[Stoch] Stochastic Parametrisation

[James Annan]

Hi Tim and all,

Thanks for the interesting read. I'm surprised no-one has mentioned to 
you before, but I believe that the GISS GCM uses a stochastic 
parameterisation, and in fact has done so since the 1980s.

I still don't understand how you draw a link between stochastic 
parameterisation (which certainly seems a good idea to me) and the need 
for model diversity (which I *also* think is important). Any single 
model, stochastic or otherwise, will surely not be able to represent 
uncertainty in the equilibrium response to any steady external forcing, 
essentially due to the law of large numbers.

I would also hazard a guess that in your Lorenz '96 example, you could 
get better results by taking the average of different stochastic 
approaches (compared to any single scheme) - certainly if you chose a 
weight carefully...

James

> Comments are welcome.
>
> Tim


-- 
James D Annan jdannan at jamstec.go.jp Tel: +81-45-778-5618 (Fax 5707)
Senior Scientist, Research Institute for Global Change, JAMSTEC
Yokohama Institute for Earth Sciences, 3173-25 Showamachi,
Kanazawa-ku, Yokohama City, Kanagawa, 236-0001 Japan
http://www.jamstec.go.jp/frcgc/research/d5/jdannan/