Bishop Hill Shaviv on models and sensitivity
Jan 11, 2012 Climate: Models Richard Betts joined in the conversation about climate models today, making some interesting comments on validation:
As I've mentioned before, the earlier climate models used in the 1970s were used to make estimates of warming over the next 30 years which were fairly close to what happened ... BH asks for tests of the projections made 10 years ago, but the problem is that with internal variability in the system you need longer than that to test the models, unless you specifically initialise the models with the conditions of (say) 2001 using data assimilation techniques, and that kind of thing was not available then, we only started doing it 5 years ago.
So yes, out of sample testing on timescales relevant to GHG rise is an important point but by definition difficult with the latest models!
One minor point is that I had said I would have been more convinced had the story of model versus data in the last ten years been different - I agree with the 30 years figure for falsification. However, more interesting is a point made in a recent post by Nir Shaviv:
From the first IPCC report until the previous IPCC report, climate predictions for future temperature increase where based on a climate sensitivity of 1.5 to 4.5°C per CO2 doubling. This range, in fact, goes back to the 1979 Charney report published by the National Academy of Sciences. That is, after 33 years of climate research and many billions of dollars of research, the possible range of climate sensitivities is virtually the same! In the last (AR4) IPCC report the range was actually slightly narrowed down to 2 to 4.5°C increase per CO2 doubling (without any good reason if you ask me). In any case, this increase of the lower limit will only aggravate the point I make below, which is as follows.
Because the possible range of sensitivities has been virtually the same, it means that the predictions made in the first IPCC report in 1990 should still be valid. That is, according to the writers of all the IPCC reports, the temperature today should be within the range of predictions made 22 years ago. But they are not!
Go and take a look at the graph at Nir's site. This seems a reasonable point to me.
Reader Comments (130)
In face of the suppurating IPCC and its self-appointed hierophants, making an objective, rational case for empirical analysis in this pseudo-scientific discipline is a mug's game.
Richard...the question arises, did you run models that wuld show projected temperatures for now from the 1980s? How are they performing?
A climate model is only as good as the component models. If any of the component models is invalid, so is the climate model. As I understand it, several of the component models are invalid. Ergo....
OK Models aren't perfect but the question remains what alternative would advocates positioned as skeptics prefer? Dr Mann asked this in his recent interview and I posed it in the ' Mann, straw man and SciAm' thread the other day, but no coherent answer has been arrived at.
diogenes
Don't know how Richard will answer, but take a look at the following Decadal Predictions/Forecasts from the MO site:-
First from 2009 including - "Previous predictions starting from June 1985, 1995 and 2005":-
http://www.metoffice.gov.uk/research/climate/seasonal-to-decadal/decadal-prediction
Second is an update starting late 2011:-
http://www.metoffice.gov.uk/research/climate/seasonal-to-decadal/long-range/decadal-fc
Draw your own conclusions.
Equally, Hengist, in the absence of knowledge of what happens next, what alternative is there to pulling out the metal pin from the strange pineapple-shaped metal thing? Without eyes and ears, where is the wisdom in stepping into the road?
You're pushing a false imperative, Hengist. The alternative to models is empirical evidence. If you don't have empirical evidence, you have nothing. So stop pushing.
Hengist, the models don't have to be perfect, but the suggestion here is that the sensitivity to CO2 is overestimated, and that the models should be run with a set of lower sensitivities. Is that not reasonable?
The climate models are based on 4 [Four] incorrect assumptions about the Physics.
Stunningly reasonable, Cumbrian Lad. I wonder why it has not been possible to come up with a similar solution before ;-)
A model is valid only if it predicts. That means in the future, not data that already exist when the model is run.
So what have these models actually predicted? -- No more polar bears or snow for the kiddies. Sea water lapping around our ankles as we walk to work.
Yes, Hengist, the models are imperfect and perhaps one day may predict but until that happens, I for one an not willing to change the economic infrastructure of the world "just in case."
A bit off topic but this should be of interest to you in the UK -- this from Bloomberg, the financial newspaper of the US.
Solar Panels SURGE!
Hasn't Prof. Michael E. Mann already explained how the models can be validated?
http://www.meteo.psu.edu/~mann/Mann/lectures/index.php
Procedure:
1) Cherry pick the range of data used in validation.
2) Use scientific terms like 'pretty much spot on'
3) Hide inconvenient assumptions (like decreased, constant, or increased CO2 production)
You'd think that Betts and Hengist could follow this line of argument - what a pity that Mann's followers are not up to the master's standard.
Not quite sure where the following quote from Richard Betts comes from:-
I guess that this relates to the 1972 Sawyer paper Richard mentioned on the "Evidence, confidence and uncertainties" discussion thread? If so, then as I pointed out on that thread (bottom of page 7), it is more reasonable to conclude from this paper that natural causes have also made a significant contribution to the warming. I imagine this applies to any estimate based on the 1967 result of Manabe and Wetherald, who do not take account of the following factors, both mentioned by Sawyer, but not taken account of in Sawyer's estimate:-
1/ The effect of increased water vapour on cloud cover.
2/ The thermal inertia of the oceans.
ZT
You forgot
4) Proclaim loudly that the "science is settled".
5) Denigrate anyone who disagrees with the Truth.
Josh -- there is a good cartoon to be had in my Solar Panel Surge. Read the article.
mydogsgotnonose
That is because you took physics and not "Natural Philosophy". We actually have that course here in the US, known as "Physics for Poets" which may explain why they take poetic license to the laws of physics with the same abandoned some do to the Iambic pentameter.
Jan 11, 2012 at 10:39 PM | Hengist McStone
OK Models aren't perfect but the question remains what alternative would advocates positioned as skeptics prefer? Dr Mann asked this in his recent interview and I posed it in the ' Mann, straw man and SciAm' thread the other day, but no coherent answer has been arrived at.
=============
James Evans gave a concise and very coherent answer on that thread.
"You don't need a wonderful alternative to shite. A simple absence of shite would be a vast improvement."
Jan 10, 2012 at 7:35 PM | James Evans
Hi Philip
I've (finally!) responded on the discussion thread (sorry to take so long) but I'll make the same point here for completeness.
As you noted in your other post, Manabe and Wetherald's model also neglects positive feedbacks from decreased surface albedo as snow and ice melts, exposing the darker underlying land / ocean surface. This acts in opposition to the negative feedbacks.
Re: Hengist
you are making the assumption that the world needs to know what the climate will be like in 50 or 100 years from now.
Whilst having this knowledge might prove useful it is not essential.
It is vulgar, and somewhat inflammatory, but such has been the misuse of models in the scientific and the political spheres that I am inclining to the view that a 'simple absence of shite would be a vast improvement'. The technical fun to be had with models is all but irresistable to those with the patience to program them up. The heroic stance that these models might help humanity is good for both the spirits of these programmers and also for their continued funding. But oh what a diversion the undue attention and weight given to these models! They are primitive toys in the face of the immense complexity of the climate system, yet they are very complex toys as well and I'd guess we fail to even understand enough about how the models work, let alone the climate system. I would like to see them relegated to a secondary or tertiary role until we know more about the modelling task and potentials, and more about the climate system. I'd like to see discussion of physical and statistical models by the field's experts without recourse to 'the GCM says this' because of course they can be made to say all sorts of things. They are not evidence. They are not core science. They are not oracles. They are not to be worshipped and sat at the foot of. They have become a weapon in the hands of manipulators and political opportunists and that abuse makes me think we would all be better off today if they had never been invented. But of course, they are there and they seem to be generously funded, and they provide endless challenges and diversions for their owners and their programmers. Let us keep them. But let us keep them in a sensible perspective.
Hengist:
"OK Models aren't perfect but the question remains what alternative would advocates positioned as skeptics prefer? Dr Mann asked this in his recent interview and I posed it in the ' Mann, straw man and SciAm' thread the other day, but no coherent answer has been arrived at."
You should have stopped at "OK Models aren't perfect..." There is no, "OK....but...." and what Mann asked was nothing more than misdirection and irrelevant. If the models are no good, they are no good whether alternatives are proposed or not. you are effectively saying that even if the models are completely wrong they have to be accepted unless one comes up with a better alternative. The fact is, if the models are wrong, they are wrong, period. Full stop. It is up to Mann and the Team to demonstrate that the models are correct, not obfuscate and misdirect.
Hi Green Sand
Thanks for that, it is what I would have shown diogenes too.
Yes the 1985-1994 hindcast didn't do so well because the cooling effect of Mt Pinatubo wasn't included in the model (there was no way we would have know that was going to happen if we'd done the forecast in 1985, so including it would have been cheating). This is of course why it is important to remember that the climate model simulations of future climate are caveated with "assuming no major volcanos erupt".
The 1995-2004 period looks OK doesn't it? The overall change is pretty good and the only instance in that period when the observed global temperature (black curve) went outside the estimated uncertainty range (the red plume) was 1997/98 El Nino - clearly that particular one was stronger than the model tends to produce.
The forecast issued in 2007 (that was an actual forecast, not a hindcast) is just about looking OK in that the observations are still just within the uncertainty plume, but yes it is pretty close and if we don't see a rise in annual mean temperature in the next two or three years then yes that 2007 forecast won't look so good...!
So yes it's very early days for initialised forecast modelling like this, but we think it is promising.
The forecast for annual global mean temperature in 2012 is here. Let's see how we get on!
Richard Betts
Thanks Richard, generally agree with your summary, time will tell.
"and if we don't see a rise in annual mean temperature in the next two or three years then yes that 2007 forecast won't look so good...!"
But more importantly and significantly won't it make the the 2011/12 forecast will look even worse?
Thanks for the link to the "Met Office 2012 annual global temperature forecast". Why does it differ to the Decadal Forecast?
"4 January 2012 - 2012 is expected to be around 0.48 °C warmer than the long-term (1961-1990)"
Whilst the Decadal Forecast "predicted global average annual surface temperature difference relative to 1971-2000 Is 0.46 °C which equates to 0.58c against long-term (1961-1990)?
Has the MO reduced the 14th Dec 2011 Decadal forecast already?
Hengist McStone Jan 11, 2012 at 10:39 PM
If a model can't be validated, it's better not to use it at all and simply say "we just don't know" than to kid yourself that its predictions can be relied on.
DPdlS,
Good points - it is by no means trivial proving that the models are perfect without actually testing them.
I've often wondered why the modelers don't simply run the models in reverse to prove that there was no medieval warm period. Surely this would be simpler than the standard climatological denigration approach to written history and would also provide proof that the models are correct. (i.e. take the absence of the MWP as an irrefutable axiom, reverse time in the models, show that historically temperature was always completely constant, pick up tickets to Stockholm (second trip).
If only Mann didn't have to deal with the mundane trivia of insufficiently bright disciples this would have been done years ago.
Green Sand
Please can you give the link to the second one? (The decadal forecast). I'm not quite clear what you are comparing so it would be good to see the original, and I can't find it on the website which suggests I've misunderstood you and am looking in the wrong place!
Richard Betts
http://www.metoffice.gov.uk/research/climate/seasonal-to-decadal/long-range/decadal-fc
ZT
Yes, you are quite right, Mann's biggest mistake (there are several) was to not prove mathematically with his computer models -- which are computer based and therefore correct and accurate -- that the written history, old Dutch paintings from the MWP and such were obviously wrong and the figment of their creators' imagination.
And I quite agree that he has missed out on his opportunity to go to Stockholm because of that foolish mistake. Obviously, I think you will agree with me that the poor man (pun) has suffered enough and we should stop questioning his obvious brilliance.
Hi Green Sand
Thanks. That page is about the decadal forecast, and says:
So it gives 5- year averages. The 2012 forecast I quoted was the annual mean for this year only.
(I still may have misunderstood you as I couldn't find the exact sentence you quoted, but hopefully this helps. If I have misunderstood then please feel free to clarify!)
Richard Betts
PS, if it helps I also have problems finding anything relevant on that site.
But wait, maybe help is at hand? http://www.metoffice.gov.uk/public/beta/
Well the last time the MO site changed, not that long ago, I spent months trying to crack the code into basic temp data, gave up, went back to CRU.
Have eventually got back in sync I think? But just why is the MO so distant?
For a simple example please look at the following two " related" charts:-
http://www.metoffice.gov.uk/research/climate/seasonal-to-decadal/decadal-prediction
http://www.metoffice.gov.uk/research/climate/seasonal-to-decadal/long-range/decadal-fc
The latter is supposed to be a continuation of the former, why all the changes?
The changes bring absolutely nothing other than obfuscation, why do it?
Richard Betts
Many thanks, tis late as I am sure it is for you.
"That page is about the decadal forecast, and says:"
Yup, I know what it says, that is why I brought it to your attention. After it was brought to my attention I studied the chart, digitised it and then a collegue asked the MO for the data from which the chart was constructed. Received earlier today was the following, which for brevity is only for the next year:-
2011.67 0.345
2011.75 0.354
2011.83 0.366
2011.92 0.382
2012.00 0.396
2012.08 0.412
2012.17 0.423
2012.25 0.433
2012.33 0.442
2012.42 0.451
2012.50 0.459
2012.58 0.461
2012.67 0.464
2012.75 0.472
2012.83 0.485
2012.92 0.501
2013.00 0.514
Average 2012.08 - 2013 = +0.460C
So the original question remains "has the MO reduced the 14th Dec 2011 Decadal forecast already?"
I have related this anecdote earlier, but Hengist's championing of Dr Mann prompts me to relate it again as it seems very relevant (to me, at least).
Many years ago, just prior to the current era of personal computers everywhere, I visited a high school science fair in a NZ provincial city and was saddened by the display created by some very bright but misguided Seventh form boys who had constructed a life-sized cardboard model of a PC; the group of boys, when I discussed their model with them, were adamant that their empty cardboard representation of a 'computer' was 'real science'. I saw it as a form of Cargo Cultism and despite Hengist's obvious intelligence, he has made a similar error. The climate models Dr Mann (and Hengist) place such reliance upon are crude and obviously wildly inaccurate representations of 'climate,' with a similar level of knowledge and understanding of climate as the stone-age Pacific Islanders had of aircraft design but who worshipped very crude model aeroplanes in the hopes of having wondrous 'cargo' that would arrive in real aircraft and be bestowed upon them as it was disgorged.
James Evans phrased an inelegant truth elegantly: 'an absence of shite would be hugely preferable to Mann's ( and thus Hengist's) tortured nonsense.
Falsifiable over 30 years?
If it wasn't politicised climate science, people would laugh at 30 years.The goalposts have been so skewed that statements like "30 years" are accepted as a base line. How about 3000? When all the chaotic natural cycles have some chance to make their presence known?
Hengis keeps talking about what is the alternative to models. You do not need an alternative, you just need some reality as to their validity. Which isn't much.
And the argument that we can predict the next 5 days weather, and those same models hold well on a yearly scale?
Just because you throw millions of pounds at the problem and create models of beautiful complexity doesn't mean complexity=validity.
Validating long time period chaotic systems? Good luck to you.
Jan 11, 2012 at 11:54 PM | Richard Betts
Please stop leaning on the "uncertainty plume." Address it.
Jan 11, 2012 at 11:00 PM | mydogsgotnonose
Agreed. The discussions involving Lapse Rate and the minor contribution of CO2 (see for example Tallbloke's blog) are simply being ignored.
As I understand it, the impact of CO2 on LR is via its specific heat, and as this includes its absorption and emission of narrow bands of IR, attempting to include this absorption process as a major driver of impedance of heat flow from Earth to space is grossly doubling accounting.
Can some-one with more familiarity with the physics please elucidate. To this ancient physicist, the thermodynamic arguments being aired over Tallbloke's seem very compelling, but I would welcome some input from Richard Betts on the topic.(Apologies in advance Richard if I have missed anything from you).
I looked at the graphs on Nir Shaviv's site.
They represent the output of 30 years work and over 100 billion dollars of investment in climatology.
And they are not accurate at providing climate forecasts.
Somebody please convince me - from the PoV of a taxpayer, not an academic
1. That it has not been a complete waste of time and money
2. That there is a reason to throw good money after bad.
What benefits have we received from this expenditure? What useful practical things do we know now that we didn't know before?
Because I am increasingly feeling that the whole exercise has been a job creation scheme for wannabee academics who found proper science too hard and/or too rigorous.
@Jan 11, 2012 at 11:35 PM
Hi Richard,
Yes, I agree about the ice, but the overall effect of these omissions is to make Sawyer's paper far less convincing as a prediction then it seems on the surface, although as I mentioned on the thread, I also think it is a very good paper, even from today's perspective.
Latimer you ask
I have seen no benefits but a great deal of harm. The list of harmful effects is seemingly endless:
wind turbines
solar panels
flood defences
science corrupted
corruption in general
education trashed
increased bureaucracy
money wasted
time wasted
people's lives ruined
etc
etc
Jan 11, 2012 at 11:51 PM | John Shade
Totally agree, can we just put the GCM's away for a while and discuss the actual science going on in the field, observations, experiments, lab work etc..
Jan 12, 2012 at 6:28 AM | Ronaldo
I can't really follow the thermodynamics examples, but I agree with tallblokes 'how could IR radiation ever get into the ocean thread'. But even accepting if CO2 heated the lower atmosphere it would seem to be every feedback the Earth would have would be negative anyway. ie extra convection more clouds etc.. (just go to the tropics and see what happens in a typical afternoon that seems to have a cooling effect.)
According to the Met Office
The only way? So, according to the Met Office, if I predict to 90% confidence that the Holocene interglacial will end in the next two thousand years I am wrong because I haven't used a computer model. When does the Met Office computer model predict the Holocene will end? I'm willing to place a bet on the outcome of our two predictions
Another alternative if people want to discuss GCM results is to discuss a fully documented GCM. I would like to know every assumption and parameter that has been made or set in the model to get the results that it did. It would also be good to have the full output (or at least something like the temp range over the full run) to see what they are good at modelling and where they struggle, rather than 'average globa; temperature' was 4.2 degrees higher in 2100 when CO2 was at 530ppm or whatever..
Richard Bettes says
In other words, as Lucia has been saying for some time, the models are on the verge of being falsified?
Unfortunately, such a line of argument has serious flaws. It is based on just one (effective) sample point, and there existed models in the 1970s that produced predictions (sorry, projections) ranging from plunging into the next ice age through to thermageddon.
The probability that we would get to today and be able to look back at those runs and observe that some of them were correct is one. That is, such an observation carries no useful information. And it is equally unsurprising that the models which matched that one sample point are the ones people find interesting today.
It reminds me a little of Gavin over at realclimate explaining the failure of polar amplification in the Antarctic. His evidence that this didn't matter was to dig out a single run from the 1980s which didn't have much warming over the south pole.
For those of us that understand conditional probability, the fact that there existed models which got around the right answer 30-40 years ago based on one degree of freedom is uninteresting because there existed models that gave pretty much every possible answer (given the confidence intervals).
If climate scientists want to convince me, they need to produce something more than 3-4 degrees of freedom in hindcast and one point out of sample which is guaranteed to be covered by some model, somewhere. But when we apply rigorous tests to models with more degrees of freedom, they fail miserably, even on statistical measures.
Furthermore, as we increase timescales (from monthly, to annual, to 30-year scales), predictions generally get worse - something exprienced modellers expect, but climate scientists brush under the carpet with their 3-4 sample point hindcasts.
The additional "e" mysteriously appearing in Richard's surname reflects his new found twitter status as a side-kick wielding, tights wearing, caped crusader...
While checking up some references I've just come across Richard's opinions from the 4 degree+ conference in Oxford in 2009. I can't even comprehend the physics that would heat the arctic by 15 degrees by 2090, surely the entire ocean would need to be heated to a similar degree to obtain that level of warming? I'm still not sure where the decreased precipitation prediction come from either.
http://www.eci.ox.ac.uk/4degrees/ppt/1-2betts.pdf
Conclusions
• Current CO2 emissions are near (but not above) upper end of IPCC
scenarios
• 4°C global warming (relative to pre-industrial) is possible by the
2090s, especially under high emissions scenario
• Many areas could warm by 10°C or more
• The Arctic could warm by 15°C or more
• Annual precipitation could decrease by 20% or more in many areas
• Carbon cycle feedbacks expected to accelerate warming
• With high emissions, best guess is 4°C in 2070s
• Plausible worst case: 4°C by 2060
Rob Burton
Thanks for that link. Where has the science gone to? Null points for Richard Betts. That presentation is a complete load of alarmist ****, obviously designed to keep politicians on-side. Hasn't Richard come here before and told us maybe they have been just a little bit too alarmist in the past and that he will do his best to make sure alarmism doen't creep into the message?
At least I am comforted to know that for the rest of my life I am going to enjoy a warmer and warmer climate with no possibility of a return to Little Ice Age conditions or even worse.
Richard Betts:
Here it is in graphical form. Pale colors are the estimated for the last years, blue vertical line is the forecast and range. I am sorry to say that it does not look very impressive.
Would you consider the forecast validated if it falls within the range?
This article from Lubos Motl is extremely relevant: http://motls.blogspot.com/2012/01/will-co2-save-us-from-next-ice-age.html
'This paper by a former student of Richard Lindzen finally managed to fix Milutin Milankovič's theory which hadn't worked and the outcome was a theory that does work. The graph below contains both theoretical predictions as well as the observed data about the Northern Hemisphere temperature and you may see that the match is beautiful:'
It predicts the next ice age. CO2 is irrelevant.
'Those people still attribute CO2 some amazing importance for the climate which it doesn't have and which it has never had. It's sad to see that most of the science journalists and probably also climatologists (and related geologists) are incompetent but it's probably inevitable.'
Now Hengist and Richard. We have a fully working and verified theory which shows the irrelevance of CO2..
Please tell your readers/viewers etc. the good news..... We've had enough of politically biased journalism based on false science.
My own thinking is that we still don't know whether Miskolczi is right that the water cycle adapts or, as seems to me likely, it's because there is very little thermalisation of IR by CO2, which is an assumption by the climate modellers without absolute proof. And please don't quote the PET bottle experiment, it is not proof of thermalisation by the gas, only of the bottle walls.
Ronaldo,
You ask about thermodynamics. As one "ancient physicist" to another I have been facinated by the concept that the lapse rate acts to maximise entropy (and minimise enthalpy?) in a near - equilibrium thermodynamic system, that constantly seeks equilibrium as a consequence of rotation. The implication is that only insolation, gravity, rotation and the mass of the atmosphere are responsible for "climate", and that doubling, tripling or whatever the mass of CO2 will have a negligable effect on the total mass of the atmosphere, and hence would not lead to any measurable effects on the "climate".
The concept is (I think) a consequece of the principle that the laws of thermodynamics subusme the effects of all energy transport mechanisms (conduction, convection and radiation - because they have to.
I found the following text useful: (taken from an abstract of a 2002 paper by Lorenz):
“Two thermodynamic principles offer considerable insight into the climatic and geological settings for life on other planets, namely (1) that natural systems tend to actually achieve the ideal (‘Carnot’) limit of conversion of heat into work and (2) if a fluid system such as an atmosphere has sufficient degrees of freedom, it will choose the degree of heat transport that maximizes the generation of work (equivalently, that which offers maximum entropy production). The first principle agrees well with results on terrestrial cumulus convection, and the mechanical energy released by tectonic activity. The second principle agrees with the observed zonal climates of Earth, Mars and Titan”
Perhaps a better qualified physicist than me (Jonathon Jones?) would like to comment?
Roger Longstaff Jan 12, 2012 at 9:43 AM
Roger, do you think this is related to what Harry Dale Huffman noticed in terms of the climate of Earth and Venus following the same profile, after correcting for different orbit diameters?
Martin,
In a way - yes. Harry's work started me thinking about this several months ago, and I have since followed the work of others. However, I came to the conclusion that the laws of thermodynamics (which include conservation of energy) MUST overide all internal mechanisms - hence the conjecture given above. If it is incorrect I would simply like someone to say why, and then I can stop worrying about it and get on with the day job.