Even with advances in climate models and computing power, ultimately our ability to provide risk estimates for the next 10-30 years are constrained by fundamental predictability limits of the climate system resulting from its chaotic nature. The presence of intrinsic, long time-scale oceanic modes that appear to modulate regional climate has inspired international endeavors to carry out decadal prediction experiments using assimilated ocean initial states. But these modes are affected by predictability, so it remains unclear at what forecast range initialized states can benefit predictions of even these structures. There is a second limit on the usefulness of initialization. At some range the impact of anticipated increases in greenhouse gases and aerosols on modal amplitudes becomes large compared to initial condition influences making the benefits of initialization relatively insignificant. The project outlined in this proposal is designed to quantify these two predictability limits of prominent ocean modes and to identify the impact of the modes on North American climate. Information about these limits is essential to any program aiming to provide decadal predictions as it provides guidance regarding the confidence users should have in predictions of the most important components of decadal projections. Predictability and the modes of variability are inherent properties of every climate system, including numerical climate models. Usually model predictability is studied with ensemble integrations from slightly perturbed initial conditions. But this approach is impractical when the predictability of many initial states and many climate models needs to be considered. In the proposed project, alternate, more efficient methods will be used so that the predictability of many initial states in many models can be examined. These include employment of linear inverse modeling and analog techniques. Furthermore, the Fluctuation Dissipation Theorem will be employed to identify the impacts of the ocean modes on North American climate. The project will address five overarching questions pertaining to modal predictability in climate models, with the first three constituting the core of the program. These include:

1. What is the structure and predictability of the leading Pacific and Atlantic subsurface modes?
2. How sensitive is modal predictability to model formulation?
3. How do intrinsic modes and their predictability change as the climate changes?
4. What initiates and maintains the dominant modes?
5. How do these modes affect predictability over North America?

These questions will be directed at the behavior of several prominent coupled climate models including CCSM3, CCSM4, three models from international modeling centers, and various CMIP5 models. In addition to quantifying predictability limits and studying processes that contribute to those limits, the study will provide information needed for the design of decadal forecasts and for development of climate forecast models. These include guidance as to the optimal duration of forecasts, the size of ensemble required to represent system evolution, and the regions that need to be well simulated if representation, initiation and forcing of the dominant modes are to be modeled well.