Long term fluctuations in the statistical properties of low-frequency variability: dynamical origin and predictability

The stability and predictability of the statistical properties of low-frequency variability are investigated by studying the dynamical origin of long-term changes arising, respectively, from the nonlinear response to variations in atmospheric forcing and from chaotic fluctuations in the energy partition between different variability patterns. Results from long integrations of a realistic quasi-geostrophic (QG) model are analysed, and interpreted on the basis of the behaviour of simple nonlinear dynamical systems. It is shown that chaotic energy fluctuations not only modify the variance explained by different variability patterns, but also may either obscure or enhance the multimodality in the distribution of model states in periods as long as several decades. On the other hand, it is argued that significant variations in atmospheric statistics may be induced by anomalies in the atmospheric forcing originated from the lower boundary. QG simulations show that a strengthening of the planetary-wave forcing due to the asymmetry in the divergent circulation between the western and eastern Pacific increases low-frequency variability in the extratropical Pacific, and may turn the distribution of model states from unimodal to bimodal. Anomalies in low-frequency variance related to changes in atmospheric forcing seem to be larger than those originated by internal dynamics, at least in certain periods and regions, suggesting a potential predictability of these statistics by coupled ocean-atmosphere models.