The Role of Stochastic Forcing in Modulating ENSO Predictability

Abstract

Predictability analysis of a 1000-yr simulated zonal wind stress anomaly in an intermediate coupled model reveals that low-frequency variations in ENSO prediction are closely linked to changes in spatial structures of the uncoupled atmospheric noise. Enhanced predictability well beyond 1 yr is attained during those decades in which the structures of the stochastic component resemble a certain optimal noise structure, while during other periods the system quickly loses its predictability when the noise has less resemblance to the optimals. The optimal noise forcing can maximize the system's predictability up to 1 yr in advance. Its spatial characteristics are such that maximum variability is located in the western Pacific at about 8°N. Within the limitations of the authors' simple model, the results suggest that changes in ENSO predictability can be explained in terms of changes in the characteristics of the noise forcing, without invoking changes in mean state and coupled dynamics. Therefore, the results offer a null hypothesis for low-frequency variations of ENSO predictability.