilde says:

    In order to deal with the ‘problem’ that global cloudiness decreased during the warming spell/period of more active sun it is necessary to recognise three things:

    i) That a reduction in cloudiness during a period of more active sun introduces a positive system feedback by allowing more solar shortwave into the oceans.

    ii) Something else has to happen to negate that positive feedback from cloud reduction.

    ii) We have observed a consequent poleward shift of the climate zones with more poleward/zonal jets. That is what alters the rate of energy transfer from surface to space and that is the negative feedback. Its power comes from the water cycle and the energy transfer capabilities of the phase changes of water.

    Shifting the air circulation systems latitudinally has two contradictory effects:

    i) Altering energy flow into the oceans by moving the cloud bands latitudinally

    ii) Altering the speed with which energy is transferred from surface to space by moving the climate zones latitudinally.

    The first is a positive feedback and the second is a negative feedback and the second always limits the first with the climate price to be paid for the necessary adjustment being a change in the relative sizes and positions of all the permanent climate zones. The sign of i) and ii) can each be reversed, thus:

    i) A reduction in cloudiness for a warming effect will be countered by a poleward shift increasing energy flow to space

    ii) An increase in cloudiness for a cooling effect will be countered by an equatorward shift decreasing energy flow to space.

    Note that it is NOT a zero sum process because the natural flow of energy from warm Earth to cold space is dominant so whereas scenario i) will limit warming of the system very effectively scenario ii) will only restrain cooling for a short while because if energy doesn’t get into the system in the first place slowing down the energy flow to space is only a short term solution.

    That is all that climate change amounts to. Anything that tries to warm or cool the system is countered by a change in the rate of energy transfer to space which alters the relative sizes and positions of the permanent climate zones. All observed climate changes fit that description.

    The underlying equilibrium is set by atmospheric pressure which in turn sets the energy cost of evaporation by fixing the amount of energy required to achieve the phase change of water from liquid to vapour.

    Much more detail here:

    http://www.irishweatheronline.com/news/environment/climate-news/wilde-weather/setting-and-maintaining-of-earth%E2%80%99s-equilibrium-temperature/18931.html

    as regards the setting of the basic equilibrium temperature and here:

    http://www.irishweatheronline.com/news/environment/climate-news/wilde-weather/feature-how-the-sun-could-control-earths-temperature/290.html

    as to how the solar changes seek to alter the equilibrium temperature (but largerly fail as per the so called ‘faint sun’ paradox).

  31. tallbloke says:

    Diogenes says
    Any approach to climate that does not take this localisation into account is just a load of round objects.

    Stephen’s comment addresses that in the large. At the local level, Willis has an interesting post on cloud forcing up today on WUWT

    http://wattsupwiththat.com/2011/09/15/cloud-radiation-forcing-in-the-tao-dataset/

    Bart has done some valuable work on this, showing the overall cloud feedback is negative, taking all temperature and radiation variations into account, albeit averaged globally. There are some wrinkles to be ironed out though, and I’m discussing these with Nick Stokes on the Moyhu blog.

    http://moyhu.blogspot.com/2011/09/greens-functions-and-desslers-data.html

  32. P.G. Sharrow says:

    @Stephen Wilde says:
    September 16, 2011 at 7:50 am
    “The underlying equilibrium is set by atmospheric pressure which in turn sets the energy cost of evaporation by fixing the amount of energy required to achieve the phase change of water from liquid to vapour.”

    Very good discription of the energy pump of the atmosphere. To the above nugget I would add to “vapor” “and back.”. At higher altitudes well as higher latitudes. You are very right about stratification by latitude as well as altitude depending on temperature vapor pressures caused by changes of energy levels. pg

  33. Stephen Wilde says:

    Thanks pg.

    There are three critical locations:

    i) The surface position of the mid latitude jets which responds to the equator/pole energy flux.

    ii) The height of the stratopause (not the tropopause) which responds to the net effect of the sun on the atmosphere.

    iii) The depth of the ocean skin which responds to the balance between the energy cost of evaporation and the solar shortwave input.

    All three are infinitely variable and acting together maintain the Earth’s temperature equilibrium.

    It is a three dimensional balancing act.

    At the base of it all is atmospheric pressure and the level of solar input to the oceans. We have three means of regulation in respect of two variables.

    Unless the atmospheric pressure is changed by GHG quantities (there is no significant effect on pressure) there can be no change to the equilibrium temperature merely a change to the speed of energy throughput from surface to space.

    So in the end the level of solar input to the oceans is the sole variable as regards equilibrium temperature. And that is intimately connected to cloudiness changes.

    Cloudiness changes are intimately connected to the speed of the water cycle and so we come full circle because the speed of the water cycle is dictated by i) above.

    i) above holds the balance between ii) and iii) above.

    That is what keeps the oceans liquid and maintains a stable enough Earth system long enough for life to develop.

    GHG variations are of no account because i) above controls the balance. Change GHG quantities and one just gets a miniscule adjustment to i) above.

    Changes in solar energy input to the oceans and changes in energy release from oceans to air make huge changes to i) in comparison.

  34. Bart says:

    TB – I see you are posting at Nick’s place. I don’t have a google or whatever account, so it will not let me post. My final comment in the CA thread is here. Nick has totally gone off the rails.

  35. tallbloke says:

    Hi Bart,
    The conversation between you and Nick is mostly over my head, but I did find out why Nick gets a positive cloud feedback with Dessler’s data. By accident, when I was replicating the Spencer CLD RF to compare to Dessler’s Eradr, I subtracted col 6 from 5 instead of 8 from 5 in Spencer’s data (Steve McIntyres flux.csv). This produced a curve almost identical to Dessler’s, with a positive slope. But col 6 was just the LW flux, not the net of the LW and SW fluxes. This shows that Dessler is using model data which hopelessly underestimates the degree to which cloud reflects incoming solar SW.

    No wonder the warmista cling to the idea that cloud feedback is positive. I’ll finish the graphs and put up a post, along with some real empirical data on the increase in cloud over the last decade from Project Earthshine (Palle et al) and ISCCP.

    Thanks for all the effort you’ve put into this, and I hope the ‘wading through treacle’ feeling you get when entering the climate debate doesn’t make you quit. It’s hard going, but it does have some rewarding moments of levity and mirth. :)

    By the way, if you really can’t resist taking Nick on his own turf, just start a free wordpress blog, and use that ID to validate yourself on his site. I met Nick in person at the Lisbon Conference in January, and although we don’t agree about climate science, he’s a nice guy really.

  36. Nick Stokes says:

    Bart,
    I checked, it will let you post, I think. Just look down the bottom of that list of Id options. You can offer a name, or be totally Anon.

  37. A. C. Osborn says:

    What do you think of this interesting analysis?

    http://landshape.org/enm/fft-of-tsi-and-global-temperature/

  38. tallbloke says:

    I think I left a comment on it 3 days ago. ;)

  39. P.G. Sharrow says:

    I just commented on the “landshape.org/enm/fft-of-tsi-and-global-temperature” post,

    The primary mover of energy from ocean to space is water evaporation and high altitude condensation or freezing and not radiation / conduction through temperature difference.

    Hopefully enlightenment may follow. 8-) pg

  40. Brian H says:

    diogenes;
    Are you disputing treelepathy? Despite all the original groundbreaking work done by Dr. Tolkien? For shame!
    >:p

  41. A. C. Osborn says:

    Rog, that was really aimed at Nick and Paul, but you were correct I hadn’t looked at the comments. LOL.

  42. A. C. Osborn says:

    And Steve of course.

  43. Stephen Wilde says:

    Hi, ACO.

    I’m inclined to agree with Bart, especially this:

    “That is, global temperature is not only a function of the magnitude of solar anomaly, but also its duration. Small but persistent insolation above the solar constant can change global temperature over extended periods.”

    I think I did mention a point of disagreement about part of Bart’s interpretations somewhere.

    I think it was about the cloud cover actually reducing while the sun was active and increasing while the sun was inactive so it was necessary to reverse a sign somewhere in the climate description.

    I think the upshot was that the negative system response is actually the latitudinal shifting of the air circulation systems with a resultant change in the speed of the water cycle rather than the cloudiness changes in themselves.

    The cloudiness changes actually cause a positive feedback which is then cancelled by the air circulation changes.

    That is because a more active sun pulls the jets poleward for MORE energy into the oceans due to DECREASED cloudiness when the sun is more active but in pulling the jets poleward the speed of the water cycle increases.

    The thing is that the observed climate changes are NOT a consequence of a temperature change for the system as a whole but rather a consequence of the climate zones shifting in relation to the geographical locations below them.

  44. Tenuc says:

    Stephen Wilde says:
    September 19, 2011 at 4:43 pm
    “…The thing is that the observed climate changes are NOT a consequence of a temperature change for the system as a whole but rather a consequence of the climate zones shifting in relation to the geographical locations below them.”

    Good summary of why climate is so difficult to understand. The energy transfer systems also change in efficiency as they move position on the global. Different topography like mountain ranges, the proportion of water to land and other features can impact results. Add in the factors which alter climate systems dependant on energy levels and the problem becomes orders of magnitude more difficult to understand. Use of averages and long-term trends are of no help to understanding our non-linear climate system.

  45. Lord Beaverbrook says:

    http://wattsupwiththat.com/2011/09/20/new-peer-reviewed-paper-clouds-have-large-negative-feedback-cooling-effect-on-earths-radiation-budget/#more-47761

    the cloud radiative cooling effect through reflection of short wave radiation is found to dominate over the long wave heating effect, resulting in a net cooling of the climate system of −21 Wm−2.

  46. P. Solar says:

    Despite Bart’s reticence for diagnosing the feedback, some hand tweaking of parameters to Spencer’s simple model with random inputs lead me to very similar results (before Bart posted his plots).

    http://tinypic.com/view.php?pic=30sfupc&s=7

    My aim was just to see if I could get Spencer’s model to produce something with a form close to the satellite data he had published. Like I said, it was ad hoc but the result came surprisingly close.

    What I have only just noticed is that the parameters I fitted give a result for the time constant so close to Bart’s method it’s remarkable.

    45/9.2 = 4.891 years.

    The result did not vary much as long as the ratio was the same , so I’m not necessarily suggesting 9.2 is the correct feedback. But again it’s very close to the value of 9.5 at the top of this thread.

  47. P. Solar says:

    TB says: “But col 6 was just the LW flux, not the net of the LW and SW fluxes. This shows that Dessler is using model data which hopelessly underestimates the degree to which cloud reflects incoming solar SW.”

    This probably explains his rather odd assertion that “all components” of TOA will give an accurate feedback.

    Funny thing, after the world record acceptance of this paper, it still has not actually been published it seems.

  48. P. Solar says:

    “The dispersion in the phase plane makes it very difficult to diagnose the sign of the feedback ”

    I agree that fitting simple ols to a scatter plot is plain silly unless you have a decent correlation. As much as anything else you are throwing away the time series information for both data set which is the key feature of the data and what ties them together.

    What I think is useful is looking at the variation of correlation with lag. Since ols is nothing but a scaled up correlation the two are equivalent for this purpose.

    In that context when you see a small correlation coeff you take it to mean there is not much correlation, not to pretend there is a small “slope”.