Climate Research

[TimberWolf]

Thanks for that considered input.
PS I have read Feynman - his is one point of view.

I've yet to hear any scientific luminary say anything complimentary about the philosophers of science. They are overwhelmingly dismissive of it.

[sasguru]

You think no statistically significant warming for 15 years, whilst CO2 levels are the highest for 600,000 years constitute evidence?

Now you're getting somewhere.
Before we carry on, please define what you mean by statistically signficant? What p-value are you starting with?

[sasguru]

I've yet to hear any scientific luminary say anything complimentary about the philosophers of science. They are overwhelmingly dismissive of it.

Maybe so. But philosophy is asking deeper questions than science does. Although the west has been hugely successful at science and technology, real intelligence is still missing.

[DimPrawn]

Now you're getting somewhere.
Before we carry on, please define what you mean by statistically signficant? What p-value are you starting with?

0.03

[kandr]

Now you're getting somewhere.
Before we carry on, please define what you mean by statistically signficant? What p-value are you starting with?

So who thinks DimPrawn is on Wikipedia right now looking up p-value.

[doodab]

No need to be insulting doodab.

All this shows is that the models are tuned in to the 1970's and on.

try hindcasting them back a bit, you will see what I mean

I never said the models were perfectly accurate, I simply asserted that they had been backtested, and provided evidence to back it up, thus contradicting your assertion that they hadn't. Given that your assertion was a) wrong and b) made without evidence I think that describing it (please note I was not referring to your good self) as "utter cock" was quite a reasonable assessment.

One thing about these sorts of models, and weather forecasting is similar, is that small errors such as the measurement errors in the input data magnify quite quickly (nonlinear dynamics, chaos theory, butterfly effect and all that) so one cannot expect them to give particularly accurate predictions a long way into the future.

So, lets say you make a forecast using data up to 1,000 years ago, I would expect the forecast to be reasonably accurate for a while after that point and then begin to diverge but by the time it gets to the present day it will be all over the shop. This will happen whatever the merits of the model, it's in the nature of the problem itself. There are ways to deal with the uncertainty e.g. using monte carlo methods to run multiple simulations with different initial values covering the full space of possible inputs and averaging them or producing a probabilistic forecast, or using pertubations to determine the forecasts sensitivity to initial conditions and taking that into account, and these will give some improvement, but the fact remains that the further into the future you try and predict the less accurate you will be.

If however you backtest 50 - 100 year forecasts at various times in the past and you get a) reasonable accuracy and b) similar errors for different periods (bias) then you get a good idea of how accurate or not your model is at predicting 100 or 200 years into the future and whether it has any obvious bias that can be corrected for. Whether it's accurate 50,000 years into the future is irrelevant, because that's not what it's being used for and it's not expected to be.

Unfortunately I lack the time and computing power to properly test any of these models myself, although if I won the lottery I probably would knock up a little supercomputing cluster and have a go because that's the kind of sad ****er I am

[sasguru]

That is perhaps the stupidest post I have ever seen on these forums, or maybe anywhere.

He has a long track record with that. Can't work out if it's a wind up or he's really that limited. Although i've learned not to over-estimate people - most really are as stupid as they sound.

[sasguru]

0.03

Thatis quite a high barrier and not usual practice.
Why did you pick that?

[DimPrawn]

So who thinks DimPrawn is on Wikipedia right now looking up p-value.

Thatis quite a high barrier and not usual practice.
Why did you pick that?

[TimberWolf]

Maybe so. But philosophy is asking deeper questions than science does. Although the west has been hugely successful at science and technology, real intelligence is still missing.

Science made progress when it shook off its philosophical roots and went empirical.