Why Models Run Hot: A Discussion of the IPCC's Overestimation

[phyzguy]

I think this is now a legitimate topic. If not, I apologize. I would like to hear comments on the paper "Why models run hot", discussed in this article. Do people on this forum agree that the IPCC models overestimate the amount of positive feedback in the Earth's climate system.? Or is this new article flawed? I've attached a pdf of the article.

 

[Quantum Defect]

I think this is now a legitimate topic. If not, I apologize. I would like to hear comments on the paper "Why models run hot", discussed in this article. Do people on this forum agree that the IPCC models overestimate the amount of positive feedback in the Earth's climate system.? Or is this new article flawed? I've attached a pdf of the article.

Lord Monckton?

 

[PeterDonis]

Lord Monckton?

He's one of four authors of the paper.

 

[PeterDonis]

Do people on this forum agree that the IPCC models overestimate the amount of positive feedback in the Earth's climate system.?

I've thought that for quite some time. It's important to keep the paper's claim in perspective, however. The paper, as I read it, is not saying that its model should be used to predict what the climate will be like at the end of the century; it is not saying that its model is based on a detailed understanding of how the climate works. It is only saying that its very simple model can make better predictions, as compared with the data over the last couple of decades, than the much more complicated IPCC models, and that should make us skeptical that the IPCC models are modeling things correctly.

 

[Bandersnatch]

I'd be wary of papers whose primary author is a politician with no relevant scientific background and a clear agenda.

 

[PeterDonis]

I'd be wary of papers whose primary author is a politician with no relevant scientific background and a clear agenda.

By PF rules, we should focus on the scientific content of papers and not on who authored them, provided they meet the basic guidelines for acceptable references. This paper was published in a peer-reviewed journal, and it looks like at least two of the four authors have scientific credentials. So it appears to be an acceptable reference.

 

[Bystander]

saying that its very simple model can make better predictions, as compared with the data over the last couple of decades, than the much more complicated IPCC models,

May I paraphrase: "a very simple model provides a better fit of data collected over the past two decades than do complex models that have been generally accepted." Implication: the complex models are "overfitting" the available data.

 

[PeterDonis]

May I paraphrase: "a very simple model provides a better fit of data collected over the past two decades than do complex models that have been generally accepted."

Yes, this is better because it avoids the word "prediction", which doesn't really fit here. (The paper also includes predictions based on the simple model, but those aren't what the claim that the complex models are overestimating feedback is based on.)

Implication: the complex models are "overfitting" the available data.

More precisely: the complex models overfit the data from prior to the past couple of decades, i.e., the data that were used to build the models. Hence, the model results for the past couple of decades don't match as well.

 

[DaveC426913]

Is climate change / global warming no longer a verboten topic on PF? It was for the longest time.

https://www.physicsforums.com/threads/physics-forums-global-guidelines.414380/#post-3929007
Apparently it is fair game again. :)

 

[Bystander]

no longer a verboten

There's a note at the head of the forum regarding this. Stick to refereed sources and omit the politics.

Beg pardon --- there was --- do not ask me where it went.

 

[phyzguy]

I have thought for a while that the IPCC climate models seem to have an unreasonably large amount of positive feedback. It appears, based on what PeterDonis said in Post #4, that he agrees. My logic was that if the climate system had a large amount of positive feedback, we would have seen more climate variability in the past due to various forcings. The authors of this paper say exactly the same thing at the beginning of Section 8.3.2, where they say,

"A plausible upper bound to f may be found by recalling that absolute surface temperature has varied by only 1 % or 3 K either side of the 810,000-year mean [40 , 41 ]. This robust thermostasis [42 , 43 ], notwithstanding Milankovich and other forcings, suggests the absence of strongly net-positive temperature feedbacks acting on the climate."

It seems possible that the developers of the IPCC models incorporated enough positive feedback to fit the temperature data through the early part of the 21st century, and are now finding that the positive feedback needs to be dialed back in order to fit the more recent data. Does anyone know whether there are other checks on the various components of feedback to verify if they are reasonable?

 

[Bystander]

checks on the various components of feedback

How many "feedbacks" have been proposed? Haven't bothered to keep up with them lately. Have they been run independently of one another? Or, do the models allow a "synergy" between/among them?

 

[PeterDonis]

Does anyone know whether there are other checks on the various components of feedback to verify if they are reasonable?

I think part of the issue is that we don't have a good understanding of the mechanisms behind the various feedbacks, so it's hard to check them independently. For example, I don't think there's agreement even on the sign (positive or negative) of the net feedback from clouds, because we don't understand the various effects clouds have well enough to know which ones dominate.

There are also people working on estimating the feedbacks by looking at satellite data; this is basically doing it "from the outside", not testing the various mechanisms separately but just looking at the overall radiation in and out to estimate global effects. AFAIK there is still a lot of spread in the possible feedback values estimated this way as well--largely because we simply haven't been taking satellite data that long.

 

[berkeman]

There's a note at the head of the forum regarding this. Stick to refereed sources and omit the politics.

Beg pardon --- there was --- do not ask me where it went.

Thanks for the heads up! The stickie is back visible again. :-)

 

[Bystander]

back visible again

Thank you --- I'm not having a "golden age moment."

 

[phinds]

Thank you --- I'm not having a "golden age moment."

Ah you probably are and berkeman is covering for you

 

[Bystander]

Ah you probably are and berkeman is covering for you :)

You'll spend the rest of your days proofreading every line you post --- one typo and you're done for.

 

[Bystander]

looking at satellite data;

I got the impression while hunting for information for Loudzoo's thread on IR & ocean evaporation that the satellite data wasn't broad spectrum. Got any details whether they're collecting UV, vis., and IR?

 

[PeterDonis]

Got any details whether they're collecting UV, vis., and IR?

I'm pretty certain satellites can get radiance measurements in the microwave and IR bands. I'm fairly sure they can get them in visible as well. Don't know about UV. As far as using satellite data to infer temperatures, since all of the temperatures in question correspond to spectrum peaks in the IR range, I'm not sure UV data would add much anyway; if you've got microwave, IR, and visible, you've covered the key portion of the spectrum.

 

[Bystander]

I'm not sure UV data would add much anyway

Thee or me either one. I haven't actually sat down and designed a "dream data set" for satellite measurements, let alone whether instrument sensitivity is adequate for my "intuitive" sense that "global" measurements are probably going to be most meaningful if made from at least lunar distance. Sorting direct reflection from scattering? Reflected IR from emitted IR? Which details are even meaningful. Hopefully NASA, NOAA, and JPL have coordinated programs for such purposes, and one of these days it'll show up in "J. Gee. Whiz" for me to play with.

 

[klimatos]

For example, I don't think there's agreement even on the sign (positive or negative) of the net feedback from clouds, because we don't understand the various effects clouds have well enough to know which ones dominate.

I believe that there was a recent AGU study by Robert Shibatani on the 31 known major aerosols that showed a net positive effect of some four watts per square meter. Regardless, we probably have less valid information on albedo forcing that any other major forcing method. We need a sensor that can read the entire planetary disc at one time, and we need data from this sensor for a minimum of thirty years. Not in my lifetime, and probably not in my children's.

 

[D H]

Do people on this forum agree that the IPCC models overestimate the amount of positive feedback in the Earth's climate system.? Or is this new article flawed?

This paper is so very bad that it will receive a good number of citations in the next few months by later papers that show just how bad this paper is. It is so monumentally bad that I won't be at all surprised if it makes its way onto retractionwatch.com.

First, for the claim that "models run hot." This is a false claim. The authors went all the way back to the first assessment report as the basis for their claim that "models run hot" (which they don't). Granted, that very first report had some flaws; it was the very first report, after all. That first assessment did not use tools such as general circulation models, which were in their infancy in the late 1980s. Note that this means that even if the paper by Monckton et al. is correct (which it isn't), that they target the FAR is a fatal flaw in their paper. Later assessments used general circulation models to drive the analyses. The switch to better models in later reports means that that the authors arguments apply only to that very first report.

Even with regard to that very first report, the claim "models run hot" is a false claim. The first assessment presented warming under multiple scenarios, the worst case being the "business as usual" scenario. Monckton et al. cherry-picked that "business as usual" scenario as if it were the report's sole prediction. This scenario assumed cars would continue to have 1990 gas mileage, coal power plants would continue to have 1990 emissions, and that the economy and population would continue growing. That didn't happen. There have been modest improvements in technology since 1990, populations in the developed world have gone stagnant, and the world was hit by two huge economic downturns. The world pumped less CO₂ into the atmosphere than predicated by the "business as usual" scenario, so of course that scenario "ran hot."

Another flaw in the paper is their use of an 810,000 year baseline as proof that positive feedbacks don't exist. That's completely wrong. The Milankovich forcings are rather small. The slight reductions in solar irradiation that result from the Milankovich cycles are too slight in and of themselves to result in a glaciation. We wouldn't have glaciations without positive feedbacks.

The biggest flaw in the paper is the authors silly Figure 5 and the arguments around it. That a sound system designer wouldn't create a system with loop gain of more than 0.1 means that the climate mustn't have a gain greater than 0.1 is a ludicrous non sequitur.

 

[Bystander]

The slight reductions in solar irradiation that result from the Milankovich cycles are too slight in and of themselves to result in a glaciation. We wouldn't have glaciations without positive feedbacks.

I'm missing something here --- or am I?

1) Global mean temperature is a function of some unknown set of feedback parameters which are also functions of global mean temperature?
2) Gains for the feedbacks can be greater than, equal to, or less than zero, but were a net positive to produce however many glaciations over the past 800 ka, and to result in whatever is taking place now?
3) Positive feedback is understood to mean that the effect of whatever "upset" condition is introduced into the global system is amplified over time until some opposite "upset" offsets it, or some other damping mechanism takes over?​

 

[PeterDonis]

The world pumped less CO2 into the atmosphere than predicated by the "business as usual" scenario, so of course that scenario "ran hot."

If I'm not mistaken, though, the "business as usual" scenario is not the only one that has over-predicted actual warming; all of them have, though not all to the same extent (obviously). And the actual CO2 increase, while it has been less than the "business as usual" scenario, as you say, has, if I'm not mistaken, been greater than rise assumed in the other scenarios (except for the "extreme" scenario that used even more CO2 rise than "business as usual", but AFAIK that one is not considered realistic). So I think it is still fair to say that the models "run hot", even if the specific comparison in the paper is not comparing apples to apples.

 

[D H]

I'm missing something here --- or am I?

Quoting from the silly paper,

A plausible upper bound to f may be found by recalling that absolute surface temperature has varied by only 1 % or 3 K either side of the 810,000-year mean. This robust thermostasis, notwithstanding Milankovich and other forcings, suggests the absence of strongly net-positive temperature feedbacks acting on the climate.​

This is sheer nonsense. What "robust thermostasis"?

The Milankovich forcings are very small, about 0.5 watts/meter². That these tiny forcings result in huge temperature swings (minor problem: the 3 K figure itself is a bit misleading; a better figure is swings of 10 K) is evidence to most climatologists that huge positive feedbacks are involved. The feedbacks are obvious. Warmer winters at high latitudes lead to more snow at those latitudes, which increases the albedo. Cooler summers at high latitudes means some of that increased snowfall becomes permanent ice. Permanent ice begets more permanent ice and even cooler summers because of the increased albedo. Cooler summers means reduced absolute humidity, and water is by far the most significant greenhouse gas. More cooling, more snow, more ice. The cooling also reduces CO₂ levels, leading to even more cooling.

The flip side, an interglacial, is also attributable to positive feedbacks. Just as the Earth couldn't enter a glaciation without positive feedbacks, it couldn't exit one without positive feedbacks.

 

[PeterDonis]

That these tiny forcings result in huge temperature swings (minor problem: the 3 K figure itself is a bit misleading; a better figure is swings of 10 K) is evidence to most climatologists that huge positive feedbacks are involved.

I wondered about that 3 K figure too; I had thought it was closer to 10 K as well.

I think it's worth clarifying a point here, though. If we're looking at the start of a glaciation, due to a small negative Milankovich forcing for example, then there can't just be positive feedback involved, because if there were the climate would never stabilize in the glaciation phase; it would just keep getting colder. At some point negative feedback must come into play in order to stop the cooling and stabilize the climate in the ice age state. (Of course similar remarks apply to the transition from ice age to interglacial.) So both types of feedbacks must be present.

Basically, on the theory you are presenting here, the climate is bi-stable: it has, roughly speaking, two stable states (ice age and interglacial), and if the climate is in one of the stable states, it takes a certain amount of forcing to kick it out of that state. Once that amount of forcing is present, positive feedback dominates and moves the climate towards the other stable state; but once it gets close to the other stable state, negative feedback dominates and the climate stabilizes in the other state.

The question then becomes: what if you start with the climate in the interglacial stable state, but you then add positive forcing--causing warming (instead of negative forcing like the Milankovich forcing, causing cooling)? There seems to be a belief that, once there is enough positive forcing present, the same thing will happen that happens when positive forcing kicks the climate out of the ice age state--that positive feedbacks will dominate and accelerate warming. However, the interglacial state is very different from the ice age state, so it's not clear to me that all the same positive feedbacks will be present in the warming direction--note that these must be different from the positive feedbacks in the cooling direction (like ice-albedo feedback) that you describe.

 

[Bystander]

tiny forcings result in huge temperature swings (minor problem: the 3 K figure itself is a bit misleading; a better figure is swings of 10 K)

Over what time scales? Patience, gang --- I'm trying to get back up to speed on just what points are being pursued on the bases of which arguments.

 

[D H]

If I'm not mistaken, though, the "business as usual" scenario is not the only one that has over-predicted actual warming; all of them have, though not all to the same extent (obviously).

That depends on which dataset one uses.

The table below summarizes the warming since 1990 per various temperature datasets.

Source     °C/decade
Satellite-based
RSS        0.11
UAH        0.16
Surface-based
GISTEMP    0.15
NOAA       0.14
HADCRUT4   0.14
Berkeley   0.16

One of the above datasets stands out, in a Sesame Street sense ("one of these things does not go with the other; one of these things doesn't belong"). The silly article used that singular source, the Remote Sensing Systems (RSS) dataset. So-called "skeptics" love RSS. Everyone else says "Hmm. That's funny. Maybe I shouldn't use that."

When one looks at CO₂ forcings based on observed CO₂ levels since 1990, the closest match is scenario B. With the exception of RSS, the observed data are most consistent with scenario B from the IPCC's First Assessment Report.

Note well: I intentionally omitted the uncertainties in the above table. The silly article did the same, so tit for tat. The uncertainties are rather large. All of the datasets are consistent with one another, to within two standard deviations.

After accounting for those uncertainties, each of those observation based datasets is consistent with each of the FAR scenarios. The uncertainties are just too large over a mere 25 year span to rule out anyone of those scenarios. The only thing that can be excluded (with the exception of RSS) is that there has been no warming whatsoever.

The RSS database exhibits the greatest uncertainty. It is consistent with both "business as usual" and "no warming whatsoever." The null hypothesis, "no warming whatsoever," is easily rejected by all but the RSS database.

 

[PeterDonis]

The uncertainties are rather large. All of the datasets are consistent with one another, to within two standard deviations.

After accounting for those uncertainties, each of those observation based datasets is consistent with each of the FAR scenarios. The uncertainties are just too large over a mere 25 year span to rule out anyone of those scenarios.

This seems to me to be the most important piece of information to keep in mind.

 

[klimatos]

The Milankovich forcings are very small, about 0.5 watts/meter².

I was under the impression that the Milankovich forcings (orbital forcings) were independent of any change in the solar constant. The incoming annual global radiation remains constant but the latitudinal and seasonal distribution of this radiation varies with the regular variations in the three controls.

The flip side, an interglacial, is also attributable to positive feedbacks. Just as the Earth couldn't enter a glaciation without positive feedbacks, it couldn't exit one without positive feedbacks.

True without doubt.

 

[Czcibor]

First, for the claim that "models run hot." This is a false claim. The authors went all the way back to the first assessment report as the basis for their claim that "models run hot" (which they don't). Granted, that very first report had some flaws; it was the very first report, after all. That first assessment did not use tools such as general circulation models, which were in their infancy in the late 1980s. Note that this means that even if the paper by Monckton et al. is correct (which it isn't), that they target the FAR is a fatal flaw in their paper. Later assessments used general circulation models to drive the analyses. The switch to better models in later reports means that that the authors arguments apply only to that very first report.

Even with regard to that very first report, the claim "models run hot" is a false claim. The first assessment presented warming under multiple scenarios, the worst case being the "business as usual" scenario. Monckton et al. cherry-picked that "business as usual" scenario as if it were the report's sole prediction. This scenario assumed cars would continue to have 1990 gas mileage, coal power plants would continue to have 1990 emissions, and that the economy and population would continue growing. That didn't happen. There have been modest improvements in technology since 1990, populations in the developed world have gone stagnant, and the world was hit by two huge economic downturns. The world pumped less CO₂ into the atmosphere than predicated by the "business as usual" scenario, so of course that scenario "ran hot."

Can you advice me any comprehensive (meta)study concerning which climate change was predicted and which temperature was actually observed afterwards?

 

[PeterDonis]

I was under the impression that the Milankovich forcings (orbital forcings) were independent of any change in the solar constant.

As I understand it, the Milankovich forcings do include changes in the solar constant; as a result of periodic changes in the Earth's orbital parameters, the annual radiation received by the Earth from the Sun changes. (The latitudinal and seasonal distribution also changes.)

 

[mheslep]

...I would like to hear comments on the paper "Why models run hot"

That Sci Bull paper is not the first journal published material to say as much.

http://www.nature.com/nclimate/journal/v3/n9/full/nclimate1972.html#access, Fyfe, Gillett, Zweir
Nature Climate Change 3, 767–769 (2013) doi:10.1038/nclimate1972, Published online, 28 August 2013

Recent observed global warming is significantly less than that simulated by climate models. This difference might be explained by some combination of errors in external forcing, model response and internal climate variability.

Figure 1: Trends in global mean surface temperature

a, 1993–2012. b, 1998–2012. Histograms of observed trends (red hatching) are from 100 reconstructions of the HadCRUT4 datasethttp://www.nature.com/nclimate/journal/v3/n9/full/nclimate1972.html#ref1. Histograms of model trends (grey bars) are based on 117 simulations of the models, and black curves are smoothed versions of the model trends. The ranges of observed trends reflect observational uncertainty, whereas the ranges of model trends reflect forcing uncertainty, as well as differences in individual model responses to external forcings and uncertainty arising from internal climate variability.

 

[phyzguy]

mhslep - thanks for posting this. I thought that it was generally acknowledged that the models were predicting more warming than we are observing, and so I was surprised by DH's vehement denial in post 22. I though that's why we see many articles trying to explain the "hiatus" in warming. This Nature Climate Change article seems to agree that the models are significantly over-predicting the amount of warming. DH - I expect you must disagree with the plots mhslep has posted. Can you tell us why?

 

[mheslep]

The "Business-As-Usual" scenario, as described in the FAR, defines man-made carbon emissions over time in the https://www.ipcc.ch/ipccreports/far/wg_I/ipcc_far_wg_I_spm.pdf. The carbon emissions BAU graph, as I can make out, gives something less than 10 GtC/yr for 2015, maybe as low as 9 GtC/yr, while the low carbon scenarios are still around 5 GtC/yr as of 2015. As of 2011 per AR5, global carbon emissions from fossil fuels and cement were 9.5 GtC/yr +/- 0.8 (http://www.ipcc.ch/pdf/assessment-report/ar5/wg1/WG1AR5_Chapter06_FINAL.pdf), growing at 3%/yr.