Moderated Global Warming Discussion

[W.D.Clinger]

I'd like to see hindcasting results from DePreSys as any model which allows for a tighter range of outcomes for specific regions is very valuable.

So what did you think of the hindcasting results shown in the paper and in its supporting online material?

 

[macdoc]

Not chance to read it - watching Obama win Is there an executive summary available? I'll accept your read on it.

 

[W.D.Clinger]

Not chance to read it - watching Obama win Is there an executive summary available? I'll accept your read on it.

Smith et al. compare DePreSys against NoAssim, "which is identical to DePreSys but does not assimilate the observed state of the atmosphere or ocean." DePreSys does significantly better at all time scales (up to nine years), both globally and regionally. Figure 3D reveals that the only regions in which the 9-year NoAssim hindcast of average mean surface temperature did better were in the oceans immediately south of Africa, west of the Americas, and east of Greenland. Even in those regions, the difference was very small.

The Fildes and Kourentzes paper is more relevant if you're interested in comparing the predictive skill of DePreSys against purely statistical models such as random walks. DePreSys has about half the error of random walks at both the 1-4 year and 10 year time scales. Combining DePreSys with purely statistical models offered significant improvement in only one instance out of eight. Fildes and Kourentzes try to make a big deal out of that, but one out of eight sounds to me like it could be a chance occurrence (which they refer to as "data snooping"). (As you can see from my previous review, I was not impressed by the care they took in drawing conclusions from their results.) Their section 3.3, which is the part that's getting all the love from denier blogs, doesn't compare to DePreSys at all; they switch to comparing against an unnamed global climate model used by Koutsoyiannis et al.. (Some of the denier blogs mistakenly assume the results of section 3.3 apply to DePreSys.) So Fildes and Kourentzes don't have anything useful to say about using DePreSys for regional predictions.

 

[macdoc]

Thanks.

 

[Trakar]

Now, hopefully, science will gain a seat at the table.

 

[macdoc]

You mean with the Obama win....could be.

 

[lomiller]

Hmm.

Should I believe you? Or should I believe the guys who published this in Science:

back-peddle noted

This supports my contention that whatever value exists here is in the original work by DePreSys not you contention that comparing that comparing it's performance to statistical models is somehow important.


Noe about that, are you ever going to provide some positive evidence for your assertion that comparing statistical models to physical models over short periods has value, and are you going to tell us what that value is or are you going to continue to dodge the issue?

 

[lomiller]

It was lomiller who attributed the predictive skill of DePreSys entirely to its ability to predict ENSO.

False. Please stick to things that have actually been said, and oh yeah. start supporting your claim as about the value of comparisons to statistical models with positive evidence.

 

[CapelDodger]

Despite that contrary evidence, you (and CapelDodger) appear to have accepted that aspect of McIntyre's argument.

My apologies, I have been somewhat off-handed due to distractions and a general feeling that it's all rather trivial. I unreservedly withdraw any opinions I may have expressed.

 

[spin0]

The Munich Re report is a major new event. Many casual deniers (those who simply can't believe it's true for ideological reasons rather than any other reasonable doubt) may possibly be swayed by economic rather than scientific evidence.

Interesting article in NYT: A Change in the Weather on Wall Street

Every year, ISO Property Claim Services, which does research and risk consulting for property and casualty insurers, measures insured losses in the United States due to natural catastrophes. Last year the figure was $32 billion. This year Sandy alone may cost $50 billion — not to mention damage caused by record heat, record drought and record wildfires. And “insured losses” covers property and casualty insurance only. It doesn’t include crop insurance — this year about a $20 billion to $25 billion loss, and taxpayers will cover all but $5 billion of that. It doesn’t include payouts by the national flood insurance program — at least $1 billion this year even before Hurricane Sandy.

And of course, it doesn’t include the expense of replacing destroyed tunnels, bridges, subways and coastal defenses. Nor the costs of cleanup, business interruption (the largest loss component of 9/11), or higher prices — drought is a big reason we are paying more for food and cotton clothing. Nor the soaring price of insurance. A report by Ceres, an organization started after the Exxon Valdez oil spill to promote sustainable investment and business practices, details the dire threat climate change poses to the insurance industry. The threat persists even though insurers have been pulling out of the highest-risk markets entirely. In those places, states have become the insurer of last resort — and some of the payouts will come from taxpayers.

...

What Hurricane Sandy has made plain is that there is a balance. We’re already paying for climate change, and that tax is only going to rise. Advocates for action can make the same argument as those who mock climate change: to do nothing is to be anti-jobs, anti-growth, anti-American values. “We are starting to see some acknowledgement that the numbers are enormous, but not nearly as much as we should,” said Mindy Lubber, the president of Ceres. “This is a fundamental economic issue. There aren’t a lot of things that have shut down the stock exchange.”

 

[macdoc]

First Big Tobacco pays through the nose for knock on effects despite decades of denial fueled by some of the same clowns in the AGW denial biz.

Is Big Oil and Coal next at the bar being judged for consequences? Exxon Valdez was a tiny event compared to the health cost of coal alone let alone vehicle emissions.

That said I suspect it's the actual operators of coal plants at risk from class action tho Koch's denial industry involvement might actually get them into the class action line up.

It took the courts to final get solid action on S02 - maybe C02 mitigation from sources will be less costly than the subsequent litigation for consequential damages.

 

[W.D.Clinger]

how science is done

Irrelevant as DePreSys only passed tests for skill in the first 1-4 years.

Hmm.

Should I believe you? Or should I believe the guys who published this in Science:

back-peddle noted

You said something about DePreSys. I noted that the authors of DePreSys contradict you. You called that a "back-peddle".

Had you called it a back-pedal, I'd say you don't know the meaning of the word. As it is, I don't know the meaning of your word.

This supports my contention that whatever value exists here is in the original work by DePreSys not you contention that comparing that comparing it's performance to statistical models is somehow important.


Noe about that, are you ever going to provide some positive evidence for your assertion that comparing statistical models to physical models over short periods has value, and are you going to tell us what that value is or are you going to continue to dodge the issue?

It has to do with how science is done.

Scientists are always looking for ways to test their hypotheses, theories, and models. The fancy-pants word for this is "falsification". The popular literature would have you believe falsification involves two theories meeting on the street at high noon. When the dust clears, only one is left standing.

With exceedingly rare exceptions, that's not how science works. In reality, falsification is mostly about testing minor details and the limits of theories and models. That's especially true with models of complex systems, which are useful precisely because they are greatly simplified abstractions of the systems they model. We know these models must have limited domains of applicability. To learn what those limits are, the models must be tested.

DePreSys is that kind of model. Its avowed purpose is to explore the feasibility of near-term climate predictions, up to a decade in advance. To learn how well (or even whether) DePreSys works, and at what time scales, it must be tested.

Smith et al. tested the predictive skill of DePreSys by comparing its hindcasts to a deliberately crippled version of DePreSys itself. DePreSys did significantly better. That establishes some degree of predictive skill, provided the crippled version wasn't doing worse than chance. It seems reasonable to assume that proviso, because doing significantly worse than chance would itself require anti-predictive skill, but it isn't hard to understand why scientists prefer tests that lead to more direct inference of predictive skill.

Fildes and Kourentzes performed that more direct test. To perform that test, they needed data that could be provided only by the authors of DePreSys. Did Doug Smith and the other authors say this new testing was redundant? Did Doug Smith object because Fildes and Kourentzes "lacked the requisite understanding of the subject" and don't have experience with physical modeling? Did Doug Smith complain about Fildes's association with "an off discipline journal"?

Evidently not:

We would like to thank Doug Smith for both supplying us with the data on which this paper is based and helping us to translate the language of climatologists into something closer to that of forecasters.

Doug Smith wasn't surprised when Fildes and Kourentzes found that DePreSys has considerably more predictive skill than random walks at time scales going out to ten years, but some people (you, for example) were surprised by that result. The authors of denialist blogs would also have been surprised, had they read and understood the paper whose results they're misrepresenting.

I doubt whether very many climate scientists were surprised. If any were, they learned something. In science, learning something is a good thing.

 

[Mikemcc]

Anybody and their dog will tell you that near-term climate prediction is a waste of time and effort. It's too noisy. It really, doesn't matter if one modelling approach thinks it's better or worse with respect to decadal predictions, next year it could be completely different. It really is mental masturbation. Decadal simulations and observations are on too short a timebase to reliably show anything, it's too open to volcanic events such as Pinatubo, or particularly strong or weak ENSO events (such as the denialists favourite year of 1998).

 

[W.D.Clinger]

Anybody and their dog will tell you that near-term climate prediction is a waste of time and effort.

So the MET office is nobody and a cat?

Science and Nature Geoscience may be off discipline journals in this thread, but some scientists still take them seriously.

 

[bobwtfomg]

Too late for 2 degrees?
report from international buisiness consultants PriceWaterhouseCooper
http://www.pwc.com/gx/en/sustainability/publications/low-carbon-economy-index/index.jhtml

The 2011 rate of improvement in carbon intensity was 0.8%. Even doubling our rate of decarbonisation, would still lead to emissions consistent with six degrees of warming. To give ourselves a more than 50% chance of avoiding two degrees will require a six-fold improvement in our rate of decarbonisation.

 

[lomiller]

You said something about DePreSys. I noted that the authors of DePreSys contradict you. You called that a "back-peddle".

Indeed I did say something about DePreSys. I said it had value, but more in the realm of theoretical research then practical climate science as the practical usefulness is uncertain.

The back-peddle here is that you were originally saying the Fildes and Kourentzes paper had provided this result, and I had already informed you that it was in fact DePreSys who were responsible for both the predicting and testing needed to achieve the results he was speaking of. Fildes and Kourentzes are irrelevant to this.

Its avowed purpose is to explore the feasibility of near-term climate predictions, up to a decade in advance. To learn how well (or even whether) DePreSys works, and at what time scales, it must be tested.

There are established methods to test the skill of a model. DePreSys had already performed such tests.

Fildes and Kourentzes performed that more direct test.

Fildes and Kourentzes compared DeRreSys to another model, but model skill is something that stands on it’s own merits. You do not need to compare one model to another to determine its skill.

Doug Smith wasn't surprised when Fildes and Kourentzes found that DePreSys has considerably more predictive skill than random walks

Why would they be surprised? DePreSys had already done skill testing of their results.

 

[lomiller]

Anybody and their dog will tell you that near-term climate prediction is a waste of time and effort. It's too noisy. It really, doesn't matter if one modelling approach thinks it's better or worse with respect to decadal predictions, next year it could be completely different. It really is mental masturbation. Decadal simulations and observations are on too short a timebase to reliably show anything, it's too open to volcanic events such as Pinatubo, or particularly strong or weak ENSO events (such as the denialists favourite year of 1998).

Yes and no. DePreSys seems to have demonstrated that they can make some model predictions with some measure of skill. What’s less certain is the usefulness of such predictions,. Since the models are functioning is a much different way than normal, it doesn’t follow that showing skill at decadal predictions means anything for the longer term predictions where really important things are happening.

I posted this a page or two back but here is a blog posting from NASA climate modeller Gavin Schmidt:

http://www.realclimate.org/index.php/archives/2011/08/cmip5-simulations/

The ‘decadal prediction’ simulations are mostly being run with standard GCMs (see the article by Doblas-Reyes et al, p8). The different groups are trying multiple methods to initialise their ocean circulations and heat content at specific points in the past and are then seeing if they are able to better predict the actual course of events. This is very different from standard climate modelling where no attempt is made to synchronise modes of internal variability with the real world. The hope is that one can reduce the initial condition uncertainty for predictions in some useful way, though this has yet to be demonstrated.

It’s an interesting emerging field, but more academic than practical at this point.

 

[lomiller]

Science and Nature Geoscience may be off discipline journals in this thread, but some scientists still take them seriously.

Fildes and Kourentzes didn't publsih their paper in Science on in Nature

 

[W.D.Clinger]

back-pedaling tedium

You said something about DePreSys. I noted that the authors of DePreSys contradict you. You called that a "back-peddle".

Indeed I did say something about DePreSys. I said it had value, but more in the realm of theoretical research then practical climate science as the practical usefulness is uncertain.

Anyone who cares can examine that sequence of posts and verify that I quoted and was referring to the following statement of yours:

Irrelevant as DePreSys only passed tests for skill in the first 1-4 years.

Returning now to what you wrote immediately after the first quotation above:

The back-peddle here is that you were originally saying the Fildes and Kourentzes paper had provided this result, and I had already informed you that it was in fact DePreSys who were responsible for both the predicting and testing needed to achieve the results he was speaking of. Fildes and Kourentzes are irrelevant to this.

The obvious back-pedal here is (1) you explicitly denied that DePreSys had shown predictive skill beyond four years, but (2) you now appear to be saying you had already informed me that DePreSys had shown predictive skill beyond four years.

Another back-pedal: You started out by saying the physical systems behave like random walks on decadal time scales. After I devoted rather more posts than should have been necessary to demonstrate otherwise, you are now pretending to have informed me of that result.

In any event, I do not believe the DePreSys group had ever shown that its predictions were significantly better than predictions made by purely statistical models such as random walks. That was the primary result reported by Fildes and Kourentzes. It is an unsurprising result, but it contradicts a claim you had repeated several times in this thread.

Science and Nature Geoscience may be off discipline journals in this thread, but some scientists still take them seriously.

Fildes and Kourentzes didn't publsih their paper in Science on in Nature

I never said they did. (Note also that Nature Geoscience is not quite the same journal as Nature.)

You edited my words to remove the links. What I actually wrote was:

Science and Nature Geoscience may be off discipline journals in this thread, but some scientists still take them seriously.

I was responding to Mikemcc, who had said "Anybody and their dog will tell you that near-term climate prediction is a waste of time and effort." If you follow the links you took such pains to delete, you will see that editors and peer reviewers for Science and Nature Geoscience disagree with Mikemcc's assessment.

Editing my statement so you could pretend I was saying something else looks dishonest to me. Can you provide an alternative explanation?

 

[truethat]

Ok there is no way in heck I'm reading through 100 pages plus so I'm just plopping this question here.

Global warming seems like the planet doing it's thing. What I don't understand is why people say that human production, waste or consumption has made a dent on the planet. Pollution I can understand. Killing habitats and whatnot, that I get.

But as far as heating up or cooling down the planet

The planet has been cooled down in the past by huge volcanic activity. HUGE. Lowering the earth's temperture 22 degrees huge.

http://www.newscientist.com/article/mg18925385.000-krakatoa-eruption-cooled-the-world.html

WHEN the Indonesian volcano Krakatoa erupted in 1883, sending 25 cubic kilometres of rock and ash into the air, it did more than generate the loudest sound ever recorded. It also cooled the world's oceans and suppressed rises in sea level for decades afterwards.

Peter Gleckler of the Lawrence Livermore National Laboratory in California and colleagues compared climate models that included volcanoes with those that did not. To their surprise they found that volcanoes seem to have a cooling effect on the oceans that lasts for up to a century after an eruption. The cooling effect of Krakatoa lasted well into the 20th century, says Gleckler.

Big volcanoes inject ash high into the atmosphere and block out sunlight for months or even years, which cools ocean surface waters (Nature, vol 439, p 675).

However, global warming appears to have diminished this long-term effect, he says

.

So obviously the planet is getting warmer as time goes on. And global warming is just another version of this. So if it is, then it's what's going to happen.

Seems to me that we can take all the precautions in the world as far as people go, but you get two or three volcanoes going off and it won't matter anyway.