Discussion > Let's get real about climate models
Ahh attp, you're here. Why don't you weigh in on the predictive powers of GCMs? Love to hear.
Simon Hopkinson
Simon,
Because I really do have better things to do, at the moment, than metaphorically beat my head against a brick wall. Maybe another time.
...and Then There's Physics
Raff, there's an important thing to note here:
I don't know what exact claims are made for models by the climate modeling community
Although it's recognised that some modellers do become wedded to their models, lose objectivity and become blind to their models' inherent limitations, broadly speaking I think most do recognise that there is a distinction between modelling and experimentation. Unfortunately human nature somewhat prevents them from speaking out generally against the confusion because a job's a job. Nobody in their right mind would shoot the golden goose, right? And it's not like they're not doing good work, after all. It's not THEIR fault that their work is being misrepresented and abused.
Simon Hopkinson
I think you’ve hit on a good analogy, there, Simon, with the lottery. Instead of a limited number of elements (49, with the lottery) giving a result of a limited number (6) of them, with an irrelevant sequence, the atmosphere embroils us with an unknown number of elements giving an unknown size of result, some (though not necessarily all) of which have to sequential, thereby allowing some limited projection on short-term probabilities, as do the results, themselves, potentially (though not necessarily) affecting future results. Instead of once a week, though, this is an ongoing lottery, which complicates it yet further, as does the number of elements affecting the results, and the actual size of the results also being variable.
Raff, if you can (or can find someone who can) produce a “model” that can give a reasonable prediction of the lottery result at least one day before the draw, then I think you would have a more accepting audience. While a model might be produced that shows patterns within all the draws since the inception of the lottery, you will find that any pattern that is found will not be of any help in predicting the results of the next draw, other than it will be a combination of six different numbers between, but not including, 0 and 50 – which is, in effect, all a weather forecast is doing: they have an approximation of the number of balls, and the numbers they represent, and roughly how many need to be drawn.
However, as I doubt you will ever be able to accept the impossibility of what is being claimed for climate models, I think I shall follow Mike Jackson’s lead and endeavour to utterly ignore you (as I already do with many of your ilk), from now on, and would recommend others do the same.
Radical Rodent
At least with the Lottery, people who choose to pay for it have a chance of gaining something in return
golf charlie
...broadly speaking I think most do recognise that there is a distinction between modelling and experimentation. (...)
Jan 15, 2016 at 11:07 AM Simon Hopkinson
Er, not in climate science (so-called).
One of the things Professor Michael Kelly took exception to was a witness talking about experiments performed using climate models. (Climatic reseach unit enquiry). Confusing what comes out of unvalidated models with physical reality is one of the things that marks climate science as not being science.
Martin A
Martin, you're right of course, but those are two distinct things. Knowing that there is a difference between a model and an experiment, which is what I pointed out, is one thing. The abuse of science that ignoring that difference amounts to, as Prof. Kelly pointed out, is fractionally different. I genuinely believe that discipline in climate science is so lacking in rigour that there really exist people within the subject who *don't know* that the distinction is important.
I *choose* to believe that the majority of climate modellers are not actively trying to confuse their work with evidence gathering but are instead merely poorly disciplined in the rigour of the scientific method - an issue which I think amounts to a systemic problem in climate science. I *choose* this because I think it's important not to attribute motive unless there is direct, prima facie evidence of wilful, explicit deception. It's a personal choice only. :)
Simon Hopkinson
Has a Peer Reviewed paper ever subsequently floundered or foundered, on account of reliance on unreliable computer models? Is it Unprecedented?
golf charlie
Thin skinned Mike, sometimes being hit by reality hurts. A little advice: next time you feel like recommending an article, get someone in the real world to check for you that it is not crazy - you don't seem to have that filter built in.
Martin, so now I've explained the updated forcings to you, what do you make of that figure 3?
Simon, the thing about that argument from authority fallacy is that it is not reversible. Just as "I'm an engineer, so you should believe me" is fallacious, so "I have no relevant experience" is an indication that whatever opinion subsequently expressed is probably fairly worthless.
Raff
Simon, Radical Rodent
A lottery is NOT a good analogy for climate. The two are opposites.
The lottery draw process is designed to be random, with minimum determinism.
The climate is deterministic, controlled by physical laws, with a smaller random element.
Entropic man
Raff, you've run out of grace I'm afraid.
Simon Hopkinson
EM, straw man. The analogy was not for the purpose of providing an equivalence on the role of determinism. I was quite specific that the analogy was for the purpose of illustrating the inordinately huge benefit of hindsight. I had thought that was obvious.
Simon Hopkinson
EM:
While a model might be produced that shows patterns within all the draws since the inception of the lottery, you will find that any pattern that is found will not be of any help in predicting the results of the next draw…… other than that there will be 6 numbers drawn from a total of 49. However, I can predict that the sum of the numbers will never be more than 279, nor will they ever be less than 21. How is that any different from making predictions about anything else? While the random element of the atmosphere might be proportionally smaller than in the lottery, it is still massive – the “butterfly effect” is as real today as it ever has been. Also, as Simon says, the benefit of hindsight does give 20/20 vision on the past, but remains pretty useless for the future.
Radical Rodent
there really exist people within the subject who *don't know* that the distinction is important.
Simon, so far as I can see, there are people within the subject who *don't know* that there is a distinction.
Martin A
Martin:
Simon, so far as I can see, there are people within the subject who *don't know* that there is a distinction.
Okay, I give! ;)
Simon Hopkinson
The climate is deterministic, controlled by physical laws, with a smaller random element.
Jan 15, 2016 at 3:17 PM | Unregistered CommenterEntropic man
EM - I feel sure if you stop to think for a moment you might want to reformulate what you said there.
"controlled by physical laws": I doubt that anyone (except the Pope, perhaps) thinks that the climate is governed by anything other than the laws of physics exclusively.
However, I imagine that you are well aware that if something is "controlled by physical laws", that does not necessarily mean it is deterministic. And that being deterministic does not necessarily mean that its future behaviour can be predicted
Where does the idea that climate is deterministic (apart from a small random element) come from?
It's something I imagined would be considered false unless there were very good evidence that it were true. And it's very hard to imagine what such evidence would look like.
Is this, perhaps, another instance of the "EM effect", where EM imagines something and, toot sweet, it becomes, for him, reality?
Martin A
Simon Hopkinson
My comment stands
For a purely random process predictability is zero, in foresight or in hindsight. You cannot project last week's lottery numbers with any more success than you can project next week's numbers.What you can do in hindsight is compare your projection with the actual outcome. You will amost certainly find that your projection was wrong.
For a completely deterministic process predictability is 100%. You can project past and future values exactly. The only difference between foresight and hindsight is that you can confirm your projections of past values by comparing them with measurements. You will almost certainly find that your projection was correct.
Climate is intermediate. In the long term states and trends are deterministic and can be projected in foresight or in hindsight using physical laws. Short term variation is not predictable, so any projection will have a mean plus error bars. Comparison between projections of past temperatures and measurement agree within their respective error bars. This gives confidence that future projections will also reflect real outcomes within their error bars.
Entropic man
EM, as much as you think your comment stands, it's not really relevant in the scope of this discussion. Riff's argument apparently was that you can discover previously unpredicted non-deterministic forces learned in hindsight and apply them forward, thereby improving a model's predictive capabilities and presumably significantly reducing error bars. I disagree.
Simon Hopkinson
For a completely deterministic process predictability is 100%. You can project (...) future values exactly.
EM, as I said before (I think I said it before) "deterministic" is not the same as "can be predicted".
For many conceivable deterministic processes, you can only "project" their future values exactly if you have *exact* knowledge of the process's present state, *exact* knowledge of the process itself, and you can do your computations *exactly*. Anything less than complete precision in any one of those can result in complete rubbish "projections" over even a short time interval. And none of them apply for any real physical system represented by a deterministic model.
. In the long term states and trends are deterministic
EM - I know that there are people who believe that but, other than "everybody knows it", there is no evidence it is true. Is there even a convincing plausibility argument for it? To me, it is one of the myths taken as gospel in the Church of Climate Science.
Martin A
Simon Hopkinson
"Riff's argument apparently was that you can discover previously unpredicted non-deterministic forces learned in hindsight and apply them forward, thereby improving a model's predictive capabilities and presumably significantly reducing error bars."
Your statement is a bit ambiguous, but here goes. You seem to have misunderstood what Raff is saying.. He is not referring to new, unknown, forcings. He is referring to unknown post-2005 values for known forcings.
The initial CMIP5 models use known values for forcings such as CO2, solar insolation, albedo, aerosols, vulcanism, ENSO, etc up to 2005. Post 2005 different model runs use different estimated values. A run with high positive net forcing will project higher temperatures. Runs with lower net forcing give lower temperatures. Together the ensemble gives policy makers an idea of the temperature outcome of different policy options.
The forcing-adjusted model runs shown in Fig.3 of Raff's link used measured forcings up to the end of 2013, and estimates thereafter. The extra eight years of data gives two advantages.
Firstly it gives an extra eight years of comparison between projections and reality. Secondly, it reduces the error bars for given dates in the future.
A projection to 2035 uses 30 years of estimated data in the 2005 based ensemble, which becomes 22 years using the forcing adjusted models.
Entropic man
Martin A
I am glad you qualified your initial remarks.
For one entertaining moment I thought you had stopped believing in cause and effect. ☺
Entropic man
Martin A
Entropic man
Martin A
Entropic man
EM - I'm not sure what your "cause and effect" refers to. The application of the Holy Climate Science Formula to compute "radiative forcing" from atmospheric CO2? A formula hypothesised to calculate a non-physically measurable quantity is not exactly a demonstration of "cause and effect" but perhaps you meant something elst.
Please answer my question [if you are willing and if you can] about why you believe "in the long term states and trends are deterministic". It might also be helpful to understanding what you are on about if you'd clarify what *you* mean by "deterministic". (And please don't just say "non-random".)
Martin A
Raff, I'm going to address some more of your post:
I have said already in this thread (so please pay attention this time), the models are explorations/descriptions/representations of hypotheses and are useful tools in that they give an indication of knowledge boundaries. Put another way, they're useful at telling us how LITTLE we know about the climate. Is that useful to know? Absolutely, it's essential. Is it worth writing and running them? I think so, yes.
See, fundamentally I object to this lining up for an argument from authority/appeal to authority logical fallacy. This is why I don't talk about my expertise or lack thereof, and would rather the veracity of what I say be assessed rather than the fallacy of experts or ignorants talking. I'm a computer programmer and modeller, in industry not in academia, and it's completely irrelevant.