Comparative assessment of predictability over mid-latitudinal and tropical climate using statistical models: Example over North Carolina and Indonesia

In this study, we address the fundamental question of whether in the statistical sense the midlatitude and tropics regimes have similar predictability? In a recent study, we showed high simultaneous correlations between the tropical Pacific sea surface temperature (SST) anomalies and the North Carolina precipitation anomalies during El Nino/Southern Oscillation (ENSO) events using pristine statistical approaches: Empirical Orthogonal Function (EOF) and Canonical Correlation Analysis (CCA). In general, for North Carolina ENSO-related precipitation anomalies are positive from November to May and negative between June and October. Similar techniques have been applied for investigating the relationships between ENSO and rainfall anomalies over Indonesia. It is found that high positive correlations mainly occur from September to March, while much lower correlations from April to August.

For these two contrasting regimes, we further investigate the previously tested EOF and CCA approaches with different lag-periods (1 to 6 months). We show that there are significant differences in predictability between the midlatitude and tropical regions. In general, in the tropics the skills of predictability decrease systematically for higher lag times, while in the midlatitude higher lag times do not necessarily decrease the skills of predictability. A detailed discussion on the implications of these results on the overall predictability of these two regimes is presented. A proposal to use dynamical statistical approaches as well as nonlinear approaches for regimes with lower predictability is presented for assessing and interpreting climate change scenarios.