Ensemble forecasting of Indian Ocean Dipole events generated by conditional nonlinear optimal perturbation method

Abstract

In this study, we applied the conditional nonlinear optimal perturbation (CNOP) method to generate nonlinear fast-growing initial perturbations for ensemble forecasting, aiming to assess the effectiveness of the CNOP method in improving the forecast skill of climate events. Our findings reveal a significant improvement in the forecast skill of the Indian Ocean Dipole (IOD) within the CNOP ensemble forecast, particularly at long lead times, thereby extending the skilful forecast lead times. Notably, this improvement is more prominent for strong IOD events, with skilful forecast lead times exceeding 12 months, outperforming many current state-of-the-art coupled models. The high forecast skill of the CNOP method is primarily attributed to its ability to capture the uncertainties in the wind anomaly field in the eastern Indian Ocean (EIO) closely associated with IOD evolution. Consequently, CNOP ensemble members exhibit significant deviations from the control forecast, resulting in a large ensemble spread encompassing IOD evolution. Furthermore, a comparison with the climate-relevant singular vectors (CSV) method in terms of IOD and El Niño–Southern Oscillation (ENSO) predictions reveals the superior performance of the CNOP ensemble forecast. Despite the initial perturbations for ensemble forecasting being generated aimed at improving IOD forecast skill, the CNOP method significantly improves the forecast skill of both IOD and ENSO events, with a greater improvement for ENSO. Additionally, the CNOP ensemble forecast system provides more reliable estimates of forecast uncertainties and exhibits higher reliability with increasing lead times. In conclusion, the CNOP method effectively captures the nonlinear physical processes of climate events and improve the forecast skill.