Bishop Hill +++Harris and Lewis+++
Sep 14, 2013 Climate: MetOffice Climate: Models Nic Lewis has published a detailed comment on the Met Office’s report on climate sensitivity, which was itself very much a response to the Otto et al paper of which Nic was an author. The comment is here.
There is a great deal of interest, not least of which is the fact that the Met Office seems to have made a series of misrepresentations of Otto et al, as well as making several mistakes.
One of these though is astonishing. This concerns Harris et al 2013, a paper by a group of Met Office scientists. The paper is particularly important as it is the source of the official UK Climate Projections. It examines how the virtual climate inside the Met Office model responds to having key parameters tweaked. Different parameter combinations are then weighted in the final analysis depending on how well the resulting virtual climate matches observations.
Nic explains what he found:
As HadCM3's parameters are perturbed, the resulting changes in ECS and aerosol forcing are closely linked. When significantly lower values for ECS – as suggested by recent observational studies – are obtained, HadCM3's aerosol forcing takes on highly negative values. The observational data strongly contraindicate aerosol forcing being highly negative, so parameter combinations resulting in significantly reduced model ECS levels (and thus highly negative aerosol forcing) are heavily down-weighted. As a result, whatever the actual level of ECS, HadCM3-derived ECS estimates are bound to be high
Essentially, the observations would lead you to conclude either that a high climate sensitivity is being masked by a quite strong cooling effect from aerosols, or that there is lower climate sensitivity and that aerosol cooling is a much smaller effect too. However, the Harris et al study behaves in a very different way. If you tweak the settings on the model in a way that gives you relatively low climate sensitivity, the model ends up with a very strong aerosol cooling effect too. The relationship between aerosols and climate is the opposite way round to the observational studies!
The aerosol cooling that appears in the Met Office model when climate sensitivity is low is very strong, completely inconsistent with recent observationally-based estimates of aerosol forcing. In fact it is so strong that it would also have prevented most of the temperature rise seen since industrialisation. In other words, the virtual climate produced with these settings of the model doesn’t match the real one. This means any scenario in which climate sensitivity is low – as indeed the observational studies suggest it is – gets downweighted in the final analysis to the extent that it doesn’t really show up in the final results. The effect is essentially to rule out low climate sensitivity as a possibility.
I’ll say that again in a slightly different way. The Met Office’s model, one used generate the official climate projections, has big temperature rises built in a priori.
Wow.
Reader Comments (85)
new David Rose article
(...)
It seems to me that the Met Office adopted a set of beliefs and constructed their models around these beliefs and today they are finding that their initial assumptions were flawed. It is now late in the day to start introducing natural factors such as the ocean oscillations, solar effects and cloud albedo so it is difficult to see how the Met Office can convert from a belief based religion to objective science without being caught out.
(...)
Sep 14, 2013 at 7:33 PM Schrodinger's Cat
There are faint signs of an ongoing debate about these matters among the Met Office management.
So far, the 'carry on as usual' team seems to remain firmly in command, notwithstanding the Met Office's new line "we just do science - not advocacy. And certainly not propaganda."
However, the Met Office has well and truly painted itself into a corner, from which there is no easy way out.
Thanks for the link to David Rose's new Daily mail article, Phillip Bratby. The explanation of Nic Lewis' work is neatly summarised there.
Lapse rates - what dey?! Aerosols! Give me some erupting volcanoes man!
No amount of tweaking, no amount of creative algorithmic nuance.......
Met Office Gollum, in cave down in Hadley nr Exeter sends email to his master @ DECC:
"......... no matter how loverly large is the computer modelses masssster - if it be based on a big fat falsehood taleses [>mmCO2=warming T's] then no matter what we doos masssster it won't be right, now give me the RING massssssssssster!"
DECC to Gollum,
"You'd better do what you're there for - make some more stuff up - potatoED is getting desperate and we're worried about our jobs too!"
P.S. "gave Brussels a ring!"
Mr. Lewis, thank you for the response. It was very helpful.
Athelstan
+ lots
Nic Lewis says:
"The Met Office climate models and GCMs generate aerosol loading and the forcing it causes from estimated emissions of sulphates, organic and inorganic carbon particles and other pollutants."
Does this mean that the natural dust loading, which is known to be large and extremely variable is ignored? This does not apply to ice ages only, during the midwest drought years of the 1930's they had "black snowstorms" in the Canadian Arctic.
Tank you Nic for wading through the model. Those of us who have not done so might conclude that there is a fundamentally wrong treatment of numerical errors. A correct treatment would possibly have highlighted the logical error.
Blimey. David Rose doesn't let up, does he? He's bludgeoning them.
Peter Stroud @ 4:02pm
I don't think you needed the word "model" after Met Office...
Oh dear. If only the play station modellers in the Met Office had listened to their colleague Peter Thorne:
. .Hats off to Andrew Orlowski and David Rose, and shame on the rest of the mainstream journalists for refusing to investigate these incompetent/corrupt scientists.
Hopefully Andrew Neil will manage to make more progress within the BBC.
Steven Mosher says:
”with that model there are something like 30-32 paramaters to be tweeked.
A full factorial examination of the parameter space is not feasible so they run certain corner cases, then they can build a statistical emulator and investigate the parameter space with that.. note any weird results and then run the full blown model.”
Anyone who builds a model like that is beyond insane unless they have very precise values for at least 27-29 of those parameters.
It reminds me of a very complex output report in a very complex information system I helped developing once. It had something like 40 choices when specifying output format (though many were admittedly simple yes/no choices). We tested several hundred of the most likely selections before putting it into service, and it worked OK for several years. Then somebody managed to find a combination of choices that were self-contradictory and which we had failed to block in the software, with the result that I got a message from “on high” demanding to know why we hadn’t checked all possible combinations.
I answered that it was because, if we tested one combination per second we would still not be finished by the time the Sun turns into a red giant and swallows up the Earth.
I never heard anything more about it.
tty: this kind of approach is used in the oil industry to effectively interpolate the output from a selected number of static and fluid flow (dynamic) uncertainty runs from a reservoir model. Its a form of latin hypercube sampling, as I recall. However, the oil industry models are linear or quasi-linear problems, climate is not, so its difficult for me to see how it would work for climate models.
And I remember having to wash my Ford Anglia in the midlands after dust storms in the Sahara.
Is this just more evidence that the Met Office model is not fit for purpose? Add its failure to predict the current hiatus, and surely it must be ignored in any future policy decisions.
Sep 14, 2013 at 4:02 PM | Unregistered Commenter Peter Stroud
Yes.
And don't forget that the Met Office's Chief Scientist has said that models are the only way to predict future climate.
Bottom line: future climate cannot be predicted.
The models are obviously wrong for many reasons.
Ignoring the obvious issue, as to whether increasing CO2 emissions in the real world atmospheric conditions leads to warming, two glaring issues stand out.
First, the thermometer record is corrupted (by polluted by UHI, poor siting of stations, and possibly by screen maintenance issues, and biased by station drop outs and endless adjustments/homogenisation the need and correctness for which is moot) and when tuned by hindcasing to this record, the tuning is off track (due to the aforementioned problems with the thermometer record) so there is no prospect of the projections for the future being correct. Wrong callibration inevitably leads to wrong rsults.
Second, the K&T energy budeget is almost certainly flawed. K&T suggest that incoming solar is 342w/m~2 of which some 77w/m~2 gets reflecting upwards by clouds and other aerosols such that it never reaches the surface ground below. A further 87w/m~2 is absorbed by the atmosphere (presumably heating the atmosphere) and it too never reaches the surface ground below. Accordingly of the incoming 342w/m~2 of solar energy only some 198w/m~2 makes it to ground level of which some 30w/m~2 is reflected (presumably predominantly by the oceans and ice) such that only 168w/m~2 is absorbed and goes to heat the surface of the Earth. However, this is incomplete. The reflection from clouds (and other aerosols in the atmosphere) is a two way street. The top of the clouds (and top side of aerosols), reflect incoming solar. K&T put this at 77w/m~2. But they overlook that the bottom of clouds (and bottom of aerosols) reflect downwards. The bottom of clouds (etc) receive some part of the 30w/m~2 of solar that has been reflected from the surface and not absorbed by the surface. Obviously some part of the reflected 30w/m~2 makes it way directly to space without interaction with clouds (etc), but not all. Some of it impacts upon the undeside of clouds and some of this is then reflected back down to the surface.
Tthe net effect of the above is that the surface does not simply receive and absorb 168w/m~2 of solar but additionally receives and absornbs some solar energy which has been reflected downwards by the clouds. This really ought not to be contentious since we witness this phenomena every day. When you see shaded areas which are in direct shaddow of sunlight, they are not completed black. They are dark, but not black. Some light is illuminating the shaddow areas and this is light reflected back to the ground surface from the underside of clouds and the atmosphere generally. So we know as indisputable fact that this is real and that the K&T energy budget leaves out some component of solar energy that teh Earth receives and absorbs.
Further, and possibly more contentious, K&T assumes that all the 342w/m@f Now are you saying that clouds (their underside) reflect back to Earth some of the 30w/m~2 which has been refle#2 of claimed DWLWIR is absorbed by the surface and that none of it is reflected. However, is that correct. DWLWIR is omni directional such that about 11% of DWLWIR has a grazing angle of 10 degrees or less. I do not know the absorption characterics of all materials to low grazing angle LWIR, but it would surprise me if all materials did not reflect some LWIR. Perhaps ice, silica and other highly reflective materials reflect some percentage of LWIR especially when the grazing angle is low. This therefore may be an oversight in the K&T energy budget.
If the K&T budget is flawed and if that budget is being used to model the atmosphere then it is not surprising that computer models are erroneous and flawed. Just think about the solar energy re-reflected solar energy from the underside iof clouds. If this is circa 3.5w/m~2 (ie., just over 10% of the solar that has been reflected from the surface according to K&T, one already has almost all the claimed additional energy (3.7w/m~2) to a doubling of CO2.
So whilst I agree with the many comments regarding the fudge factor introduced by aerosols, I consider that there are also two further fundamental errors (being tuned to a corrupted thermometer record, and not taking account that the clouds and atmosphere re-reflect downwards some of the solar irradiance which the surface has reflected rather than absorbing such that the surface receives and absorbs moe than 168w/m~2), and possibly an error in not considering whether some part of the the DWLWIR is reflected without absorption. Indeed, it is possible that the oceans cannot really absorb much of the DWLWIR (in view of the absorption characteristics of LWIR in water such that about 50% of all LWIR is absorbed within just a few microns - not millimeters- and absorption of this amount of energy can do little more than power evaporation, and some part of the DWLWIR will be absorbed in the fine mist of water droplets-., windswept spray and spume in the immediate atmosphere above the oceans befiore it reaches the ocean below).
Further to my above post at 11:11am.
The 6th para is a little garbled. It should have read:
"Further, and possibly more contentious, K&T assumes that all the 342w/m~2 of claimed DWLWIR is absorbed by the surface and that none of it is reflected. However, is that correct? DWLWIR is omni directional such that about 11% of DWLWIR has a grazing angle of 10 degrees or less. I do not know the absorption characterics of all materials to low grazing angle LWIR, but it would surprise me if all materials did not reflect some LWIR. Perhaps ice, silica and other highly reflective materials reflect some percentage of LWIR especially when the grazing angle is low. This therefore may be an oversight in the K&T energy budget."
No sign of the BBC take on this yet. It must be a quiet weekend at the beeb, otherwise the'd be all over the story.
tty
" Does this mean that the natural dust loading, which is known to be large and extremely variable is ignored?"
Natural aerosol loading is not ignored, but it is certainly one reason for the large uncertainty in aerosol forcing (which is defined as the change since an average level in pre-industrial times).
"Anyone who builds a model like that is beyond insane unless they have very precise values for at least 27-29 of those parameters."
I agree that the large number of uncertain important parameters in AOGCMs is a major problem. It's probably not quite as bad as in your software case, since (AFAIK) strong interactions exist between only a subset of the key parameters.
The problem in this case doesn't seem to be that the study didn't explore different combinations of key parameters in HadCM3 thoroughly enough. They certainly tried quite hard to do so, although as you note computational limitations prevent a complete exploration. But the problem here is that no possible combination of parameters appears to result in HadCM3 exhibiting both a modest level of aerosol forcing and a modest level of climate sensitivity. The model is structurally rigid in regard to aerosol forcing becoming more negative at lower levels of climate sensitivity, however its parameters are changed.
Richard Verney, Willis Eschenbach had a post at WUWT recently which suggested that the cloud response to solar radiation in the tropics can change over a relatively short space of time - 5 to 10 minutes from memory - but even at this short time interval it has a massive affect on the amount of solar that makes it to the surface bearing in mind that the average solar radiation in the tropics is substantially greater than the average taken over the whole surface of the earth. I don't think the MO can model to this level of granularity.
As an aside, I don't accept that the climate can be modelled so every penny spent on climate models is a penny that could be spent in so many other and better ways.
Philip Bradley
Re your Tokyo paper, that means it might not be just an impression that the best weather occurs when most people are at work, it might actually be a fact?
"Wow. What a coincidence! To me it seems at least 97% likely."
Billy Liar
Then you are in agreement with those scientists who regard the increased aerosols due to wildfires, increased coal burning by China and increased 21st century volcanic activity as sufficient to cancel out the ongoing warming effect of increased CO2
You will also agree that since none of these were predictable at the time the models were compiled, the models should not be expected to show their effect.
The basic problem is the climate models require an overall energy balance (as does nature) and do not currently take account of ocean warming below the depth of 300m or so which is not of an obviously cyclical nature. They include an ENSO (El Nino Southern Oscillation) component, although that is not necessarily that predictable either. However, for the decade of 2000-2009 there has been a net global uptake of heat by the oceans, following a net return of heat during 1990-1999, whereas El Nino has an average cycle of five years - analysis from "Distinctive climate signals in reanalysis of global ocean heat content", Balmaseda, Trenberth and Källén, GPL 2013 volume 40. So within Nick's trials the models would have to find some way of reducing the ECS temperature rise without having the (correct) option of stashing the heat deep in the oceans.
Putting this into context the 1990s and 2000s added together have a similar average ocean heat uptake (-0.26, +1.19 W/m^2 respectively) to the 1970s and 1980s individually (0.47, 0.58 W/m^2) . So if the ocean heat absorption during the 2010s reverts to the long-term average then the climate models would get back on track in terms of prediction..
The climate models are designed to provide long-term predictions of climate rather than expecting to get it right over a single decade. So it is not unreasonable that they react to out-of-scope ocean heat exchange in an invalid way. There has been a recent suggestion that they should be changed to cope with such long-term ocean heat uptake behaviour in which case in the future they are likely to have more options to play with if flat surface temperatures are used as input. However, this would imply a change in the objectives of such models to include more short-term predictions of changes to the climate.
30 parameters to tweak?? Tell me it isn't so.
I thought we have been assured that all this Climate Science stuff, every single bit of it from top to bottom, was conducted solely with fundamental equations for physical processes and phenomena that have been well established for centuries.
I don't recall seeing Euler, Gauss, Cauchy, Poincare, the Bernoulli boys, among many others, tweaking parameters. Constructing response surfaces certainly never entered those minds.
It is especially distressing to learn that one of the more basic phenomena that explains Settled Climate Science, radiative energy transport, involves tweaking parameters that are the most fuzziest of them all; the optical properties of aerosols. We don't even know how much of these are in the atmosphere, never mind what they are and what the look like relative to radiative energy transport.
A focus on macro-scale system response functionals plus tweaking the fuzziest parameters is an approach that is certain to lead to the right answer for the wrong reasons, at best, and simply wrong answers at worst.
The Sagan aerosol optical physics is wrong. In reality the sign of the effect is reversed.
The climate alchemists can't accept this because it shows much if not all the real AGW has been from polluted clouds, not CO2.
This also explains the end of ice age, the real reason plus much other science.
Sep 15, 2013 at 10:42 AM | splitpin
And I remember the glaciers in the Pennine Alps were pink with Saharan dust last August.
Don't let the b'ast-hards ever forget that it was said that the science was settled. They gave free reign (and rein) to the policy makers, the green brew and the revenue gatherers.
Any complex dynamic system, defined as any random-recursive time-series with three or more interacting variables, will exhibit chaotic/fractal features described by Edward Lorenz and Benoit Mandelbrot c. 1960 and 1974, respectively. These self-reflexive patterns invariably derive incommensurable decimals, numbers such as base e of the natural logarithms, pi, and the square root of 2, which never end in zero and never repeat.
On this basis, constructing a simple "logistic equation" where each successive value y = Kx(1-x), that is where y / x(1-x) converts to an equation by means of a constant of proportionality K, derives a nonlinear, oscillating function similar to a modified Markov Chain-- wholly deterministic on one hand, completely indeterminate in detail, deriving a set of dual dynamic "strange attractors" self-similar on every scale.
Without delving into Godel's Theorem or Claude Shannon's Information Theory, any math/statistical neophyte can say one thing for sure: That any "model" pretending to extrapolate an incommensurable "logistic ratio" must necessarily fail by exponentially increasing margins over time. So it is, not only with Met Office hocus-pocus but with each and every other "false precision" Global Climate Model (GCM) purporting to rationalize projections from intrinsically irrational inputs.
As primarily a function of geophysical plate tectonics rather episodic atmospheric/oceanic currents, cosmic rays affecting cloud cover, or Milankovich cycles affecting TSI, "climate" in itself is a vexed concept. As Lorenz himself wrote, "Does the Earth have a climate? The answer, at first glance obvious, improves on acquaintance."
The big problem with predictions is that most of them are about the future. This means that they could turn out to be wrong. Predicting the past is usually much safer, as long as you have a good memory or your wife reminds you.
The best option, of course, is to predict the present. That way you have at least a 50% chance of being partly right.
Yeah! The official response was that they run it for weather forecasting several times a week. I have a small suspicion that truth and honesty are not their strong points.
My guess is that at some time during the creation of the next generation of climate models, someone is going to be looking at one of the modules from HADCM3 and think "shouldn't that + be a -"?
And with this single character change, it will all suddenly fall into place.
OUCH..! Someone just tweaked my parameters..
Hey - stop that, will you..? You don't know what the outcome might be...
entropic man its it 'lucky that by chance ' increased aerosols due to wildfires, increased coal burning by China and increased 21st century volcanic activity' happens to be just the right value to ' to cancel out the ongoing warming effect of increased CO2' I mean what are the odds on that ?
Excellent work by Nick. There needs to be an easy to understand guide to the various isssues. Cartoons by Josh combined with explanations from a writer who can reduce the confussion to no more than 10 points . Perhaps someone with a technical writing /advertising background. I would start with pictures of models outputs versus balloon and satellite data.I do not know how climate sensitivity could be reduced to a few cartoons and a few words. Cartoons of natural changes in sea level rises would be important as well changes in pH.
I would also add number of peope employed as civil servants, non-sceintific advisers etc, etc involved in climate change policies and their costs- including costs of offices, travel, heating , hotel expenses, etc, etc. A cartoon showing where everyone flew ftrom to attend a conference and the costs and then equate with how many people this would feed.
The most basic maths of climate change rapidly reach post A Level standards at which point most people reduce the discussion to noise- they switch off. Josh's cartoon of a drawing to scale showing the depth of the drilling at Balcombe compared to the shallow elevation of water supplies demonstrated the baseless fear mongering. There needs to be much clearer, funnier and shorter explanations why fears over AGW are wrong and government policies are actually going the harm the poorest and enrich some wealthy people.
This from Bob Tisdale
http://wattsupwiththat.com/2013/09/17/a-quick-look-at-the-hadgem2-es-simulations-of-sea-surface-temperatures/
Say no more !