A study of atmosphere-ocean predictability on long time scales

Two different methods, one diagnostic and the other prognostic, are used to investigate the predictability of the coupled atmosphere-ocean system on time scales of months to decades. The diagnostic approach analyzes the output of a 200 year coupled model simulation for evidence that the long time scale variability of annual, pentadal, and decadal means of surface air temperature are "potentially predictable". The prognostic "perfect model" approach analyzes predictability of the system by calculating the rate of separation of a small set of coupled model simulations in terms of monthly and annual means of surface air temperature over the globe. Both approaches give reasonably coherent results. At the shorter of the time scales considered, there is little predictability over land but some evidence of predictability over the tropical Pacific, the Southern Ocean and to some extent the extratropical northern Pacific. At longer time scales there is some slight evidence of predictability over certain land areas and more definite evidence of predictability in the tropical Pacific and Southern Ocean. A new area of predictability emerges in the tropical Atlantic. There are plausible physical mechanisms which may explain these long time scale areas of predictability. Both approaches have the potential of shedding light on the long time scale predictability of the coupled system in more extensive predictability studies of this kind.