* [A new approach to studying ENSO predictability: Conditional nonlinear optimal perturbation](http://link.springer.com/article/10.1007/BF03184224).

**Abstract.** A new approach, the conditional nonlinear optimal perturbation (CNOP) is introduced to study the predictability of El Niño-Southern Oscillation (ENSO) using a theoretical coupled ocean-atmosphere model. The differences between CNOP and linear singular vector (LSV) are demonstrated. The results suggest that the nonlinear model and CNOP are superior in determining error growth for studying predictability of the ENSO. In particular, the CNOP approach is used to explore the nature of the ‘spring predictability barrier’ in ENSO prediction.