Wednesday
Aug282013
by 
Bishop Hill
Bishop Hill Masters and Dessler do climate sensitivity
Aug 28, 2013 Climate: sensitivity I chanced upon Andrew Dessler's video on climate sensitivity a few weeks ago.
At the time I sent it to Nic Lewis to see what he made of it, but I think he was away at the time and it slipped my mind. This morning, however, I was reminded of it by this response from Troy Masters. I would say that the case for low climate sensitivity is still looking pretty strong.
Reader Comments (61)
The higher the sensitivity, the colder it would now be without AnthroGHGs. Pick a sensitivity that frightens you, and calculate how much colder the Earth would now be without the input of human endeavour.
=================
simply ignoring everything that doesn't support a position you've already taken. AKA Pot calling Kettle Black
Bish
Please explain what makes sensitivity models plausible when other climate models have been seen as totally unreliable?
I am simply applled by the notion of feedback used by so called climate scientists,
http://judithcurry.com/2011/10/10/climate-control-theory-feedback-does-it-make-sense/
Not only does Dessler (and Spencer) use a totally incorrect mathematical formulation of feedback - what they are defining mathematically is damping, the statistical methods that they apply are shakey to say the least.
On another issue, I see that your favorite troll, ZedDeadHead, has re-emerged. What is happening, is that there are an increasing number of estimates, better derived from observations, of climate sensitivity that are at the low end of the "theoretical" distributions. Your comments simply show your lack of scientific education.
Via an link on Annans blog from your link I saw this paper:
“The phase relation between atmospheric carbon dioxide and global temperature” Humlum et al."
Which concluded that the Mauna Loa CO2 must come from the ocean. Annan of course dismissed it merely because it would mean that he is wasting his life chasing a fools errand. There was a more fulsome comment though here:
http://www.sciencedirect.com/science/article/pii/S0921818113001562
which focussed on "logical deficiencies". Now arguing logic is something that is not done enough in this debate so here is my logical reply to theirs:
1) what could be the sink for fossil fuel CO2 emissions, if neither the atmosphere nor the ocean plays a role?
Answer: Well logically the land! In fact the idea that the ocean is a net carbon sink is contradictory to Henrys law and is only a postulation due to the lack of any other ideas for a sink. Hence the idea that the ocean is a source rather than a sink agrees with established physics/chemistry. Moreover there is still a "missing sink" in the IPCC carbon accounting. The paper just proposes a bigger missing sink than that already accepted by the IPCC. As it turns out those who have bothered to look for land sinks have generally found them.
http://www.nature.com/climate/2007/0708/full/climate.2007.35.html
http://www.engineering.iastate.edu/eri/files/2012/06/EcosystemsDesertScienceVol320June08.pdf
2) What is the alternative explanation for ocean acidification if the ocean is a net source of CO2 to the atmosphere?
Answer: This argumentation is circular logic. Ocean acidification is a postulated consequence based on the unphysical idea that the ocean is a net sink. We have only a minute amount of data supporting such an idea. Many studies though have shown that ocean acidity varies wildly and that much greater set of obs would be needed to weed out any spatial bias.
3) The authors have obviously ignored the reconstructed and directly measured carbon isotopic trends of atmCO2 (both d13C, and radiocarbon dilution) and the declining O2/N2 ratio, although these parameters provide solid evidence that fossil fuel combustion is the major source of atmCO2 increase throughout the Industrial Era.
Answer: The logical conclusion of such a study is that nature somehow rejects anthropogenic CO2 which, being around 3% of the total CO2 flux in nature, is an absurd idea. The IPCC even contradicted these carbon ratio studies by proposing the "overflowing bathtub" analogy whereby mans emissions are indeed absorbed by nature but the total is just too much for nature to cope with. This was also illogical but less so than the idea of nature somehow separating out and rejecting manmade CO2. Therefore by logic these ratio studies must be flawed. As it turns out the error bars in the studies are larger than the signal so they cannot be used as evidence of anything.
So yes we should reconsider the entire field of climate science. The logical conclusion is that vast amounts of money are being wasted on a non-issue. that computer models cannot predict the last 17 years of non-warming and that the stratosphere has not cooled either since 1995 give full support to the idea that climate science has led us all the wrong way, presumably because we paid them to find a problem so they found it. If we had paid for a natural explanation for this miniscule rise of 0.6 degrees/century they would likely have found that instead. The previous solar theory was of course the dominant "consensus" throughout human history including through previous CO2 scares by Arhennius and Callendar both of whom saw a rising trend and thought they knew the cause - yet both were confounded by a fall in temperatures.
I am currently involved in nuclear energy so I have no dog in the fight between believers and non-believers, just a long experience of "end of the world" popular fantasies starting with the predictions of a coming ice age in the 70's. Personally I'd prefer warming to cooling though - along with most of planetary life.
"chief climate change advisor to WWF"
Wouldn't it save money if they just looked out of the window..?
Dung:
That question isn't directed to me and is to my mind too imprecise to be very useful. However, I'd like to pick up on the apparent premise. Is it true that all climate models are totally unreliable?
If the premise was that GCMs are unreliable in their attempt to model regional rainfall, say, then yes, I think from everything I've picked up that's fair enough. But what about Guy Callendar's model for global average temperature (anomaly) that Steve McIntyre recently rediscovered?
That's a climate model - albeit a simple one - and it turns out it does better than much more complex GCMs with this iconic number (whatever its real-world importance or otherwise). I don't think it would be fair to call Callendar's model totally unreliable but there are no rule books defining the 'totally' or the 'reliable' that I know of.
The important thing to grasp is that climate model is a very general concept. You have to judge each case on its merits. I'll leave the rest to the Bish and others if they wish to say more.
RC Saumarez & Richard Drake
A far more erudite post than I could hope to emulate
was made by Rhoda in a previous thread on Climate Sensitivity. Rhoda asked how the effect of CO2 could be isolated from all the other factors when we do not even understand what all of those factors are. If the climate has any sensitivity to CO2 at current levels then ALL other factors must be known.
Mr Saumarez could you possibly illuminate this blog and indeed the whole world by describing those other factors? Please do not insult us by using the phrase "Natural Variability".
Dung: Thanks for mentioning me but aren't you shifting the goalposts here? Do you accept my point about climate models, having considered Guy Callendar's 1938 model? If you wanted to redraft your question for the Bish as a result that would be grand. I'm not up for a general debate of what Rhoda said on a completely different thread, that you haven't linked to, on this one. One step at a time. If necessary baby steps.
Richard Drake
Not even Steve McIntyre suggested that Callendar's 1938 model was a good model.
Do stop making patronising comments?
Dung: It wasn't patronising, unless I was patronising myself. Baby steps is often what I find I need in order to arrive at deeper understanding. Meanwhile you've changed the goalposts again in what you've said about Steve. The issue I chose to examine, based on your question to the Bish, was whether all climate models are totally unreliable. There's a measurable gap between that and good. But I hope it has helped you - or at least some reader - to think again about Callendar's amazing 1938 effort as a climate model. Some models are certainly better than others.
As I have said before "climate sensitivity" seems to be a notion from "climate science", based on the same old stuff (radiative forcing etc), but which even many climate sceptics seem to accept as having some sort of meaning and being worth trying to calculate.
But, to adapt a phrase coined recently by Lord Monckton, is there any reason to believe that computing climate sensitivity is anything other than numerical masturbation, blinding its practitioners to the truth?
Ha, how did Monckton give you that idea? URLs please.
Richard Drake
Callendar's model might seem to work better or at least as well as vastly more complex GCMs but it isn't very good really. See, for example, how it completely misses the current standstill in warming. One of the guys who commented on climate audit made quite a full dissection of the fluctuations in global temperature that the model missed. I am not sure that the correlation between the model and actual temperatures would be very good.
diogenes: Sure, but is it 'totally unreliable', in Dung's original phrase? Not as much as the GCMs it seems. That was my only point here. Models in themselves aren't the problem. Getting them wrong is.
http://wattsupwiththat.com/2013/08/27/the-200-months-of-the-pause/
"That, in a single hefty paragraph, is why the models are doing such a spectacularly awful job of predicting global temperature – which is surely their key objective. They are not fit for their purpose. They are mere digital masturbation, and have made their operators blind to the truth. The modelers should be de-funded. Or perhaps paid in accordance with the accuracy of their predictions. Sum due to date: $0.00."
Tossers.
@Dung.
I'm not sure that I understand your question. What I am saying is that the idea of feedback used widely in climate science is not mathematically correct. If you take a system and describe it by a first order differential equation, as done by both Spencer and Dessler in their spat a couple of years ago,you are not describing feedback in the sense that there is a state that is altered by an inpout which then "feedbacks" to modify the response to an input. A first order ODE is characteristic of a damped system. The effect of water vapour that may, or may not increase temeperature by virtue of its "greenhouse" effect is not a feedback but an amplification of the initial radiant energy into the system. Uncontrolled this would be expected to produce a run away increase in temperature. If you hypothesise that this causes clouds to form and the latent heat of condensation allows energy to escape from the atmosphere inyo space, this is a transport phenomenon.
If, on the other hand, you hypothesise that cloud formation reduces the radiation striking the oceans and so limits the its effects, this is feedback as stated by Spencer and rejected by Dessler. The problem is that this model, which is stated by both Dessler and Spencer in a hopelessly naive way, is not reflected in their mathematics since a feedback system such as this must conform (at least) to a second order ODE. Given this, the statistical methods they use to quantify feedback is wrong.
Feedback is a widely used term in the climate verbosphere but I have yet to see a proper mathematical analysis of feedback in system terms. You might argue that a systems approach is too naive to capture the complexities of climate, but are incorporated implicitly as subsystems of more complex models. I would suggest that a systems approach used by Spencer and Dessler, in which they use a single, real number to represent feedback is both naive and wrong. If you follow the link in the post above, and if you read the attached pdf, you will see these arguments expanded in a very simple way.
Howver, the point is that feedback is bandied around and is used incorrectly. If their mathematical description is incorrect, then their concepts may well be incorrect.
As regards other factors which you wish me to define, without invoking natural variability, I would would merely comment that a mathematical model is device to enable one to understand something about a system and data derived from it. If that model is incomplete, or incorrectly parameterised, it is unlikely to predict future results.
RC Saumarez
I was responding to this comment you made about climate sensitivity models
@Dung.
Non-uniform priors!
I stand in awe of your knowledge of statistics and its terminology, perhaps you could now wow me with your ability to communicate with non statisticians?
Dr Saumarez -
You said "If you take a system and describe it by a first order differential equation, as done by both Spencer and Dessler in their spat a couple of years ago,you are not describing feedback in the sense that there is a state that is altered by an inpout which then "feedbacks" to modify the response to an input. A first order ODE is characteristic of a damped system."
I have noticed you say similar things before and (I think) you have said that, if a system involves feedback, then it is described by, at least, a second order differential equation.
This is something I don't follow. For example, in electronics, an op-amp with a capcitor between its output and its inverting input, and with the input signal applied via a resistor (Like this) is an integrator, described by a 1st order differential equation* - and there is no question that it is a feedback system.
How to resolve the discrepancy between your "2nd order" statements and my comment?
Regards
___________________________________________________________________________________
*[Of course, I understand that, in reality, op amps have their own dynamics and that no physical capacitor is a lumped parameter element described by a 1st order differential equation. But in the frequency ranges of interest the 1st order equation describes the system with great accuracy.]
I like it simple. I reckon the climate sensitivity (to CO2) is 1.0. If not why haven't we boiled dry sometime in the last few billion years? If it's not 1.0 then maybe something else regulates it to keep it 1.0.
Hearing the term "feedbacks" makes my toes curl.
Dessler is a climate/enviro politician first. climate scientist, distant second.
He was science advisor to Al Gore, the Al Sharpton of climate.
As to the paper itself: This paper is not really very different from the series of climate promotion papers that has heat hiding in the upper troposophere, deep in the oceans, or simply hiding the (inconvenient) declines.
It serves as a nice bit of sciencey propoganda to keep the faithful in line.
Dessler is a climate/enviro politician first. climate scientist, distant second.
Aug 28, 2013 at 9:31 PM lurker, passing through laughing
Yes.
Another Dessler video whose link was given at the end of the one above was titled: "Should we listen to the 97% or the 3%?"
The article concludes: "I would say that the case for low climate sensitivity is still looking pretty strong."
Many would say that is an understatement, given that 33 years of satellite data shows absoulutely no measurable relationship between CO2 emissions and temperatures. During the satellite era, their temperature measuremenst do not show any linear increase in temperature on a decadel scale, but instead a one off solitary climate shift (temperature shift) in and around the Super El Nino of 1998. Unless that El Nino event was in some way caused by CO2 (and I do not understand that anyone alleges such), then according to satellite temperature data, Climate Sensitivity is so low that it is unmeasurable.
Further, of the other temperature data sets, there has been no warming for approximately 17 years (in some data sets even slightly longer). What is the chance that natural variation these past 17 years or so has operated with a downward forcing exactly equal and oposite to the positive forcing of CO2, if CO2 in real world conditions (not the laboratory) actually results in warming? This is highly implausible and the longer the temperature anomaly stasis continues the more implausible this becomes.
These two facts put together, which are based upon observation, as opposed to some theoretical model, strongly suggest that to the extent that Climate Senitivity exists, it is so small that we cannot measure it within the sensitivity, resolution and inherent errors of our present day measuring equipment.
As I have repeatedly commented, personally, I consider that it is logically impossible to assess Climate Sensitivity until absolutely everything there is to know and understand about natural variation is known and understood; in particular each and every forcing that goes into making up natural variation is identified and understood including the upper and lower bounds of each and every constituent forcing within the natural variation umbrella. It is only once you can eliminate natural variation being the cause of any temperature change from year to year or from decade to decade that you may be left with the signal of temperature changes brought about by CO2 (or some other manmade activity). Until we fully understand natural variation, we cannot begin to seperate the CO2 signal, if any, from the noise of natural variation.
In my opinion, these studies investigating/pontificating upon Climate Sensitivity are tauntamount to fraudulent. They are disengenuous in the extreme and show a blatant disregard for the scientific principle.
@ Dung
The posterior probability depends on the distribution of the prior probability.
Since yout original post was so offensive, I suggest that you go away and learn some maths.
@ Martin A
The theory of op-amps depends om am "infinite" input impedence. I will leave you to ponder whether this is the case in climate feedback,
I discuss this at some length in my original link. I would caution you against applying the theory of op=amps to climate.
Thanks for the response. If you think that answers my question, fair enough. I wasn't suggesting that op-amp theory is relevant to climate; I was just giving an example of a feedback system described by a 1st order differential equation.
I'd point out that the theory of op-amps does not in fact depend on infinite (or even large) input impedance. However, it does depend on very large gain to ensure that the inverting input terminal is a virtual earth.
@MartinA
Phase dear Boy Phase. Do you imagine that everything is wholy real?
Martin A
The paleo data shows a pattern of shifts between two equilibrium states. Glacial periods show CO2 levels around 200ppm and temperature around 10C. Interglacials periods show peaks of 280ppm and 15C.
The transition takes some 10,000 years to equilibrium, with orbitally induced temperature changes of ~1C as the driver and CO2, by conventional greenhouse theory, acting as an amplifier. If you do not regard this as a feedback system, how would you describe the processes involved?
Richard Drake
saying that Callendar's "model" is of any value then might snare you into thinking that CO2 and equivalent "forcings" are the only things that matter in determining the climate. In fact, it might drive you closer towards the CO2 is "primus inter pares" camp.
Is that what you think? My take is that even Callendar's model lacks explanatory and predictive power.
"What is the chance that natural variation these past 17 years or so has operated with a downward forcing exactly equal and oposite to the positive forcing of CO2, if CO2 in real world conditions (not the laboratory) actually results in warming?"
richard verney
p=1.0
It is happening.
@MirtinA
Having thought about your comment, if you reduce the gain, does this lead to a first order ODE?
I would be interested to see your working.
I have engaged in this futile debate with illuminati of the blogosphere on several occasions who have quoted EE101 at me.
I would suggest that the op-amp analogy is misleading.
Actually I have had to deal with this in the pratosphere on many occasions.
You will notice that I have never mentioned op-amps. I have dealt with the mathematical basis of a system with feed back in a highly simplified way, As you make the assumptions more realistic, the maths becomes less simplistic.
The analogy you suggest dends on high gain (as you rightly point out) and a high input impedence. I am unconvinced that these have anything to do with climate models.
RC Saumarez
Do tell what degrees you have in Meteorology, Geology, Biology, Paleontology, Physics, Chemistry and Astronomy? If you believe that a degree in maths and/or statistics is required before being qualified to proffer an opinion about climate change then degrees in these subjects are also required. Go away and learn humility.
I note that my original post was indeed so offensive that it left you quite unable to answer the fairly "simple" question that was asked.
Martin A,
You are correct, feedback does not require a second order response. I do not know where RCS gleaned this requirement, but is is simply false. Feedback is very simply the process by which some of the past output is saved and added back into the current input to modify the current (immediate future) output, i.e., past outputs are fed back into the system. The most simple is an integrator: y(n) = x(n) + y(n-1). An electrical circuit with a very simple single pole response is a capacitor, which has a single pole at s = 0, i.e., DC. First order feedback responses represent basic storage systems.
Don't get me wrong, climate science completely borks feedback and related terminology - RCS is absolutely correct about that. It is also likely that there are many more poles (higher order) than just one, though there is a point at which they no longer describe much of the variance in a system's output.
Mark
Aug 29, 2013 at 12:28 AM | Entropic Man
///////////////////////
I guess that is why the satellite data shows no CO2 related warming these past 33 years.
Aapart from an ENSO event, the globe does not appear to have warmed these past 33 years!
The Met Office predicts that there will be no return tio warming before 2017. I have no confidence in any prediction made by the Met Office, but if that is so, it means that there will then be 37 years worth of satellite data showing no warming and the other thermometer records will show over 21 years without any warming.
But what if, the stasis in the temperature anomaly contines beyond 2017? What does that say about Climate Sensitivity? At that stage we will have
1) some 37 years of satellite temp data that shows no first order correlation with CO2 and just an isolated one off ENSO related temperature shift; and
2 the other thermometer records showing a number of warming events prior to 1940 which were not CO2 induced, one solitary beriod between the late 1970s and mid/to late 1990s, ie a period of less than 20 years in which there is some warming coincident upon a rise in CO2 emissions, and then over 21 years of no temperature rise notwithstanding rising CO2 emissions and notwithstanding CO2 levels of over 400ppm.
Given that the rate of warming between the late 1970s and mid/to late 1990s, is no greater than the rate of warming between the 1920s and 1940s, it would suggest that Climate Sensitivity to CO2 does not add anything to the natural variation that appears to have brought about the 1920s to 1940s warming. On top of that, there must be question marks as to whether there was in fact any significant warming during the period between the mid/late 1970s to mid/late 1990s since during the period of overlap, the satellite data set does not show anysuch warming. One obvious explanation is that the the other thermometer records show warming during this period not because there was any real warming but simply as a consequence of pollution via UHI and poor siting issues and poor screen mauintenance issues, and basterdisation through endless adjustments to raw data/homogenisation thereof (the need for which is moot) and biases creeping in due to station drop outs.
Put all of this together, and the observational case for CO2 induced warming becomes very weak. It becomes even weaker when one considers the paleo record (which of course contains uncertainties) but there are many periods within the paleo record of anti correlation, ie., rising CO2 levels yet global cooling, falling CO2 levels yet global warming, the enetring into ice ages with high levels of CO2, coming out of ice ages with low levels of CO2, and of course the evidence that CO2 lags temperature by some 600 to 1000 years etc.
I do not dispute the physical laboratory properties of CO2, but how CO2 operates in the real world, ie the complex flux of Earth's atmosphere is a completely different matter, and observational evidence suggests that Climate Sensitivity is so low that the signal, if any, cannot be seperated from the noisy variation of natural variation. As matters presently stand, it is so low that it is unmeasurable/detectable within the limitation of the sensitivity, resolution and errors of our present day equipment.
The fact (and I am not asserting that this is a fact, but merely putting the assertion forward for the sake of argument) that CO2 does not, in real world conditions, replicate what one might assume would be the outcome based upon its laboratory behavoiral characteristics is not surprising. Consider Ulsain Bolt. He can run 100 metres in around 9.5 seconds. that is his track speed. If you like his laboratory capabilities. Now see how fast he runs 100 metres along the pavvement of Oxford Street at the busiest time of the day on the busiest shopping day before Christmas. He would perform no better than you or I, and would take minutes to cover 100 metres. This is not because his theoretical speed has been reduced. It is due entirely due to complex interactions with other people who crowd the street/pavement. The same may be so of CO2 in Earth's complex atmosphere. Just a thought.
@MartinA
Phase dear Boy Phase. Do you imagine that everything is wholy real?
Aug 29, 2013 at 12:04 AM RC Saumarez
I hope your granny responds well to lessons on the art of sucking eggs.
@MirtinA
Having thought about your comment, if you reduce the gain, does this lead to a first order ODE?
I would be interested to see your working.
If you have a system with high-order dynamics (ie N s-plane poles or described by an Nth order differential equation) and feedback around it (no dynamics in the feedback path), then the system with feedback still has dynamics of order N, irrespective of the value of the loop gain. So, in principle, reducing the loop gain has no effect on the order of the differential equation relating input and output.
However, as you reduce the loop gain (making assumptions about the system not having some pathological configuration of its poles), yes, you can then neglect the shorter time-constant poles. They haven't gone away but they no longer affect things significantly.
In frequency response terms, the higher order poles then only produce significant phase shift at frequencies where the loop gain has fallen to fractional values. So you can forget they are there.
The whole art of designing a feedback system to have a non-overshoot transient response consists of getting it to behave as if it were a 1st order system, despite the combustion chamber or what ever it is you want to control having high order dynamics.
Your assertion that, if there is feedback it can't be a 1st order system - it does not add up.
If you do not regard this as a feedback system, how would you describe the processes involved?
Aug 29, 2013 at 12:12 AM Entropic Man
Entro - I don't think I said anything that implied I did not regard this as a feedback system. Nor that I did regard it as a feedback system.
@Martin A
The climate system is not an op-circuit. I am perfectly well aware that an op-amp circuit can be built that behaves as a 1st order system.
This requires a) High gain
@Martin A
The climate system is not an op-circuit. I am perfectly well aware that an op-amp circuit can be built that BEHAVES as a 1st order system.
This requires a) High gain
b) High input impedence
c) Power consumption by the op amp. Note that an op-amp circuit will not function without drawing power.
The transfer function of a basic feedback circuit is Y(s)=G(s)/(1+G(s)H(s)). If this is reduced to an ODE, It is 2nd order. You can use op-amps in which the phase characteristics are negligible at the frequencies involved - but this is not the same as saying that they theoretically do not exist.
However, in my view the whole op-amp analogy is a waste of time since any climate systems model does not have the characteristics of an op-amp.
RCS - We're just not communicating. Let's drop it.
Martin A - I agree. All I said was that the formulation of Dessler and Spencer is not feedback. I stand by that.
@Dung.
The reason that the ECS has changed is that it is based on a Bayesian model. Therefore one has to make assumptions about the distributions of prior probabilities. When these distributions are changed from uniform to something approaching what is observed, the median value of the ECS falls
RC Saumarez (12:06 PM) -
"When these distributions are changed from uniform to something approaching what is observed..."
Isn't it rather than the prior should not be uniform but uninformed? Setting the prior to "something approaching what is observed" is not in the Bayesian spirit. A priori should not include observations.
RCS - If I could just confirm that I have understood your viewpoint.
- I think you hold that if, in Y(s)=G(s)/(1+G(s)H(s)), H(s)=constant, then it is not considered feedback.
- Dessler and Spencer have H(s)=constant, so what they are discussing should not be considered feedback.
Have I got it? Y/N would suffice.
Martin A
Yes. But they call it feedback.
But they write a 1st order ODE which in general cannot contain feedback. Both the forward network which is water vapour forming contins a delay (phase response) and clouds forming also contains delays, which they freely admit and then try to calculate "lags"
RCS - OK thanks. You've confirmed that I now understand what your viewpoint is.
MartinA,
RCS is still wrong... Even his revised system is not limited to 2nd order. He is doing exactly what he accuses Spencer and Dessler of: redefining feedback to fit his necessity of a 2nd order system.
Mark