Understanding Decadal Climate Predictability in the Global Ocean
Due to considerable social and economic implications, there is a continuously increasing demand for decadal climate predictions. However, decadal predictions remain a challenging problem largely owing to the insufficient knowledge of decadal predictability. The overarching goal of this work is to understand decadal climate predictability in the global ocean. Specifically, this work is motivated by current challenges in decadal predictability and has three major objectives.
The first objective is to investigate the limits and mechanisms of decadal predictability, particularly the unresolved role of internal atmospheric noise in decadal predictability. The interactive ensemble (IE) coupling technique is used to quantify how the internal atmospheric noise at the air-sea interface impacts decadal predictability. We apply the nonlinear local Lyapunov exponent method to the Community Climate System Model comparing control simulations with IE simulations. The global patterns of decadal predictability are shown for both models and observations and we find that the impact of internal atmospheric noise on decadal predictability is not a linear question and largely dependent on the background coupling and dynamics.
The second objective is to address the so-called “signal-to-noise paradox”. The essence of the paradox is that the signal-to-noise ratio in models can be unrealistically too small and models seem to make better predictions of the observations than they predict themselves. We introduce a Markov model framework to represent the ensemble forecasts and reproduce the paradox, which is primarily dependent on the magnitude of the persistence and noise variance between models and observations. The Markov model framework is applied to the North Atlantic Oscillation index based on the coupled models from the fifth Coupled Model Intercomparison Project (CMIP5). The results suggest the widespread existence of the signal-to-noise paradox that may exist at different timescales.
We re-examine decadal predictability from the lens of the signal-to-noise paradox in the context of CMIP5 models for the sea surface temperature and sea level pressure fields. We demonstrate that decadal predictability is generally underestimated in CMIP5 models, which is closely related to the paradox. Models are likely to underestimate decadal predictability in regions where it is likely to have the paradox.
The third objective is to determine if this underestimate of decadal predictability is, at least partially, due to missing ocean mesoscale processes and features in CMIP5 models. A suite of coupled model experiments is performed with an eddy-resolving and eddy- parameterized ocean component. Again, the results are discussed through the lens of the signal-to-noise paradox. Compared with eddy-parameterized models, less chance of existence for the paradox is seen in eddy-resolving models, particularly over eddy-rich regions, where increased decadal predictability is also identified. This enhanced predictability is possible due to the enhanced vertical connectivity, which is demonstrated through ocean vertical structure and the relationships between the deep ocean and surface processes. We argue that the presence of mesoscale ocean features and associated vertical connectivity significantly influence decadal variability, predictability, and the signal-to- noise paradox.
Overall, this work summarizes major challenges facing decadal predictability and aims to understand decadal predictability from the perspectives of the internal atmospheric dynamics, signal-to-noise paradox, and ocean mesoscale features. These findings not only suggest potential opportunities to advance decadal climate predictability in future studies, but also provide guidance on future model development and initialization.