Global Warming Predictions, Computer Modeling, and Error Reporting 4

[TGS4]

It's been a while since we had a good AGW discussion here. I read an article yesterday that discussed the climate models and their error (in other words ±1%, 5%, 10%, etc).

http://www.skeptic.com/the_magazine/featured_articles/v14n01_climate_of_belief.html

I work with computer models (FEA) for a living, so I can appreciate the issue of error reporting. My models are approximations, so they have error such as discretization error, boundary condition error, material property error, etc. I always report on those sources of error in my reports. Why do none of these discussion about future climate modeling ever discuss the error in the models?

So, forget whether or not the planet is currently warming. Maybe it is, maybe it isn't. I personally don't have enough knowledge to know for sure. However, if the climate models that we are basing predictions on are actually on the order or +4°C ± 100°C in 100 years, then I call BS. Who cares whether or not the different models agree with each other, what about the agreement with observation?

 

[cranky108]

I don't know about your weatherman, but ours is wrong quite often.

So if he can't predict the weather for tomorrow, how can we trust him to be correct in 100 years.

 

[rb1957]

yeah, the only way to make accurate weather forecasts is to forecast yesterday's weather !

maybe i'm in denial but i don't believe that changes in a trace gas (0.03% to 0.06%) have a significant effect on the global climate; and to compound this by saying that cow farts are responsible ... good grief.

we can certainly be more efficient in our consumption of carbon fuels. we should consider encouraging the developing nations to leap-frog the carbon fuel economy and go directly to an electricity based economy, and not just so we can have all the carbon fuels.

 

[KENAT]

Have we ever actually had a good AGW thread?

Posting guidelines faq731-376

http://eng-tips.com/market.cfm?

(probably not aimed specifically at you)
What is Engineering anyway: faq1088-1484​

 

[TGS4]

KENAT - good in the context of a lively discussion/debate. Those threads have been cooling for several months now (all double entendres intended).

 

[zdas04]

We can absolutely do a better job on fuel efficiency. No doubt about it. Eventually fuel will get expensive enough that people will be unwilling to drive their 18 wheeler to the Quick Stop for a pack of smokes. It is not near that point yet.

There are economic reasons to conserve and as engineers, a big part of what we do is trying to find an optimum mix of efficiency and price. It has always been that way.

Back to the OP, I recently put my opinions on modeling in a blog called Computer Models Never Prove Anything. Like TGS4, I use computer models a lot and find them to be a very valuable engineering tool, but also like TGS4 I'm aware of their limitations. I like the example (I think I learned this on eng-tips.com, but I can't remember who posted the first version I saw) of opening an Excel sheet and in cell A1 put 0.99999999999999999. In Cell B1 put 1.000000001. Then in the next row square the previous line. Continue that for a few dozen iterations. Two numbers that start with approximately zero variance in a hundred operations column "A" is zero and column "B" is as close to infinite as Microsoft will allow. A climate model operates on approximate data thousands of times per step. Infinitesimal inaccuracies swell to dominate the model.

I wish this discussion would go away, but there is so much money to be made that the self interest of the Politicians keeps getting in the way. Already there are several new billionaires in Europe from their Cap & Trade scheme. I expect that within a year of the U.S. implementing this nonsense, Chicago will have 10-30 of the top 100 richest people in the world. Just ask Kofi Annan how much your conscience hurts after the first billion (and "Oil for Food" was tiny compared to Cap & Trade). There is enough money here to purchase any number of scientists, politicians, and media whores.

David Simpson, PE
MuleShoe Engineering

http://www.muleshoe-eng.com

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

"It is always a poor idea to ask your Bridge Club for medical advice or a collection of geek engineers for legal advice"

 

[TGS4]

David - fantastic blog. I could write the exact same thing about FEA rather than fluid flow models, but the concepts are identical.

What do think the discretization error is for a 173 mi X 259 mi (44,712 sq miles) grid? Forget about the assumptions around the equations INSIDE the grid, just the discretization error?!?

Now, if only there wasn't the obscene amount of money to be made in certain schemes... Say, in that vein, how does a simple engineer get on-board a scheme like that to rake in the Billion$?

 

[zdas04]

Thanks for the kind words.

The discretization error is around +/-100%, but they have tweaked their boundary conditions to the point that the program will stand up and sing Yankee Doodle Dandy in whatever key is requested. Every year they move a few more variables from the "boundary condition" column to the "determinant" column, but the boundary conditions are still mostly first order effects (they don't have either the Gulf Stream or the Japanese Current in the model, to name two significant thermal masses for example).

The really big bucks are going to be "made" by the same sort of bottom feeding slime that made huge money off Oil for Food. I wouldn't climb in that bucket for any amount of money.

A lot of consulting engineers are going to make a ton of money ethically doing things like Technical Review of Western Climate Initiative Proposals to Meter Fuel and Control Gas. I wrote that document to try to head off the WCI from causing many billions of dollars of damage to industry to force them to report worse data. There is also a bunch of money to be made designing CO2 Sequestration projects. I've done the conceptual design on four of them so far and it pays well.

A bunch of people will make their fortunes selling snake oil "solutions", but none of them will be ethical engineers (most will be engineers, but I have to question their ethics).

David

 

[corus]

Here's a good response to the inaccurate maths used throughout that article

http://skepticsplay.blogspot.com/2008/05/innumeracy-in-global-warming-skepticism.html

Tata

 

[jmw]

Tracking back the bruin skeptics, the criticism is authored by a physics student at UCLA.

He assumes random errors which tend to cancel out one year to another while the author of the article responds that he was looking at systematic errors which do not cancel out.

http://bruinskeptics.org/2008/05/27/innumeracy-in-global-warming-skepticism/

The comments are an interesting read and Miller wriggles politely.


JMW

http://www.ViscoAnalyser.com

 

[owg]

TGSD4 - I like one of your closing sentences.
"Aren't we better off accumulating resources to meet urgent needs than expending resources to service ignorant fears."
That summarizes my position quite well.
Do we really spend a lot of time in the UK damning the folks who cut all the trees down to make masts (power) for their ships? We should deal with the present and accurately predictable future.

HAZOP at

http://www.curryhydrocarbons.ca

 

[zdas04]

That article makes many bold statements that are collectively and individually nonsense. As best I can tell, he is saying that weather is random, but climate is set on an immutable, predictable path. That is like saying that each roll of a pair of dice is random, but in the long run the patterns will be obvious, predictable, and can be modeled statistically. Many people have lost their life savings at the crap table with "systems" like this.

The discussion of errors being cumulative is just dreck. He starts with the erroneous statement that "even deniers will agree that the models can predict +/- 10C". Well this denier will not accept that. I don't accept that the models can actually "predict" anything and my analysis (as a modeler myself) is that there are so many things that they have as "boundary conditions" that are actually variable with time and space, and so many interactions between variables that are ignored that the model's only function is to produce pretty graphs to feed the fear monster. Then the author goes on to say that statistically the error does not vary with time, but with the square root of time. Either statement is nonsense. If I roll a pair of dice, I have exactly the same probability of the result being a specific number as I have every other roll of the dice. If I roll them a million times, the next roll has the same probability of a double 6 as every other roll--even if double 6 has shown up 500,000 times in the million rolls. Random means random.

Fundamentally, he is looking at a mite on a flea on a squirrel in the forest and saying he can predict forest fire activity in 30 years. He can't. When climate models were science instead of politics, they were moving towards being interesting (I don't know that they were actually moving towards useful, but definitely towards interesting academic exercises). Now that they are pure politics they are moving towards theatrics.

The discussion of ignored data, modified data, and manipulated results are all tempests in a very fragile teapot--the real issue is that MODELS CANNOT PROVE ANYTHING and it is a good thing for the climate scientists that our leaders wouldn't know proof if it bit them on the nose.

David

 

[rb1957]

is this the 5 minute argument, or the full 1/2 hour ?

 

[satchmo]

Thanks rb. I needed that laugh at the end of the week.

 

[jmw]

I'm staying out of any coin tossing arguments. There are too many pitfalls and counter intuitive scenarios and before you know it I'd be playing crown and anchor and losing my pocket money.

Suffice it to say that I am not convinced by the computer models and especially having seen Harry's notes which suggest the computer models may have been written by high school dropouts.


JMW

http://www.ViscoAnalyser.com

 

[321GO]

The "climate" is highly complex and probably poorly understood at best. Number of factors involved enormously.

Now, how can we estimate something we don't even understand to a satisfactory level i ask?

 

[Qshake]

I've satisfactorly stayed out of the earlier discussions on this but since some brought up the analogy regarding FEA and other computer modeling, I'm going to have to intervene and call a spade a spade.

You're engineers, for goodness sake. You care more about the interpretation of the results and in the necessary context. In some cases, many people's lives depend on it.

The reason you never see any error analysis in the articles you've read on this is that these are likely mainstream articles wherein the intent is to produce panic among the common folk to deter their evil ways. It is not to produce a scientific discussion or dialogue.

In the context of geological time and more importantly in human time, we've already experienced what is now called the little ice age in Europe. This has been credited with such things as farming wheat since it was hardy and the nice things that fell out of it, one of which was beer/wine. But I digress. The point is that there have been major fluctuations in temperatures already, both hot and cold.




Regards,
Qshake

Eng-Tips Forums:Real Solutions for Real Problems Really Quick.

 

[zdas04]

Nobody, but nobody is saying that the the climate is not changing. The climate has always changed. It will always change.

The discussion is about two things: (1) have man's activities forced the changes to be in a specific direction; and (2) can we predict the outcome of efforts to change those activities. In my lifetime we've had horrible scares that the next ice age is upon us and the current "global warming" hysteria. I personally don't believe that there is any credible science that ties human activities to climate change, nor do I believe that efforts to predict future changes have the robustness required to develop meaningful projections.

This thread is about modeling and several of us who have participated have familiarity with computer models of complex systems. There are valuable insights that can be gotten from a properly crafted model, but the model doesn't prove the things it illuminates, merely shines some light on them.

In 100 years this discussion will either rank up there with the small portion of society that thought the earth was flat in the 15th century, or it will be seen as the deniers preventing action that could have saved the planet. I won't live to see either side proven right. I just hope that I don't have to watch the world's economy collapse from the weight of the regulations on the horizon.

David

 

[SomptingGuy]

It is simply wrong to compare (steady-state) FEA models with (transient) CFD models. Steady-state models are not sensitive to initial conditions. There may be a unique solution for the stresses and strains inside the domain that are consistent with the material properties, model geometry and boundary conditions. CFD models with cyclic boundary conditions (like engine models) can converge to a cyclically repeating solution. But completely open models (like weather models) will always diverge in a direction dictated by their initial conditions. You can't make a weather model more accurate by mesh or time-step refinement in the same way that you can for an FEA model or a cyclic CFD model.

- Steve

 

[btrueblood]

What Steve said. The most telling point of the article is where it points out the inability of any global circulation model to reproduce its own results; i.e. the results over long time periods become dependent on the growth of numerical instabilities in the codes themselves, and not so much on physical (real world) mechanisms.