Saturday, October 23, 2010

Climate Models

Here is a fairly boring academic talk by Mike Hulme on climate models. It gives 4 measures of reliability and discusses the current climate models. From my perspective, it should give people some caution about swallowing "modeling results" holus bolus and without some skepticism.

Watch the video here.

Skip the first 18:00 minutes which include an introduction and some introductory remarks.

Having built computer models in my previous life, I watch this and chuckle at the careful terminology and tepid characterization of the climate modeling community's efforts. It is obvious that Hulme is highly constrained by a desire to keep his "academic standing" so his remarks end up being carefully hedged and qualified to ensure no academic reputations are threatened.

One oddity of this talk: why spend an hour talking about the need for reliability (on four different fronts) if there isn't an issue of reliability. You will notice that he never admits to any problems with model reliability. But he does give broad hints that this is badly needed. He sounds like a politician and not a scientist.

His only clear words are in the introduction where he talks about the origins of science and the need for experimental verification. This of course is completely impossible for "science" based on model predictions. Reading the tea leaves, it sounds like he is most keen on moving the modeling from the current "private" models jiggered in the back room to open platforms using open software standards where the modeling is released along with the model results to ensure that reliability can be evaluated by a wider public. This makes sense only if the current secretive modeling is in fact questionable (which it is). But he never clearly says "there is a problem which can be seen from this, and this, and this". Instead it is all airy fairy "we need reliable models and here are four directions in which you make models more reliable".

He mentions the UEA CRU e-mails but gives no indication of their relevance to a need to improve model reliability or indicate why the stench of scandal hangs over them. He admits that over the last "few years" there has been a decline in public acceptance of climate modeling "results". But he doesn't connect the dots.

No comments: