But all the government stuff is public domain, so you can look it up yourself if you like.

Note that since the American people own the source code, you can freely dl that and really get into it, should you so desire:
http://www.giss.nasa.gov/tools/

And most climate modellers are physicists. That’s what makes them so insufferable.

Hansen, J., G. Russell, D. Rind, P. Stone, A. Lacis, S. Lebedeff, R. Ruedy, and L. Travis, 1983: Efficient three-dimensional global models for climate studies: Models I and II. M. Weather Rev., 111, 609-662

I think if you want to cast doubt on people’s work and demand some kind of high level review, it would help to do some more homework and not rely on anecdotal evidence.

I totally agree. Like I’ve said before I didn’t expect to convince anybody by the tidbits I mentioned on this thread. I only included them to give people a feel for the types of things I was worried about, to see if there was anything parallel in this galaxy formation case. This whole thread is part of doing my homework.

-Sam

This is very optimistic. At some level, all simulation codes are similar: you have a bunch of PDEs and are approximating them with a finite difference scheme. On the other hand, the devil is in the details. Have you ever written a CFD simulation code or tried to use, test, and verify one? I did for my thesis and it took me an entire year to get to the point of trusting it, even though I started by adopting almost wholly a code someone had already written, and even though the code was primitive by modern standards and solved a highly simplified CFD problem. I would not expect to be able to vet someone else’s code in a different field without a huge effort. People in different fields need to make entirely different sets of approximations, so for example the techniques used in CFD in astrophysics can be very different from those used in atmospheric studies or oceanography. Also, particle physicists, bless their hearts, are not all competent to teach fluid dynamics, which is widely regarded as a scary nonlinear subject, and worst of all, applied science.

when a graduate student mentioned to me the sensitivity of that model to a change in the way vegetation is handled

I think if you want to cast doubt on people’s work and demand some kind of high level review, it would help to do some more homework and not rely on anecdotal evidence.

Some useful, non-technical summaries of the history of climate modeling can be found at
http://www.aip.org/history/climate/simple.htm
http://www.aip.org/history/climate/GCM.htm
I strongly recommend the whole site
http://www.aip.org/history/climate/

Makes sense… I suppose I should always be explicit in sensitive matters. Let me mention for the sake of further harmony and agreement that should an engineer (or even a biologist!) also have experience with complicated simulations, I’d want that person’s opinion, too. I’m not sure about a social scientist, however (me physicist so smart definitely no need whatever they do).

BTW Julianne (I think) wrote a really funny post on this a while back… the punch line was how we love to talk about bioligists, enginners, chemists… but we never mention mathematicians because we secretly fear they are smarter than us :).

-Sam

This is a much more nuanced explanation of your concerns than what you wrote above. The comment which pushed my button really reeked of “Me physicist so smart I figure it ALL out myself!” This is an attitude which some physicists certainly give off, and drives my non-physicist science colleagues absolutely crazy.

scott

I don’t think this is correct, but it would certainly make me feel better about the whole field if it were. My understanding is that the temperature increase from man-made c02 would be negligible without the many poorly understood feedbacks, which must be simulated (and some still can’t be, such as cloud cover changes).

However I should say that if your sentence does turn out to be correct, then, as honest scientists, that is what we should tell Congress. Instead, we have Hansen presenting his model predictions with dramatic errorbarless increases out a hundred years, or giant international bodies putting uncertainties on numbers from simulations that, in line with your above sentence, we know just aren’t useful.

-Sam

The expertise of physicists that I refer to is a familiarity with what level of quantitative significance should be ascribed to results from complicated models, given what tests on those models are applied. For example, when a graduate student mentioned to me the sensitivity of that model to a change in the way vegetation is handled, my immediate reaction was “why should I believe anything that has ever come out of these models, or anything that will come out of them in the future, until they demonstrate that their results converge on some value as they add more and more physical processes?” Isn’t that your reaction, too?

All I hear in physics and astronomy colloquia and discussions is how (complicated) simulations are sometimes nice to have around but can’t be trusted quantitatively. I can’t count how many times somebody starts a sentence about their work or idea with “if you believe simulations…”. So, we don’t trust simulations. Then I hear the climate pepole talking about their simulations, which are immeasurably more coupled, non-linear, and tuned/made up, so I suggest that maybe we shouldn’t believe those, either. Maybe we should even be purportionaly more distrustful of those. At this point, somebody starts talking about Kyoto, or the oil companies, or tells me what some other “skeptic” once told them, as in this thread. But I just want to know why we trust simple N-body simulations and the most complicated non-linear systems ever attempted, but nothing in between. It doesn’t make any sense to me. Do you at least see where I’m coming from?

(By the way, I’d love to continue this in private some time, if you’re interested. It can be very frustrating in a group or online, but I was hoping to get Julianne’s opinion, so I posted.)

-Sam

We would only be convinced that there are no errors in the simulation if we could pin down why the simulation deviates so much from the rough estimate. But that is effectively upgrading your back of the envelope model.

Now, we already understand climate at the “back of the envelope level” and computer simulations agree with such rough estimates. The climate skeptics, who mostly lack physics training, are the ones who argue in favor of unphysically low values for climate sensitivity.

On what possible basis can you claim that physicists have the expertise to learn about the (highly coupled and nonlinear) dynamics of climate, yet climate experts don’t have the expertise to learn about physics experiments???

Sheesh; exhibit A for the public perception of physicist arrogance!

The difference is that climate scientists don’t nearly have the expertise required to learn about particle physics experiment and theory, whereas physcists who already do simulations do (easily) have the expertise to learn about, and critically evaluate, climate models. They’re just fancier versions of the same old stuff physcists do all the time (I think including these semi-analytic models; I hope Julianne will respond to clarify that). I agree that it would irritate climate scientists, and convince few of them, but in my mind, as I’ve said, it’s about the integrity of science. The only real “practical” application would be to ensure that scientists maintain the trust of the public. The public (especially the right) is already quite mistrustful. If global warming doesn’t continue as the climate scientists tell congress it is going to, why will they ever believe us again? Why trust us when we say we’re sure of the big bang, or of evolution, if we were so confident, and yet so wrong, about climate change?

-Sam

I’m intrigued. What is the alternative, apart from throwing up your hands and saying we won’t understand this until fresh observational or experimental evidence (driven by looking at other problems) comes along?

Is it the analytic or the semi- part that’s the issue? You’d think the more you can make something analytic the better, if only because it’s easier to explain what’s going on.

David

I’ll simplify the climate change issues under debate into two questions: (1) is anthropogenic global warming happening? (2) for a given rate of CO2 production, how large is the temperature rise going to be N years in the future? Basically, (1) is a question that can now be answered by looking at the record, since the models are actually constrained by the data, and (2) is extrapolation. The uncertainties in complex climate models mostly affect (2) now. However, many climate change skeptics persist in attacking (1), although virtually all serious climatologists now accept it as established. Testing the models, then, only affects the question of the future prospects: whether it looks sort of bad, bad, or really bad.

Sorry for continuing the threadjacking. I could try to make analogies to some issues in semi-analytic modeling, but for safety’s sake I better stop now.

I presented my opinion of (much of) climate science and (many) climate scientists without justification. I qualified it as an opinion of a non-expert in the hopes that people wouldn’t get mad. Maybe I should have also said “Lots of people know more than I do, but I’m still allowed to have an opinion.” It seems obvious to me, but you can never be too careful in this world. In any case please understand that I wasn’t hoping to convince anybody to agree with me based on this one-sentence non-argument. I guess are should have made that clearer. I’m happy to try to convince you (or be convinced myself), but this isn’t the right forum. If anybody wants to discuss it with me, you are welcome to email me at capook@gmail.com, or stop by physics if anybody is in Chicago.

-Sam

Lab Lemming–Thanks for the pointer.

Thanks for the response. Do you think that if the galaxy-formation models became (with money, time, and effort) as sophisticated as the climate ones, then they too would be useful? I think that by “sophisticated” you probably mean either “take in to account more processes” or “do a better job of modelling these processes”. You mention full hydrodynamics, which presumably falls under the category of “do a better job”. But the models also contain–and are very sensitive to–many elements that are just as crudely modelled as what you mention from galaxy formulation (your “crafty juju”). For example, one of the big models changed its long-term prediction by a few degrees when the model for earth’s vegetation was changed from “one-leaf” to a supposedly more accurate regional cover model.

But I don’t think it’s worth focusing on climate in this forum. What would be most helpful to me is if you could point out the similarities and differences between the respective models, using your expertise from astronomy. The name of the game in climate simulations is to parameterize your ignorance of the details of all the physical processes you think should play a role, and then determine the free parameters by fitting to past data (the period from ~1980 you mention). As I understand it there are typically about 100 free parameters that are tuned in this way. Is there an anlogous step in semi-analytic galaxy formation, and how many free parameters are involved?

Also, is falsification of these models ever discussed (or they entirely post-dictive as you suggest)? The model-based IPCC predictions have been falsified at least twice, but only for “secondary” predictions–the artic ice melt rate and the warming of the oceans. Is there anything analogous in the galaxy-formation case?

Your response is eagerly awaited…

-Sam

One of James’ pet professional peeves is the shape of the probability density function of climate sensitivity (the global temperature change caused by a doubling of CO2).

A variety of methods give a climate sensitivity of around 3 degrees. But how fast this tails off as you get substantially higher or lower than this average is one of his areas of research.

A good place to start is his post “Climate sensitivity is 3C”:
http://julesandjames.blogspot.com/2006/03/climate-sensitivity-is-3c.html

In climatology, “Alarmist” can be defined as, “Anyone whose preferred sensitivity value is substantially higher than my own.” Similarly, the epithet “denialist” means “Anyone whose preferred sensitivity value is substantially lower than my own.”

Since we don’t know what your preferred value for climate sensitivity is, we have no way of knowing who you refer to when you use the “A-word”. But the links above (and papers referenced therein) explain why the actual value is unlikely to be less than 2 or greater than 4 degrees per doubling of CO2.

The first speaker would have only a one-word slide, “yes”, and then everyone could go home.

That rather depends on who ends up being selected to be first speaker at our workshop! And which “we” is being referred to in the title. There are certainly those who are deluded into thinking that semi-analytics provides a “real” view of the Universe, but there are also those who use it as an appropriate tool.

At least in my view, semi-analytics is at its most useful when it doesn’t work. When you take a very generic parametrized model of the Universe and try to twiddle all the knobs and dials and switches to reproduce the real thing, it is pretty much uninteresting when you succeed. But when you cannot reproduce the full set of observations then that is telling you that even your generalized fully-tunable model is missing some fundamental ingredient.

Unfortunately, most proponents of semi-analytics are more interested in advertising the successes of their codes than their failures.