Estimates of Decadal Climate Predictability in the Interactive Ensemble NCAR Climate Model

Great progress has been made in seasonal climate prediction over the past several decades, which is mainly driven by better understanding of the limits and mechanisms of seasonal predictability (e.g. ENSO). However, decadal predictability and prediction remains a challenging problem. The purpose of this paper is to use the newly developed interactive ensemble coupling strategy to quantify how internal atmospheric dynamics noise at the air-sea interface limits decadal predictability. The interactive ensemble approach couples multiple realizations of the atmospheric model to a single realization of the ocean model. The interactive ensemble has proven useful in quantifying how internal atmospheric dynamics limits interannual predictability. Here we focus on decadal time scales and apply the recently developed Nonlinear Local Lyapunov Exponent method to the NCAR Community Climate System Model comparing control simulations with interactive ensemble simulations. We found a generally larger limit of decadal SST predictability with the interactive ensemble compared control, which can be attributed to a significant reduction of atmospheric internal dynamics noise in the interactive ensembles. The specifics of how atmospheric internal dynamics noise limits the predictability is diagnosed in terms of regional phenomena and the relative importance of turbulent heat flux noise versus momentum flux noise and the spatial coherence of the noise.