Large-scale modes have potentially reproducible patterns, possibly enhancing the predictability of the ocean's response to atmospheric perturbations. On the other hand, prediction of the phaseof a mode excited by atmospheric forcing can be challenging if the low-frequency variability is chaotic—this phase uncertainty degrades the predictability of the forced response. Since chaotic trajectories diverge exponentially, intrinsic variability can be regarded as an instability of the unperturbed flow. A small forcing or initial condition perturbation may excite an unstable mode which grows exponentially until it dominates the forced response.
The role of intrinsic variability in oceanic predictability is explored in a series of simulations using a coarse-resolution global ocean/ice model forced by an annually repeating atmospheric state. Large scale interannual variability spontaneously develops if dissipation is sufficiently small. The most prominent mode of intrinsic intrinsic variability takes the form of large-amplitude long Rossby waves radiating from Western Australia due to a baroclinic instability of the Leeuwin Current. These waves refract around South Africa and contribute to low-frequency variability in the South Atlantic Ocean. The North Atlantic supports a number of additional intrinsic modes which appear unrelated to those originating in the Southern Hemisphere.
The modes of intrinsic variability have repeating spatial and temporal patterns, but the amplitude and phase of the modes varies chaotically in time. Small perturbations cause nearby trajectories to diverge exponentially with an e-folding time of a few years, suggesting that the predictability time for the ocean model is relatively short. However, removing the projection of highest amplitude intrinsic modes from the forced response allows the forced response to be distinguished above the level background variability for more than a decade. Further, temporal and spatial structure of the intrinsic variability imprints itself on the system's predictability, with boreal fall the least predictable season globally and pockets of enhanced variability in the North Atlantic during boreal fall and winter. These results imply that knowledge of the spatial structure and evolution of the intrinsic variability can indeed be used to extend the predictability of ocean forecasts.