[ExI] Climate models

[Anders Sandberg]

John Clark <johnkclark at gmail.com> , 30/3/2014 5:42 PM:
On Sat, Mar 29, 2014 at 3:14 PM, BillK <pharos at gmail.com> wrote:

 > I wouldn't fret too much about climatologists having difficulty putting clouds into their climate models.

You damn well should fret about it if you're using those climate models to plan the future of 7 billion people! 
"Remember that all models are wrong; the practical question is how wrong do they have to be to not be useful." (Box & Draper)
In the case of climate models the useful output is (1) it looks anthropogenic gas emissions are big enough to cause warming in the generic case, (2) the distribution of outcomes (like sensitivity factors) is relatively skew so big deviations are not too surprising, (3) detailed outcomes are really hard to model well (and real world data also give evidence for this - the Arctic sea ice disappearing faster than expected is deeply worrying since this ought to have been modelled well and represents a potentially big feedback factor). 
3 tells us that making ever more elaborate models will not give us much more precision, since there is enough irreducible uncertainty in the system, science and models: hence we should plan for changes in climate, but not bet too much on any particular temperature/precipitation interval. This is reinforced when you look at the mess of the IPCCC working group II (trying to turn climate scenarios into effects on human society): we do not have the science there, and human society is changing and adapting anyway. That is both good and bad news, but it shows that the only smart policies are robust ones that hedge their bets.
I would be watching reinsurance companies for ideas of what to do.

Anders Sandberg, Future of Humanity Institute Philosophy Faculty of Oxford University
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20140401/5065c0bf/attachment.html>