From NCAR:
Future warming likely to be on high side of climate projections, analysis finds
November 08, 2012
BOULDER—Climate model projections showing a greater rise in global temperature are likely to prove more accurate than those showing a lesser rise, according to a new analysis by scientists at the National Center for Atmospheric Research (NCAR). The findings, published in this week’s issue of Science, could provide a breakthrough in the longstanding quest to narrow the range of global warming expected in coming decades and beyond.
NCAR scientists John Fasullo and Kevin Trenberth, who co-authored the study, reached their conclusions by analyzing how well sophisticated climate models reproduce observed relative humidity in the tropics and subtropics.
The climate models that most accurately captured these complex moisture processes and associated clouds, which have a major influence on global climate, were also the ones that showed the greatest amounts of warming as society emits more greenhouse gas into the atmosphere.
“There is a striking relationship between how well climate models simulate relative humidity in key areas and how much warming they show in response to increasing carbon dioxide,” Fasullo says. “Given how fundamental these processes are to clouds and the overall global climate, our findings indicate that warming is likely to be on the high side of current projections.”
The research was funded by NASA.
Moisture, clouds, and heat
The world’s major global climate models, numbering more than two dozen, are all based on long-established physical laws known to guide the atmosphere. However, because these relationships are challenging to translate into software, each model differs slightly in its portrayal of global climate. In particular, some processes, such as those associated with clouds, are too small to be represented properly.
The most common benchmark for comparing model projections is equilibrium climate sensitivity (ECS), or the amount of warming that eventually occurs in a model when carbon dioxide is doubled over preindustrial values. At current rates of global emission, that doubling will occur well before 2100.
For more than 30 years, ECS in the leading models has averaged around 5 degrees Fahrenheit (3 degrees Celsius). This provides the best estimate of global temperature increase expected by the late 21st century compared to late 19th century values, assuming that society continues to emit significant amounts of carbon dioxide. However, the ECS within individual models is as low as 3 degrees F and as high as 8 degrees F (, leaving a wide range of uncertainty that has proven difficult to narrow over the past three decades.
The difference is important to reconcile, as a higher temperature rise would produce greater impacts on society in terms of sea level rise, heat waves, droughts, and other threats.
Clouds are one of the main sticking points, say the NCAR authors. Although satellites observe many types of clouds, satellite failure, observing errors, and other inconsistencies make it challenging to build a comprehensive global cloud census that is consistent over many years.
However, satellites perform better in measuring water vapor, and estimates of the global distribution of relative humidity have become more reliable. Relative humidity is also incorporated in climate models to generate and dissipate clouds.
Fasullo and Trenberth checked the distribution of relative humidity in 16 leading climate models to see how accurately they portray the present climate. In particular, they focused on the subtropics, where sinking air from the tropics produce very dry zones where most of the world’s major deserts are located. The researchers drew on observations from two NASA satellite instruments — the Atmospheric Infrared Sounder (AIRS) and Clouds and Earth’s Radiant Energy System (CERES) – and used a NASA data analysis, the Modern-Era Retrospective Analysis for Research and Applications (MERRA).
The seasonal drying in the subtropics and the associated decrease in clouds, especially during May through August, serve as a good analog for patterns projected by climate models.
“The dry subtropics are a critical element in our future climate,” Fasullo says. “If we can better represent these regions in models, we can improve our predictions and provide society with a better sense of the impacts to expect in a warming world.”
Accurate humidity yields higher future temperatures
Estimates based on observations show that the relative humidity in the dry zones averages between about 15 and 25 percent, whereas many of the models depicted humidities of 30 percent or higher for the same period. The models that better capture the actual dryness were among those with the highest ECS, projecting a global temperature rise for doubled carbon dioxide of more than 7 degrees F. The three models with the lowest ECS were also the least accurate in depicting relative humidity in these zones.
“Because we have more reliable observations for humidity than for clouds, we can use the humidity patterns that change seasonally to evaluate climate models,” says Trenberth. “When examining the impact of future increases in heat-trapping gases, we find that the simulations with the best fidelity come from models that produce more warming.”
The authors focused on climate models used for the 2007–08 assessment by the Intergovernmental Panel on Climate Change. The next-generation models being used for the upcoming 2013–14 IPCC assessment were found to behave in a similar fashion, as described in a preliminary analysis by the authors in a supplement to their paper.
“In addition to providing a path forward and focus for improving models, results strongly suggest that the more sensitive models perform better, and indeed the less sensitive models are not adequate in replicating vital aspects of today’s climate,” write the authors in the paper.
About the article
Title: A Less Cloudy Future: The Role of Subtropical Subsidence in Climate Sensitivity
Authors: John Fasullo and Kevin Trenberth
Journal: Science
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
KR:
It would be hard to find a more clear example of psychological projection than your post addressed to me at November 10, 2012 at 7:33 am. It purports to be a reply to my post at November 10, 2012 at 1:08 am which was a reply to your egregious and mendacious post at November 9, 2012 at 4:20 pm.
My post you purport to reply was a clear and factual response to your typically fallacious, misleading and offensive missive, and yet your reply says to me
Clearly, you don’t know the meaning of argumentem ad hominem.
I gave the facts. I did not rant. And I did not give any insults.
The evidence I presented was clear. Indeed, I numbered each point.
I have as much interest in your “speculations” as I have in your “beliefs”; i.e. none.
Importantly, my “opinions on those two papers” are completely consistent. They are not “self-contradictory” in any way which indicates (IMO) that your assertion is yet another demonstration of your lack of ability at reading comprehension.
Importantly, you repeat the falsehood – which my post refuted – that S&B 2011 used the same method as F&T 2012. They did not. As explanation of the fundamental difference between their methods I need do no more than to quote my post which explained the difference and which you purport to be replying.
Please note that you included this same quotation in your post which I am replying, and that says all anybody needs to know about the validity of your assertions.
I know it is hard for you when you so often display your problems with reading comprehension, but please try to understand what you read before responding to it.
Richard
richardscourtney – Repeating your error does not correct it.
* S&B 2011 evaluated the output of climate models against measurements (temperature and TOA radiation from CERES), and drew conclusions regarding climate sensitivity based on which models outputs are best correlated with those measures.
(Although as Dessler 2011 and others pointed out, they neglected to show the models _they tested_ that did not agree with their conclusions, and their calculations require that clouds force temperatures rather than the other way around)
* F&T 2012 evaluated the output of climate models against measurements (relative humidity from AIRS and CERES), and drew conclusions regarding climate sensitivity based on which models outputs are best correlated with those measures.
It’s a new paper, and will be assessed in due time.
Apparently, though, you cannot or will not recognize the similarities, the common methods, between these two papers. Phrasing your descriptions of the methods slightly differently (with insults) doesn’t change anything in that regard. I believe the similarities are quite clear – and when you treat one as “empirical evidence”, and the other as “pseudoscience”, your confirmation bias is quite evident. To be blunt, you’ve contradicted yourself, and have shown that your conclusions are more important (to you) than the evidence.
—
Based on prior exchanges, I do not expect Richard to admit to any error, and I do not expect any further discussion with him to be productive. But I would ask that readers take a look at what’s going on here. Supporting a methodology when it gives results you agree with, but denigrating it when that _same methodology_ gives results you don’t like? Hmm…
Au revoir
Philip Bradley: But there is a fundamental limit to how much they can be improved.
Let’s say the Forcings model, with currently recognized forcings accounts for 50% of climate variability, then that is the limit of the predictive accuracy of a model that embeds the theory and recognized forcings.
I agree. I accidentally implicitly wrote only of models that have CO2 as a major driver. When I write that “models can be improved” I intend to include models that exclude CO2 in the mix, and the possibility that even models with CO2 will eventually be parameterized such that the effect of CO2 is negligible.
Gail Combs: At this point the emphasis should be on trying to determine the different things that effect the climate. Instead the emphasis is on trying to bury anything that might show that CO2 is not the driving force in the climate.
For the first sentence, I agree. For the second sentence, there is more variety among modelers than you allow for.
richardscourtney: Clearly, I need to explain the matter yet again.
Your explanation was perfectly sufficient the first time. You are just narrow-minded about modeling. You should pharmacokinetics, neuronal modeling, and ecological modeling for other examples of models. Or even a larger section of physics and its history.
oops “you should study pharmacokinetics” etc.
KR:
I notice that you continue your egregious behaviour at November 10, 2012 at 10:06 am where you write to me saying.
I made no error and you have not stated one I made.
You are plain wrong and I have repeatedly explained your error.
And I know from past experience that you will continue to make meaningless noises about it.
Richard
Matthew R Marler:
Your post at November 10, 2012 at 10:54 am confirms you know my correction to your misinformation is correct. Your post says in full
I know my explanation is sufficient. That is why I posted it.
And you know my explanation demolished your misinformation. That is why you have not stated any flaw you have found in it. If you were able to find such a flaw then you would have stated it instead of posting irrelevant and untrue nonsense.
I am very supportive of (n.b. NOT “narrow-minded about”) modelling which is why I object to the climate models being presented as models of the Earth’s climate system when – as I explained – they are not.
Richard
Here is a comment on model accuracy that I put up on Judith Curry’s bloc “Climate Etc”, and I thought I’d repost it here.
The spatio-temporal average temperature is 288K. If the Earth were in equilibrium, as the usual “basic, simple science” assumes, the entire Earth would be 288K all the time. Today’s mean temp of the Arctic is 255K, which I’ll take as the approximate temp of the Antarctic, for illustrative purposes.
The error in each case works out to 33K, which in percentage terms is 33/288 time 100%, or 11%. In lots of fields of study, an 11% error is pretty good, especially when it is near the maximum error. The squared error is 0.0121 (sq pct); the sum of the two squared errors is 0.0242, and the mean squared error, in this case, is 0.0121, and the square root of the mean squared error (RMSE) is 11%. Now, we could do this with every thermometer on earth, and every day, am and pm, and the errors range from 0% (in abs value) up to about 15%. Average the squared error over all the spatio-temporal specific measures, and the RMSE might be 5%. In many fields, that is a really good fit of the model (in this case equilibrium) to the system that is modeled.
However, the modeled change induced by CO2 doubling is something like 1 – 3 K, in the equilibrium model. This is a change of about 0.3% to about 1%, which is much lower than the RMSE of the model. Therefore (though this is empirical, not exactly a logical deduction) It is extremely unlikely that the equilibrium model is sufficiently accurate to make an accurate calculation of the effect of CO2 doubling. Add in other things like the possibly negative feedback of the dynamic effects of clouds, it becomes even less reasonable to think that we have either the correct sign or the correct magnitude of the effect of doubling CO2.
In my experience and in my reading, most people in most professions who use the results of model calculations are very uncomfortable with this kind of reasoning. In medical care, where deviations from modeled values occur all the time, the amount of random variation about the modeled value (e.g. plasma concentration of a drug following a standard dose) is consistently underappreciated.
The paper that is the focus of this thread is a step toward models of the Earth climate that have smaller RMSE. There will never be a model with RMSE = 0, but a model with an RMSE of 1% will at least be approaching something that might be reliable enough for planning purposes.
Matthew R Marler:
Your post at November 10, 2012 at 1:58 pm is meaningless twaddle.
The models are each individually fudged to agree with past rate of global warming by adoption of an assumed amount of ‘aerosol cooling’ which is unique for each model (see my post above at November 9, 2012 at 1:48 pm for referenced explanation of this).
In this circumstance it is not surprising that as you say
The surprising fact is that the fudge fails to provide better agreement than you say.
Each model emulates a different climate system and that is why they are each fudged with a different amount of assumed ‘aerosol cooling’. So, at most only one of the climate models emulates the climate system of the Earth, and there is good reason to suppose that none of them do.
The paper of the above article assumes the models emulate the Earth’s climate system but they don’t and, therefore, the paper is bunkum. Live with it.
Richard
The comments of Richardscourtney continue to provide me with a heads up on contributors to WUWT who clearly know the science concerning the subject under discussion. In addition to KR, I now know that Matthew R Marler is also such a person. If you have not already, I would recommend that everyone reread what these two contributors said – some good science does leak through here at WUWT. And thanks to Richardscourtney for the service he is providing – by alerting us all to those moments with credible science is trying to leak through. Eric
Matthew R Marler says:
November 10, 2012 at 10:46 am
I agree. I accidentally implicitly wrote only of models that have CO2 as a major driver. When I write that “models can be improved” I intend to include models that exclude CO2 in the mix, and the possibility that even models with CO2 will eventually be parameterized such that the effect of CO2 is negligible.
That’s gratifying.
Modelling has its place in investigating any complex system. The main problem I have with the current crop of models is they embed theories with little empirical basis, and the predictive acurraccy of a model cannot be greater than that of the theories it embeds. Well, they can, by fortuitous error, but errors in climate models is a whole other discussion.
I believe better climate models requires general acceptance by modellers that they are only a means to an end, better theories. It’s better theories that will give us better models.
The teenagers who come to the park near our house to fly the model aeroplanes they designed and built have a healthy perspective on models and a sound grasp of the aeronautical principles they have learnt from the design-and-build process. Trenberth et al, not so much.
It is distressing to admit Dr T is a fellow Kiwi.
KR and Matthew R Marler:
You now have clear evidence that you need to ‘raise your game’; i.e.
at November 10, 2012 at 5:06 pm ericgrimsrud says he likes what you have written.
Richard
Richard S Courtney: Your post at November 10, 2012 at 1:58 pm is meaningless twaddle.
Don’t understand MSE, is that it? How are you with R^2 and 1-R^2?
What do you think of the quadratic integrate and fire model for neurons, compared to the Hodgkin-Huxley model? Granted they are not climate models, but you address the general issue of deciding the adequacy of complex models of complex systems. What other techniques do you like for assessing the comparative worth of models: AIC, BIC, F statistic? Do you have a favorite technique for ranking the accuracies of non-nested models?
I am sure that everyone wants to know.
Matthew R Marler:
Your post at November 11, 2012 at 3:53 am attempts to justify the meaningless twaddle of your post November 10, 2012 at 1:58 pm with falsehood, innuendo and irrelevance.
At November 10, 2012 at 2:45 pm I explained why it is meaningless to undertake a statistical analysis of the efficacy of the fudges applied to climate models. The analysis can only provide a misleading indication of the models’ performance.
Your hallmark falsehoods, innuendos and irrelevances do not – and cannot – change that.
Richard
So we are now drawing attention away from the main model argument are we?
1) Aerosols put into models are wrong for the cooling period from the 1940’s to 1970’s.
2) The mechanism involving CO2 requires a positive feedback of water vapor increasing.
3) Increasing water vapor in the atmosphere results in increases in cloud albedo.
4) The model assumes CO2 is reducing water vapor in the tropics with no scientific evidence.
5) The argument that CO2 sensitivity is high with claims of matching it with drier regions in the tropics are bogus, when scientific evidence doesn’t support a CO2 mechanism.
6) Any claims by a model need to be verified with observed scientific evidence corresponding with the scientific method.
7) Scientific evidence supports that CO2 has no effect on reducing water vapor in any regions. Therefore the conclusions of this model are based on natural variations in the tropics.
Richard S Courtney: Your hallmark falsehoods, innuendos and irrelevances do not – and cannot – change that.
I’ll leave it at that.
Well that is clearly a picture of Orion.
“””””…..Philip Bradley says:
November 10, 2012 at 6:53 pm
Matthew R Marler says:
November 10, 2012 at 10:46 am
……………………….
Modelling has its place in investigating any complex system. The main problem I have with the current crop of models is they embed theories with little empirical basis, and the predictive acurraccy of a model cannot be greater than that of the theories it embeds. Well, they can, by fortuitous error, but errors in climate models is a whole other discussion.
I believe better climate models requires general acceptance by modellers that they are only a means to an end, better theories. It’s better theories that will give us better models.
I don’t see ANY distinction between a “theory” and a “model”. Any real “model” will function in exactly the manner described in the “theory”. If the model fails to act as the theory says it should, or acts in ways the theory doesn’t say it should then it is not a model of that theory, but of some other theory.
The thing about models is that they behave in ways described by mathematics, which we invented to describe the operation of the model. It is the real world which the model may not accurately emulate; not the theory. The real world is far too complex to be described by our limited mathematical tools; which is why we costruct models.