Predictability of Indian Ocean Dipole Over 138 Years Using a CESM Ensemble‐Prediction System

Abstract

AbstractIn this study, we performed a long‐term ensemble hindcast from 1880 to 2017 (138 years) using the Community Earth System Model (CESM) and conducted a comprehensive investigation of the Indian Ocean Dipole (IOD) predictability. We found that the CESM can produce the IOD prediction skill comparable to that produced by some of the best state‐of‐the‐art coupled general circulation models, achieving a correlation skill of 0.5 for a lead time of one season over the 138 years. The Brier skill score shows one season of the effective probability prediction skill for the below‐ and above‐normal events and no probability prediction skill for the near‐normal events. The potential predictability of the IOD is much higher than the actual prediction skill; for example, the information‐based potential correlation is as high as 0.8 at a 6‐month lead time, suggesting a large scope for improvement in current IOD predictions. Compared with the dispersion component, the signal component dominates the variation in the relative entropy and the relationship between the potential predictability and deterministic prediction skill. An analysis of the IOD prediction skills suggests that the strength of IOD events plays an important role in the IOD prediction skills, regardless of the measurement metrics used to evaluate the prediction skills. Our study also suggested that the remote forcing from the tropical Pacific and the local sea‐air interaction in the tropical Indian Ocean would be two major sources of IOD predictability.