http://onlinelibrary.wiley.com/doi/10.10...
http://gradworks.umi.com/15/48/1548236.h...
http://adsabs.harvard.edu/abs/2012AGUFM....
There needs to be more work done on it and you can't predict exactly what it will do in the future because it has random elements, so if you see the average model output then this hides the PDO because averaging over lots of things with a random PDO tends to 'average it out'.
I don't really know much about climate models, but I would guess that they use average values for long range forecasting, so the answer would be "not at all!". I understand that they are very accurate when reproducing current temperatures since actual data can be used. I would expect that a good model, using average values, would over estimate temps in the PDO cool phase and underestimate them in the warm phase.
As Lin says, it would be nice to know for sure.
Near as I can tell, models ignore the PDO. It is predictable. Do a little research, it switches every 30 years or so. It switched to the cold phase about 2000, coinciding with the halt in warming. (go figure). Go back 30 years and you are at 1970, when we were worried about an ice age. 30 years before was 1940, the end of the "dirty 30's." Here's a good graph of the PDO:
Great question. Scientists couple GCMs with ocean models, but I think PDOs are very hard to reliably predict occurrence and magnitude based on a temporal trend, so probably not.
This just came out in PNAS. It may be more informative.
http://www.pnas.org/content/107/5/1833.f...
I don't think that they do.
I think the problem is that they're not predictable.
I think that they change where the Pacific waters are warmer.
However, I'd certainly like to know, and understand, where I'm wrong.
(Wouldn't it be amazing to hear an AGW denier say something like that. :)
Edit: They're not predictable in that we don't know, a long time ahead, when they'll occur, or how long they'll be. We do know that they'll happen and kind of average out. They do tend to mask some of the results, but there's not a lot that can be done about that.
No you can program the same numbers to predict a outcome
and it could come out no harm or the worst thing possible
or something in between .
It depends on the sophistication of the model.
