The study presents results of seasonal‐to‐decadal climate predictions using a coupled climate model called the MIROC6, which contributed to the CMIP6. MIROC6 is initialized every year for 1960–2018 by assimilating observed ocean temperature and salinity anomalies and full-fields of sea ice concentration and by prescribing atmospheric initial states from reanalysis data. The impacts of updating the system on prediction skill are then evaluated by comparing hindcast experiments between the MIROC6 prediction system and the previous system based on MIROC version 5 (MIROC5). The skill of seasonal prediction is overall improved in association with representation and initialization of El Niño/Southern Oscillation (ENSO), the Quasi‐Biennial Oscillation (QBO), and the Northern Hemisphere sea ice concentration in MIROC6. In particular, the QBO is skillfully predicted up to 3 years ahead with a maximum anomaly correlation exceeding r = 0.8. The prediction skill for the North Atlantic Oscillation in winter is also enhanced, but the prediction still suffers from the model’s inherent errors. On decadal timescales, MIROC6 has a larger fraction of areas of the globe with better surface temperature skill at all lead times than MIROC5, and it has predictive skill in the annual‐mean sea surface temperature (SST) in the North Atlantic and the Pacific. In particular, MIROC6 hindcasts at 2–5 years lead time are able to capture the spatial structure of SST changes in the North Pacific and the eastern tropical Pacific associated with the 1970s regime shift better than MIROC5 hindcasts. The impacts of a large ensemble are also discussed.

In addition, I will briefly highlight our efforts on earth system predictions and mechanism studies on tropical climate variability.

Lieu
École normale supérieure – PSL
24 rue Lhomond – aile Erasme
salle Claude Froidevaux – E314