A simulation of variability of ENSO forecast skill

Abstract

In many prediction schemes, the skill of long-range forecasts of ENSO events depends on the time of year. Such variability could be directly due to seasonal changes in the basic ocean-atmosphere system or due to the state of ENSO itself. A highly idealized delayed oscillator model with seasonally varying internal parameters is used here to simulate such behavior. The skill of the artificial forecasts shows dependence on both seasonal and ENSO phase. Experiments with ENSO phase-locked to the seasonal cycle, but with no seasonal variation of model parameters, show that the ENSO cycle alone can induce variability in skill. Inclusion of seasonal parameters enhances seasonal skill dependence. It is suggested that the seasonal skill variations found in practice are due to a combination of seasonal changes in the basic state and the phase-locking of the ENSO and annual cycles.