I’ve been reading through the AGW debates here and I really like that there are well reasoned folks on both sides of the issue. I haven’t seen that anywhere else yet and I feel it’s important.
This caught my eye in that if this is true, then what basis is there for anyone making policy based on those models? Also, how much time would have to go by to get out of the “statistical noise” portion?
Thanks,
Eben
p.s. to those interested in my bias, I’m a philosopher with nothing published (ie: only a BA) and lots of interest to learn more. On the subject of AGW I’m generaly of the idea that resources spent fighting GW or Global Climate Change would be better spent in other persuits. So now you know my main positions, but I’m willing to listen and consider other evidence, since I didn’t always hold this view.
Actually, I would disagree with aptonym’s statement, at least without the qualifier about latest and best models. Jim Hansen issued a projection in the late 1980s that so far looks to be pretty accurate, although the noise is admittedly still large enough that one can’t test it to great precision…On the other hand, noone has ever claimed that we know the climate sensitivity to great precision. (At any rate, the model that he used at the time had a climate sensitivity that is on the high side of what it is believed to be today…so one might expect that today’s best estimate would fall somewhat below the estimates from then.)
Also, there are lots of different ways for scientists to test the AGW hypothesis without us having to run the full experiment for many, many years here on earth (which is obviously undesirable…although there are some people who are doing their best to try to prolong that experiment as much as possible). For example, one can infer equilibrium climate sensitivity from the past climate record (usually with the help of climate models when the record is over timescales that are short enough that the system doesn’t get close to equilibrium…but sometimes, e.g., for the glacial - interglacial transition, without the use of climate models).
Another thing that one can do is vary the parameters in climate models over ranges that one believes to be plausible and see how robust the predictions are. (The answer from the climateprediction.net experiment was that the climate sensitivity could vary quite a bit in the models in the high direction but that it didn’t get below about 2 C per CO2 doubling on the low end. It is worth noting that they did not attempt to constrain their models by looking at features of the resulting current or historical climates that the models produced and comparing them to reality…and some have argued that doing so would likely have allowed them to do a better job of constraining the high end.)
Still another is that one can look at the historical record and pattern of the warming that has occurred and compare it to what is expected from the greenhouse gas mechanism and what is expected from other mechanisms.
And, one can check that various important features of the models that help determine the climate sensitivity are being captured correctly. For example, is water vapor in the atmosphere increasing with the warming in the way that the climate models predict (which is important because this predicted increase in water vapor provides an important positive feedback).
And, there are probably many other ways that I haven’t thought of. But, to summarize, while being able to perform the full experiment is always the best check on a theory, it is not always possible (and, obviously, AGW isn’t the only example of this conundrum…just look what the people challenging evolution say). Nonetheless, there are still ways to test the theory…It is just that they take more work and ingenuity.
As I noted, the climate sensitivity in that model is at the high end of the likely range…so we might hope it will start to diverge a little. However, it is questionable that it is doing so (i.e. diverging from Scenario B, which is the closest scenario to the actual forcings that played out) in any statistically-significant sense. I’d also caution against using the climateaudit version, which made the post-hoc decision to align the measured temperatures and the simulations relative to each other in a different way than was originally done by Hansen, with the effect that the measured temperatures are shifted down a bit (and I think intention can comment more on this as it is my understanding that he is at least partly responsible for that graph).
Furthermore, I believe that Hansen has noted that the data set that there was no land-ocean data set that he was using back at that time…and the one more in line with what he was using is the actual station data. Finally, I believe the last point in the “RSS” satellite data on that plot may be incorrect because RSS had a data error that was causing it to diverge from all the other data sets…and it has since been corrected.
Here is the better version to compare to, showing both the land-ocean data set and the station data set that Hansen argues is more comparable. Admittedly, it only goes through 2005 although you can visualize where 2006 and 2007 would fall by looking here.
Well, the null hypothesis is that the model is incorrect . . . so the proper question is whether Hansen’s projections are accurate in a statisticially signficiant sense.
In any event, the climateaudit article explains the particulars of how the graph was constructed. I haven’t had time to digest it all, but it seems reasonable on the face of things.
I suppose I will wait to see if intention has any insight to share.
All this “null hypothesis” stuff is just bologna. It is example of applying techniques used in one field of science to another field where they do not apply. This is not a binary either-or situation. It is a question of how accurate the estimate of climate sensitivity is.
You gotta be kidding me. Where do you think the concept of “statistical significance” comes from, anyway? And if statistical hypothesis testing has no application in climate science, then why did you mention statistical significance in the first place?
I am not saying that statistical significance plays no role. I am arguing against simplistic binary (either-or) formulations. I.e., what I am saying is that even if climate deviates in a statistically-significant way from the Hansen’s projection, that doesn’t mean that we must accept the null hypothesis that there is no AGW…or that AGW is not a problem…or whatever. What we would have to reject the hypothesis that Hansen’s 1988 model is precisely correct…and given that it is 20 years old and known to have a climate sensitivity that, while possible, is toward the high side of current best estimates…I am pretty willing to go on record as saying that I am already almost certain that this hypothesis is incorrect!
I’m still left with a basic question. Is there any difference between these two positions:
1)The tuned models are working so far reasonably well. We should (as a nation/world) create policy based upon these models.
2)The tuned models don’t prove anything yes. We should (as a nation/world) wait until we have more evidence before making policy.
The question being which of these two options has more scientific support. To give a metaphore to the subtext here, I feel a little like we’re in a position where we’ve come up with a wonder drug to cure a disease, but we haven’t done very thourough trials on the drug and we’re not sure what causes the disease but we’re being pressured to take the drug.
Eban: Well, I don’t really like your phraseology too much. First of all, the phrase “tuned models” is somewhat loaded. The models are based in large part on basic physical equations. Yes, some things are parametrized and there are thus some parameters that could conceivably be tuned but there is a limited degree of “tuning” one can actually do with these parameters in order to fit to the historical temperature record, even if that is what you were determined to do, and there are also many more pieces of data from current climatology…and the basic physics of the process that the parameter represents…that you also want to get right.
Second of all, as I pointed out in the previous post, the argument for AGW involves a lot more than the predictions of climate models and, furthermore, there are also lots of ways to check that various aspects of the models are getting the physics correct (and, also, as I pointed out with the climateprediction.net study, to vary various parameters in the model and see what range of climate sensitivities one can get through such variations of parameters).
I also think that your analogy might not be the best one (although it is a little ambiguous given that it is not clear how serious the disease is and what the alternatives are). Let’s look at another analogy: If you were buying fire insurance for your house, would you first want to ascertain that your house is definitely going to catch fire before you decide to buy it. Or, perhaps you should look at it this way, which I think is most realistic: There is a possible problem in our future but we can’t pin it down enough to be sure which of the range of scenarios is going to come to pass. By the time we do know, it may be too late to prevent considerable damage…or at least it likely will be much more costly to do so than if we were to start taking preventative action now. What is the right balance between the extremes of doing nothing to prevent this problem for occurring and doing what we would do if we were quite sure that the worst cases would come to pass?
Along these lines, here is a recent piece in Science Magazine. I am not claiming that this paper is perfect but at least it takes a crack at this sort of question. It is also worth noting that the U.S. National Academy of Sciences and the analogous academies in many other major countries have been quite clear that they think it is time to take action:
I think that pretty clearly answers your question regarding which position enjoys more scientific support (see also this Wikipedia article summarizing the scientific views on climate change).
By the way, it is worth remembering that, in addition to climate change, there is another concern we have about CO2…ocean acidification…that thus far has received much less study.
You’re right that my phraseology was loaded. I’m not sure what to use instead, since it’s really not possible to have an untuned model of any usefullness. The word “tuned” doesn’t really add anything, so my appologies. Added to that, the negative connotations of the tuning are likely to be directly proportional to your pre-review opinions on the subject rather than something gauged after reading whatever reports and methods are involved.
The abiguity was intentional in my analogy. The cost/benefit ratio seems to be such a guessing game when it comes to AGW. Every time I read something about how inexpensive it would be to reduce our environmental effect, I read something about how that same thing will keep third world countries from developing. I’m sure that both sides use hyperbole to demonstrate their case, and that doesn’t help much either.
I can see here my Philosophy training has done me a disservice, not for the first time. (Would you like fries with that? -joking, I work in an office) I habitualy distrust appeal to authority, yet in science there’s no other real option. The best founded science is the largest number of authorities telling me that math I don’t understand works out to some conclusion. I guess I need to start trusting, but it sooooo goes against my training/instinct/brainwashing.
Oh, on the subject of insurance. I don’t like it. Mostly because of the profit angle of it. The generic idea of being ready for something “just in case” is a fine idea. The idea of someone making money so that I’ll be prepared (for fire, accident, death, whatever) rubs me the wrong way. So if AGW Insurance doesn’t lead to profit, I’m potentialy in. But if someone (Looking hard at Gore) is making money from fearmongering regarding this insurance, I’m skeptical in the sense that it gives me a bad feeling and I’m going to look for holes to poke in it.
I’m aware of that. What you are saying is that the concept of a “null hypothesis” has no application in climate science. While simultaneously talking about “statistical signifcance”
Actually, I will say that in the physical sciences, I don’t see simple binary choices having much applicability, period (except, for example, perhaps in particle physics where they are looking for a particle with some specific energy). In fact, in over 20 years working in the physical sciences, I don’t think I have read one paper where they talked about the “null hypothesis”.
Please link to a peer reviewed paper that talks about the “statistical significance” of a result but makes no mention, explicitly or implicitly, of a null hypothesis.
An asteroid has been identified as heading towards Earth’s direction. Using the best modeling available scientists are able to state that there is a 95% probability that it will hit the Earth directly causing massive global destruction, a 3% probability that it will scrape across the atmosphere causing moderate destruction, and a 2% probability that it will miss the Earth entirely. They do not really know, they can just make probability estimates. As it gets closer they can fine tune those predictions but the only actual “test” of them will be to see if it hits the planet or not. Building a device (whatever that device might be) that could alter the course must start now if it is to be at all effective. It would be a sizable but not economy hobbling investment and would require international cooperation. Some might make profit off of that investment. Alternatively we could invest in how to deal with the destruction once it hits. Some would profit from that. Or we could do nothing.
Would it make sense to make policy based on those (unproven) models? Would any decision be able to be made without consideration of those models?
This point becomes salient only once you accept that using models that can “only” give probabilities is indeed a reasoned (if not the only) basis for making policy decisions. The issue here changes from your op to one of which of those cost-benefit analyses is more likely accurate and how to value different outcomes at different time points. Some of those questions are indeed more based in the science, but some is more right up your alley of philosophy.
Let us just consider the perspective of Third World countries in isolation of any other consideration. In a short term it is indeed possible that international cooperation on mitigating global climate change could increase energy and materials costs and impede short term economic development in some emerging economies. OTOH the Third World in general would be most hardest hit by the effects of global climate change as it ramps up more significantly over the next half century and have the least capacity to deal with the consequences. Philosophically how much more value should be placed on the short term effects over the more severe long term ones that also have a lesser degree of certainty as to magnitude?
Extend this to a worldwide question. In the moderate term certain nation states may benefit from global climate change (Russia will likely have greater access to oil resources as ice melts and opens ship passage to resource rich regions for exploitation and will have longer growing seasons across most of their country). Others may have relatively only moderately severe effects and have the resources to mitigate them albeit a great cost (perhaps the United States) whereas other regions will be more severely affected and without resources to deal with the consequences. Those most contributing to the problem currently and in the near term future would have the immediate costs to dealing with the problem whereas they may have a lesser degree of consequences from the problems resultant longer term. So cost benefit analysis depends on whose setting the values of the different outcomes and what ethical obligations nation states believe they have. That’s not science, that’s values, morality, and ethics. Your training I believe.
Look, the point is that the “null hypothesis” that would be implicit in seeing if there was a deviation from the Hansen prediction would simply be that there is a statistically-significant discrepancy between Hansen’s prediction and the actual global temperature. However, I can already predict for you that, unless Hansen is the luckiest guy on the planet, eventually that null hypothesis will be correct because it is extremely unlikely that Hansen nailed the climate sensitivity exactly correctly with that model (and, also, although over short enough intervals we can probably decide that one of the forcing scenarios, such as Scenario B, was actually what was borne out, eventually we will get the point where the forcings don’t necessarily follow closely enough any one scenario).
So, whether there is any statistically-significant deviation between Hansen’s prediction and reality alone is not a very useful test. A more useful discussion would not take a binary yes-no approach but would actually try to study how well Hansen’s prediction is doing compared to reality…e.g., how large are the deviations and what can they tell us about the likely climate sensitivity compared to that of Hansen’s model.
This is a good analogy to choose. I can see that the analogy choice itself is largely dependant upon preconceived notions. In the case of the disease there’s not enough data to make a good decision, so it’s weighted towards a wait and see/further testing conclusion.
The asteroid theory has enough previously shown to be predictively accurate (we can track space-borne objects fairly well) that there is not much doubt about this new object’s status. So in that case the analogy itself weights towards a preventative choice.
This is very indicative of the status of argumentation I generaly see on the subject of AGW. Not to hijack my own thread to badly, but a good portion of my not yet accepting AGW is based on my own look at the science of astrophysics. The predictive usefulness of the Electric Universe Theory seems to be completely overlooked by mainstream science, and to me it seems that that theory does a lot more predictively regarding the Sun’s output and effect on Earth than climate models account for. So if in my view the climate models are missing a big chunk of data (say for example they don’t know if the asteroid in the analogy is rock or just gasses) I’m not sure what level of prevention is appropriate. At this point it does come as more of a philisophical problem, which is where I entered into the whole ball of yarn actually. I’d be glad to discuss the philisophical ramifications of that in another thread if anyone’s interested.
-Eben
p.s. heck, I’ll discuss ramifications even if no one’s interested, I’m a philosopher after all!
Oh no, not the Electric Universe Theory!!! I think we had a proponent of that start a thread a while back. I think the (pretty overwhelming) consensus at the time on the board was that it looked like a pretty crackpot idea. It’s another one of those “theories” dreamed up by someone who has just enough knowledge and intelligence to actually believe he has come up with something profound but not enough to realize just how large the body of work is that supports the conventional view and doesn’t support his view. Unfortunately, for every Albert Einstein who is really capable of creating a paradigm shift in a field, there are at least 1000 Albert Shmobergs who have views of similar (or even greater) eccentricity without the genius to actually come up with something that is even plausibly correct.
