Decadal scale predictability is not difficult because of the models, it is difficult because of the character of the turbulence that makes it even impossible for models that are perfect to machine round-off (Lorenz’s study). GCMs have to get the jet stream, equator-pole heat and moisture transport, annual cycle, continental effects, stratosphere, etc., right to even have a chance at being used for climate. Each model may have different biases in regional climate details, but only when they have generally similar responses to forcing changes is it considered significant in the IPCC report. There are things as pointed out by CK that may not be modeled, e.g. collapses of ice shelfs, slowing or stopping of major ocean currents that are more likely under climate forcing changes, or just solar variations and volcanic activity, so no one can say the prediction is completely reliable, and is just a guidance about the least that could happen in that sense.
↧