Sunday, July 3, 2011

Testing climate models

Climate models are certainly improving, albeit more slowly than many would hope for. They are far from perfect.


Paul Valdes from Bristol University has an interesting paper in the last issue of Nature Geosciences about the reliability of climate models. He basically claims that climate models fail to reproduce past climate, in particular some abrupt climate change that happened in the deep past of Earth's history. This is not really a scoop, as Paul Valdes explains in this interview with Quentin Cooper in Material World, a BBC Radio 4 program. Does it mean that climate change will be worse than so far predicted by climate models ? Answer: we do not know.

37 comments:

Anonymous said...

We do not have the computing power nor the time to input all the necessary algorithmic values, it is an impossible task and imo, ultimately futile - chaotic systems are of incalculable variation.

It makes 'good copy' for the shitehawks at the IPCC and particularly - the big business Carbon traders who all need this AGW scam to go on and on, ergo; "computer models say...." doesn't mean jack shit though in the real world.
But lets face it, AGW is science fiction.

Anonymous said...

As long as one cannot demostrate periodicity (like in ice ages) these kinds of events are impossible to model. If one doesn'¨t put in any constraints for abrupt effects that happen in such a complex system experiencing a steady change, it's not surprising surprises occur. The modelling could possibly be done in a very large grid (say 500*500 km), but the verification of their results would depend only on a few occurrences and thus be a bit uncertain (or at least thay can be claimed to be uncertain) so it's a bit of a lost effort since the prediction of a such model could be verified only afterwards

events that are "never before seen" are outside of experience, and experience is what is used to predict so it's easy to fall into false security

Anonymous said...

Computer models are very useful.

Very useful to impress stupid people.

lucia said...

eduardo--
He basically claims that climate models fail to reproduce past climate, in particular some abrupt climate change that happened in the deep past of Earth's history.
In at least one instance, the difficulty is that the models can't predict the quasi-steady state prior to the abrupt event. Being unable to predict quasi-steady behavior or slowly evolving events ought to be at least as unsettling as the models being unable to predict the abrupt event.

Stan said...

I'm curious why there are so many different climate models. I thought the science was settled.

Freddy Schenk said...

I think that is one crucial point:

In two cases, the models did not adequately capture the basic climate configuration before abrupt change ensued, and in the remaining two examples, to initiate abrupt change the models needed external nudging that is up to ten times stronger than reconstructed. The models seem to be too stable.

I don't know if there might be also a relation to the problem that the (global) models are too zonal - or they are too zonal because of being too stable. However, e.g. ECHAM5 highly improved blocking action compared to ECHAM4 even without increasing spatial resolution. So things are stepping forward.

The desertification of northern Africa. Between about 9,000 and 5,500 yr ago, the region that is now the Sahara was much wetter and supported a steppe-type vegetation. The transition to the current desert state occurred in decades to centuries. Complex climate models fail to simulate the vegetated state, and can not therefore capture this event of rapid change.

I remember from the studies that the position of the "Findlater-Jetstream" (cross-equatorial flow from African tropics to India over E-Africa) in the past is supposed to have suddenly changed together with the monsoon system so that humidity could enter North Africa. Somehow its strange that this jestream is extremely stable now - not only in the model. As this circulation between African tropics and India is highly connected to convective processes (clouds?) - the problem might be here (also resolution problem)?

Whatever, exactly such problems with no green Sahara in the model makes confidence in future scenarios smaller - the Sahara could also get wet and green again when the globe is warming up like in the past? How often and when was the Sahara green in the past? And was it allways warmer? Most interesting would be a warm period with a dry Sahara...

@Stan #5:
I'm curious why there are so many different climate models. I thought the science was settled.

The problem is that they share a lot of similarities, i.e. the models are not really independent...

Punksta said...

"...claims that climate models fail to reproduce past climate...Does it mean that climate change will be worse than so far predicted by climate models ? Answer: we do not know."

Indeed. That climate models are a failure, could just as easily mean that climate change will less serious than predicted by these failed mschanisms. Perhaps even completely unrelated.

Anonymous said...

Dr. Gerhard Löbert, 85598 Baldham, Germany. July 5, 2011.
Physicist. Recipient of the Needle of Honour of German Aeronautics.


Climatological Facts

1. There exists an extremely close correlation between the changes in the mean global temperature and the small changes in the rotational velocity of the Earth in the past 150 years (see Fig. 2.2 of www.fao.org/DOCREP/005/Y2787E/y2787e03.htm), which has been ignored by the mainstream climatologists. (Temperature change is quasi-periodic with a period of some 70 years). (Rotational velocity leads temperature by about 6 years).

This close correlation leaves little room for a human influence on climate. As is shown in Ref. 1, this correlation results from the action of galactic vacuum density waves on the Sun and on the Earth.

2. The Earth's rotational velocity passed a maximum in December 2003, so that temperature should have peaked at about 2009. Actually, as high-precision satellite measurements in the US and the UK show, the mean Earth temperature has been decreasing since 2008.

3. Because of the periodicity of climate change, the mean Earth temperature can be expected to decrease until 2040 to 2045.

4. As every good climatologist knows, there exists a close correlation between solar activity (for example, mean sunspot number) and mean Earth temperature. Low solar activity (as for instance in the Maunder and the Dalton minima) results in low Earth temperature (by the Svensmark process). In a post of July 1, 2011, the Marshall Space Flight Center brought to attention that the current sunspot cycle will probably have the lowest intensity since more than 100 years.

All of this shows that the mean Earth temperature can be expected to continue to decrease until 2040 to 2045.

References
1. www.icecap.us/images/uploads/Lobert_on_CO2.pdf
2. The post of Sept. 19, 2008 in Google "Gerhard, pakteahouse"
3. The posts of March 16, September 26, and October 6, 2009 in Google "Gerhard, pakteahouse"

Werner Krauss said...

This post seems to attract more rude or strange statements than others on klimazwiebel. Lots of scorn, sarcasm, bad language, weirdness. Same is true for the other post "Weather link can no longer be ignored".
Normally, we don't have this kind of guests. I wonder why that is!

Freddy Schenk said...

@Punksta #7
Indeed. That climate models are a failure, could just as easily mean that climate change will less serious than predicted by these failed mechanisms. Perhaps even completely unrelated.

Hm, if you would have drawn the conclusion that climate change might be more serious than projected when models are too stable, OK. The other way round is for sure not the right conclusion. You may argue that e.g. the Sahara will get green instead of desertification as a consequence of global warming. However, what the article says is, that models fail to simulate ubrupt changes - these are rather related to negative consequences.

The best conclusion is maybe just "we don't know". And "don't know" combined with strong anthropogenic pertubations of atmospheric composition and landuse should be even more a reason to be cautious.

Anonymous said...

Do we need the ability of GCMs to predict abrupt climate changes or does the existence of those changes in the past suffice to be concerned?

If Valdes is right in GCMs failing of abrupt climate changes, the consequence would be that the IPCC predictions are a sort of optimistic.

Andreas

Punksta said...

Fredy:
The point is that the models are unreliable for abrupt change (as well as stability). Is essence, useless. So we just don't know - they could be either over- or under-estimating future temperatures.

Punksta said...

Andreas: "If Valdes is right in GCMs failing of abrupt climate changes, the consequence would be that the IPCC predictions are a sort of optimistic"

Yes, the earth may for example cool disastrously.

eduardo said...

Yes, computer models are very complex, but apparently not complex enough to be dismissed by many. Climate models as a product of human ingenuity have drawbacks and deficiencies, but they can boast some success as well. They certainly can be improved. Perhaps the most serious errors by climate modellers is to present climate models as infallible. Very few people do this , but they are quite vocal.

Why are there many climate models? In economics there are also many models. Each central bank hast its own model, in Germany each of the most important economic research institutions has its own model , the IFM has its own model. They deliver different projections, and yet they are considered seriously for political decisions.

Anonymous said...

Mandelbrot thinks that:

http://www.perlentaucher.de/buch/22123.html

"dass die Mathematik des Fraktalen den in der Wirtschaftswissenschaft derzeit gängigen Modellen überlegen ist. Diese gehen, so seine These, von zu viel Stabilität und Normalität aus, während in Wahrheit das Risiko auf dem Finanzmarkt sehr viel größer ist als von den derzeitigen Modellen vorgesehen."

Hans Erren said...

here is a summary of the climate models in 2005:
http://members.casema.nl/errenwijlens/co2/tcscrichton.htm

* Cloud cover is still very badly modeled in both polar regions (fig 4.11, page 50), indicating that moisture transport to the poles is poorly understood.

* All of the coupled models exhibit a “split” intertropical convergence zone (ITCZ) in the Pacific Ocean.

* difficulties remain in the simulation of ENSO (e.g., amplitude, seasonality, and periodicity),

* The models have difficulty in generating the observed level of variability as a function of space and time scale.

* the spatial error indicating the model response was not as strong as observed. This shortcoming was systematic across the models analyzed,

* The discrepancy among models and the Levitus climatolgies are most prevalent in the Arctic Basin.

I wrote in 2005:
"Now this document gives a nice summary of the state of the models, lets see what's improved in five years time."

Are the answers available?

Anonymous said...

Dr. Gerhard Löbert writes:
A mere 1% reduction of cloud cover produces most of the temperature increase of the last 150 years. Do the climate models consider the influence of cloud cover adequately? Probably no.

Do the climate models show the decreased temperatures (relative to the IPCC projection) of the past decade? No.

Can the climate models reproduce the temperatures of the Dalton and the Maunder minima? Probably no.

I think that it is irresponsible to alarm the whole world on the basis of climate models with such deficiencies.

As the world temperature continues to decrease for the next 30 years, the mainstream climatologists will slowly become aware of their false alarm.

Anonymous said...

@ Hans Erren

Yes, models are far from being perfect. But it seems that the importance of models are rather overestimated. In parts of public discussion you could get the impression as if nearly all predictions of climate science were based on models.

But let's listen to the word of one "modeler", James Hansen:

"Well, I had been working with climate models for decades, and I Knew that some of the most recent models predicted ice sheets would grow with global warming, causing sea level to fall, defying common sense and empirical evidence. Models are no better than the representations of processes that are put into them [...] In the case of ice sheets, some of the most important processes were not even included in the climate models.
I prefer to start with paleoclimate, the lessons of history, which provide our best measure of how Earth responds to changing boundary conditions or forcings. Second, as a measure of how rapidly climate can change, we need to look at what is happening now - observations of the ongoing climate response to fast-changing human and natural forcings. climate models come third. Models aid interpretation of past climate, and they are needed to project future changes. So models are valuable, but only when used with knowledge of their capabilities and limitations."

(Hansen, Storms of my Grandchildren, p. 75ff)

Sounds reasonable, doesn't it?

The next question would be: Are models used in IPCC-reports with knowledge of their capabilities and limitations? Maybe Eduardo can say some words on this issue.

Andreas

eduardo said...

@16,

Hans,

we all know that climate models have deficiencies, in some aspects very large deficiencies and modellers are aware of them and trying to improve them all the time. Since it will never be possible to construct a perfect model , not only in climate science but in any other area of science, the deeper question is how bad is bad. For instance, a numerical weather prediction model does not simulate the atmospheric heat transport down my street, and yet it produces reasonable forecast at some spatial and temporal scales. Other example, nobody knows if the solar system is stable (at least when i was a student nobody knew, and 30 later the situation is probably unchanged). Yet, we can launch satellites to space and predict eclipses, etc. So models are not good or bad per se, but good or bad for some specific target.
In the case of climate models, I would agree that deficiencies in ENSO simulations, clouds, etc, are worrying, but nobody really knows if this is really critical for climate change prediction. Probably clouds are critical, probably ENSO not that critical

eduardo said...

@ 17
'Do the climate models consider the influence of cloud cover adequately? Probably no.'

Could you explain why 'probably no' ?

'Can the climate models reproduce the temperatures of the Dalton and the Maunder minima? Probably no.'

Definitively yes, they can. Many publications show these tempertature minima simulated by climate models.


'I think that it is irresponsible to alarm the whole world on the basis of climate models with such deficiencies'

This is your opinion and you are entitle to it. However, the alarm or the warning does not only stem from climate models. It stems from more basic physical reasoning: CO2 is a greenhouse gas and its concentrations is higher now than any time in the last million years. perhaps its effect its tiny, but nobody can be sure that it is.
So while you think there is not reason for alarm- and in some sense I can understand your reasoning. you could also try to understand that others may honestly think that there is a reason to worry - apart from any climate models

Punksta said...

Yes, the CO2 situation is surely a valid area of concern, even of we don't yet know if/how serious it is.

Anonymous said...

Dr. Gerhard Löbert writes:
Reply to eduardo

There is no direct connection between CO2 emission and climate warming. This is shown by the fact that these two physical quantities have displayed an entirely different temporal behaviour in the past 150 years. Whereas the mean global temperature varied in a quasi-periodic manner, with a mean period of 70 years, the CO2 concentration has been increasing exponentially since the 1950's.
The sea level has been rising and the glaciers have been shortening practically linearly from 1850 onwards. Neither time trace showed any reaction to the sudden increase of hydrocarbon burning from the 1950's onwards.

With regard to the Maunder minimum, did the models come out with a temperature decrease of 4 deg. C ?

Freddy Schenk said...

Punksta said... #12
Freddy:
The point is that the models are unreliable for abrupt change (as well as stability). Is essence, useless. So we just don't know - they could be either over- or under-estimating future temperatures.


Like the bold quotation of Hansen in #18 and Eduardo said in #20, models are tools to elaborate questions but they cannot replace our own physical expertise nor do they think for us.

Conversely, if we want to say something about climate, we first rely on physical arguments (sometimes maybe only a reasonable hypothesis) based on more or less basic physics. Often, the models confirm our expectations (as the "known" physics is used in the model), but sometimes not. In most cases of disagreement, the model will be modified. I think it is really a rare exception when it is inverse - that we rewrite physics based on model results. And if, it will not be based on the model results but rather on new physical knowledge being in line with connected physical laws.

Anonymous said...

Eduardo:

So while you think there is not reason for alarm- and in some sense I can understand your reasoning. you could also try to understand that others may honestly think that there is a reason to worry - apart from any climate models

I find it quite reasonable. And acceptable, I would say, for almost every "sceptic". But this is not the way it is being handled in the political discussion. By far not.

In this perspective, with such a doubt, the first question should probably be to address the margin of the "wait and see" option. And since around 2002 - 2005 the margin seems to be growing. But what you hear all the time, and everywhere, in the political arena, is the "there is no time left" motto. At least in Europe. Do you think this should change?

Dr. Gerhard Löbert said...

Eduardo, you haven't commented my remark and my question of July 7 ,12,55 pm.

eduardo said...

@22

Dear Dr. Löbert,

'Whereas the mean global temperature varied in a quasi-periodic manner, with a mean period of 70 years, #

I would not agree with your observation. Do you mean that global temperature has not risen in the last 150 years ? On which observations do you base this assertion? I think it has risen and therefore it has not vary in quasiperidic manner.

'The sea level has been rising and the glaciers have been shortening practically linearly from 1850 onwards. '
this observations would not fit with your previous assertions that temperatures have evolved quasi-periodically. What was the cause for the shortening of glaciers and the rise of sea-level ?

How do you know that sea-level has risen linearly ? If I read the published data, tade gauges indicate a mean rise of 1.8 mm/year during the 20th century, whereas in the last 30 years satellites show a rate of 3 mm/year. Could you please cite the observations you are based on ?

'With regard to the Maunder minimum, did the models come out with a temperature decrease of 4 deg. C ?'

I can produce with a climate model any temperature drop you wish. I only need to reduce the solar constant accordingly. As the value of the solar constant during the LMM is not well known, the temperature drop is no constrain for climate models. It does not demonstrate either that models are good if they can simulate a cooling.

Freddy Schenk said...

#24:
In this perspective, with such a doubt, the first question should probably be to address the margin of the "wait and see" option. And since around 2002 - 2005 the margin seems to be growing.

I don't exactly understand what you mean with "margin" here? Do you mean the 2K-Limit and the "wait and see option" = "business as usual scenario" (BAU)?

But what you hear all the time, and everywhere, in the political arena, is the "there is no time left" motto. At least in Europe.

Some even argue that it is already too late (if you consider the +2K goal). Models can be a useful tool here to study what this could mean - and a new paper exactly motivate to look on the upper 90th percentile of unlikely but still plausible emission scenarios based on energy use:

Sanderson et al. 2011: The response of the climate system to very high greenhouse gas emission scenarios

I found chapter 1 and 2 quite interesting (if you're not dealing with that topic every day). The simulation results themselves appear however very theoretical as very important negative feedbacks like aerosols and an interactive carbon cycle (see e.g. Jungclaus et al. 2010 are not implemented.

However, the paper clearly states possible limitations as the simulation is designed as a simplified sensitivity study. I think it is a good example for what are models made for: experiments under given assumptions.

Dr. Gerhard Löbert said...

Dear Eduardo,

I said QUASI-periodic. After a steep rise before 1872, the mean Earth temperature decreased steadily for about 35 years from 1872 to 1913, then rose steeply for about 35 years from 1913 to 1942, decreased steadily for about 35 years from 1942 to 1974, rose steeply for about 35 years from 1974 to 2009, and will probably decrease steadily for about 35 years from 2009 to about 2043.(see Fig. 2.1 of www.fao.org/DOCREP/005/Y2787E/y2787e03.htm).
In contrast, the CO2 concentration has been increasing exponentially since the onset of hydrocarbon burning in the 1950's.

If you google "sea level chart", Wikipedia shows a chart in which the onset of massive hydrocarbon burning from the 1950s onward does not show up.

While CO2 concentration has been increasing steadily, the mean Earth temperature has levelled off in the last decade and is even decreasing steadily since 2009.

SUMMARY: THERE IS NO DIRECT RELATIONSHIP BETWEEN CO2-EMISSION AND GLOBAL WARMING.

I kindly ask you to read the references given in my contribution of July 5, 10.00 am.

eduardo said...

Dear Gerhard,

I dont agree with the logic of your argument. CO2 is not the only factor that affects temperatures, there are many others superposed to CO2, especially in the 20th century: solar variations, aerosols, land use, urban heat island, ozone, etc.

Consider the following, argumentation: the temperature in Hamburg peaks every day at about 3pm and reaches a minimum every night at about 6am (just rough numbers); CO2 does not display these oscillations ; therefore CO2 cannot affect temperatures. logically this is the same argument you are proposing, but it is clearly wrong.

I did not understand your response well. Are you saying that the temperature has not risen in the last 100 years, and that it has quasi-oscillated around a mean value; or are you proposing that there is a temperature trend superposed on some quasi-oscillation ?

I think that by now the temperature trend in the last 100 years is well establish (?). If you agree what is the reason for this trend in your opinion and how could you prove it ?. You point to a chart showing a linear increased sea-level. What was the reason that caused this sea-level rise ? if sea-level is rising there must be a physical agent that is causing it. which one ? by the way the rate of sea-level rise is not physically related to the global mean temperature, but rather to the global mean heat-flux into the ocean or into the land-ice. Surface temperature and heat-flux are to some extent related but not very directly related.

Finally, temperature in the last decade is not decreasing: all data sets that I am aware of (HadCRU, GISS, NOAA, satellites RSS and UAH do show a small but positive trend, which for one decade alone is not remarkable. We have had much larger and smaller decadal trends in the past, and climate models driven by observed forcings also show this kind of decadal trends, up and down. Even simulations with strong increase of CO2 concentrations in the future show decades with zero trend. So, a trend over a decade does not say much about CO2. and does not say much about models either, for or against them. We have discussed this already quite frequently here at Klimazwiebel. We have also mentioned that if the trend remains small for another decade, then that would contradict climate models. It is still however too soon.

Dr. Gerhard Löbert said...

Dear Eduardo,

If you don't look up the charts that I am indicating to you and if you don't read the references that I am giving to you, there is no sense in continuing our discussion.

Kind regards, Gerhard.

Hans von Storch said...

Herr Löbert,

ich muss Sie ermahnen - wenn Sie hier mit diskutieren wollen, dann haben Sie die Argumente, die angeboten werden, zur Kenntnis zu nehmen und - wenn Sie meinen - versuchen zu widerlegen. Ein einfaches Wiederholen Ihrer eigenen Argumente ohne Berücksichtigung der inhaltlichen Antworten reicht da nicht aus.

Also, entweder: konstruktive Teilnahme am Austausch oder bitte wegbleiben. Das Missionar-artige Verkünden von "Wahrheiten" sollte unterbleiben.

eduardo said...

@ 24
Plaza Moyua,

In the overall discussion there are many sorts of 'skeptics' and 'warmist'. Perhaps one should try to understand the arguments and not just spot weak links to score political points. In the skeptic camp, for instance, some people defend the view that no climate change is possible because God promised not to harm human race again after Noah's arch. Here, some readers are putting forward that ' this correlation results from the action of galactic vacuum density waves on the Sun and on the Earth', something that reminds me of other political astronomic alignments that you may very well know. Should one take those as solid skeptic arguments ? There is wheat and chaff almost every where

jgdes said...

It is misleading to ponder that the models are far from perfect when they are really not even adequate. The usual assumption is that as they are all we have, we should use them for policy, an idea that is fundamentally wrong. More people need to get out and collect real data and use it properly when they get it..

Also the number of models is not unusual; there is competition to come up with an adequate model. However it has been assumed, with no justification whatsoever, that a combination of models gives a better answer than individual models.

Edouardos parallel with economics models is more profound: Many of the economics models were as sophisticated as the climate models, some moreso, and were programmed by usually cleverer, much higher paid people, using more modern languages and techniques. They all fail because they all rely on overly simplistic assumptions. Nobody ever seems to bother to test out these base assumptions withreal data. A parallel indeed.

The other idea being put around - that some models have shown some success - is also flawed, because on those very rare occasions, more models predicted exactly the opposite scenario. Ergo it was just as likely to be pure chance that one or two got something correct.

To improve models, they need to concentrate on spatial variabilities rather than this obsession with predicting future thermageddon.

jgdes said...

Edouardos comment about a zero trend in a decade being meaningless is of course at odds with the fact that we were all previously told to expect that temperatures would be climbing ever higher without any pause because CO2 was supposed to be dominating natural variation by this time. They think we all have as short and convenient memories as they do.

However you can discuss atmospheric variations till the cows come home but the more important metric is the sst where the flat trend since 2003 utterly contradicts the theory, leading th the "missing heat" conundrum. It was not expected, it is unusual and it challenges the catastrophists to come up with ever creative adjustments and physically unrealistic theories to explain it.

All such handwaves about the current pause in heating of course rely on the natural variation that was supposed to be now in decline. It's all a PR game to keep funding going. By this time, by any independent measure, the CO2 domination theory should have been discarded long ago, but that would make the funding dry up.

Hans von Storch said...

jgdes,
would you mind telling me to whom you referring when you assert "we were all previously told to expect that temperatures would be climbing ever higher without any pause"? Who told you? Media? Politicians? NGOs? Scientists? Can you provide names and quotes, please?

Anonymous said...

@ jgdes

"The usual assumption is that as they are all we have, we should use them for policy, an idea that is fundamentally wrong."

I've often heard this argument, but always from skeptics. I think, it's simply not true, see for example my post #18 (Hansen: "Models come third").

In my opinion the strongest argument to be concerned are paleoclimate data. For example:
Last time, when temperatures were 2-3°C above our levels, sea level was about 25m higher. No model prediction, just data.

Andreas

Freddy Schenk said...

@ jgdes #33
However it has been assumed, with no justification whatsoever, that a combination of models gives a better answer than individual models.

That's only partly true. Reto Knutti gave an excellent talk around one year ago at the MPI Hamburg where he exactly evaluated the question, how an ensemble can be estimated best for different models (benchmarking, weighting of ensemble members, number of members etc.). He showed a plot with a skill score as a function of different numbers of models included in an ensemble. As far as I remember, the skill was already decreasing when more than 8 (?) of the best ensemble members were used. Hence, the assumption, that the ensemble of a many “good” models will always represent a better ensemble mean than an ensemble of some “good” models was falsified.

The most critical point is however the definition of "best members" - there are defined criteria e.g. for IPCC ensemble members.

One question of Reto Knutti was e.g. if this is a good idea. The problem might be that the best models for a given validation are not necessarily the best also for the past or future. So it might be that a strongly different model excluded from the IPCC models might be well designed for the past/future. In addition, developers of models are supposed to "tune" their model to fit the IPCC criteria rather than following the pure physics. This would lead to the problem I mentioned before: different models are not necessarily largely independent anymore... This might be a serious problem if e.g. all “accepted” models would just by chance be better than excluded models for a limited validation period due to the same maybe wrong reason…

Unfortunately, I don’t know if our how he included this topic in a paper. Maybe it is done in the first two papers in the section “Articles submitted or in press” in the link above, e.g. from Tebaldi.