A nonlinear local Lyapunov exponent method based on monthly sea surface temperature data is employed to explore the predictability limit of warm El Niño–Southern Oscillation (ENSO) events. Results using observational data show an asymmetry of the p...
A nonlinear local Lyapunov exponent method based on monthly sea surface temperature data is employed to explore the predictability limit of warm El Niño–Southern Oscillation (ENSO) events. Results using observational data show an asymmetry of the predictability limit between the developing and decaying stages of the warm ENSO phase. To wit, predictability of the developing stage of warm ENSO events is found to approach a limit of 10 months, less than that of the mature and decaying stages. This asymmetrical predictability limit is also found in a long climate model simulation and may explain the asymmetry in operational forecast skill of warm ENSO events. Through exploring the error growth rate as represented by nonlinear local Lyapunov exponent and the instantaneous error growth rate, it is shown that error growth, especially during the first 8‐month lead forecasts, is the primary contributor to the asymmetry of the predictability limit of warm ENSO events.
El Niño events have a characteristic development and decay cycle, whereby most events start developing in late boreal spring, peak in boreal winter, and subsequently decay. Model skill in operational forecasting of El Niño events depends on whether the forecast is done during the developing or decaying stage of the event. This study shows that this asymmetry in forecast skill is associated with an asymmetry in the inherent predictability limits of the developing and decay stages of warm El Niño–Southern Oscillation events. These inherent predictability limits stem from the fact that small perturbations of a dynamical system may grow into large forecast errors after a period of time.
The warm ENSO phase exhibits an asymmetry in its predictability limit between the developing and decaying stage
The asymmetrical inherent predictability limit explains the asymmetry of operational forecast skill for warm ENSO events
Nonlinear error growth, especially during the first 8‐month lead forecasts, contributes to the asymmetry of the predictability limit