We propose to investigate the role of atmospheric noise (due to internal dynamics) at the air-sea interface on the limit of decadal predictability of tropical and North Pacific regions using the NOAA-NCEP Climate Forecast System (CFS). There is increasing evidence from observations and modeling studies that the Earth’s climate system possesses natural variability on decadal timescales. Numerous physical mechanisms have been proposed for decadal variability in the tropical and North Pacific areas. However, it is not well understood which of these mechanisms underpins the decadal predictability and if the state-of-the-art climate models show any decadal forecast skill. One of the ingredients of the physical mechanisms is the stochastic weather noise (due to internal atmospheric dynamics) randomly forcing the ocean through the surface turbulent fluxes. From a climate modeling perspective, the problem is further complicated because it has to be understood as a problem of separating the predictable signal from the unpredictable background noise. We propose to use the interactive ensemble coupling strategy, which is designed to filter out the noise, to investigate the role of noise on the limit of decadal predictability. The CFS has been exploited mostly as a monthly and seasonal forecast tool. It has also great potential for forecasts of the longer timescales, which recommends it as a suitable candidate of a multi-model ensemble forecast system. This proposed project has the following main objectives: 1. investigate the role of weather noise on the internal decadal predictability of tropical and North Pacific SST; 2. produce a set of ensemble decadal hindcasts with CFS between 1981 and 2001; 3. evaluate the effects of systematic errors on the decadal forecast skill. We expect that the results of this study will unify the three elements currently competing to explain factors which limit the decadal predictability of the SST variations. Initial conditions, boundary conditions and weather noise might all be required to explain the reality. The proposed directly contributes to the Climate Variability and Predictability (CVP) in the main priority areas of (i) understanding the limits of decadal predictability, and (ii) developing a decadal climate prediction system.