A large ensemble of decadal predictions using MIROC6

The climate fluctuates on various timescales and in various patterns, giving rise to extreme events over the globe. Skillful predictions of such climate variations would therefore benefit society, and there have been substantial efforts. For the CMIP6 Decadal Climate Prediction Project (DCPP), we performed decadal predictions with ten ensemble members using the Model for Interdisciplinary Research on Climate version 6 (MIROC6). However, since models tend to underestimate signal-to-noise ratio in some sectors, such as the Atlantic, a large ensemble size appears to be required for skillful predictions of those variations. To better understand the predictability on timescales out to a season to a decade, we have prepared a set of initialized predictions using MIROC6 that consists of 10-year-long hindcasts starting every November between 1960-2021, with 50 ensemble members. Compared to the original 10-member ensemble hindcast, both seasonal and decadal prediction skills are broadly improved (e.g., SAT and SLP over southeast China and Scandinavia for the first winter, North and South Pacific SSTs for decadal prediction). Regarding the decadal prediction skill, the impact of initialization is seen up to lead year 7-10 for the North and eastern tropical Pacific Oceans.
Also, building on our experience with decadal climate predictions, we have been working on decadal carbon predictions in recent years. Our efforts on earth system predictions will be introduced as well.