Initializing a coupled model to observed initial conditions is a difficult challenge, and critical to climate predictions. Sea ice is one of the most sensitive model components to an incorrect initial state. This figure shows a successful initialization of the PCM to observed conditions of 1995. Panel A: Model sea ice cover in January 1995 � the initial state. Panel B: Model sea ice cover in January after running the model forward for 30 years with no constraints. The good agreement with the initial conditions is just one indication that the model drift away from the imposed initial conditions based on observations is small.

Research Objectives
To predict natural climate variability in the North Pacific on the decadal time scale. This is a key region, as it is known that the state of the North Pacific sea surface temperature (SST) field is well correlated with wintertime precipitation and temperatures over the U.S.

Computational Approach
Our primary tool is the Parallel Climate Model (PCM), a fully coupled general circulation model. The individual ocean and atmosphere components from this model are used separately as well. The resolution of the atmospheric component of the models used in this study is T42, while that of the ocean components varies from about 0.5� to 1�, depending on the latitude and longitude.

Accomplishments
In previous work we had identified a preferred 20-year timescale of climate variability in the North Pacific in a coupled ocean-atmosphere general circulation model. Our objective in FY 2000 was to show whether or not this preferred timescale arises out of coupled ocean-atmosphere interactions (and therefore might be predictable by a coupled model) or is a manifestation of various so-called �stochastic resonance� mechanisms, which can give enhanced spectral variability at a preferred timescale even in the absence of coupled ocean-atmosphere interactions.

We ran the ocean model component from the fully coupled model in standalone mode, forced first with the saved surface flux fields (heat, fresh water, and momentum) from the fully coupled model. This control run showed that if the ocean model is driven by these saved flux fields, it reproduces the enhanced spectral peak at the same timescale (20 years/cycle) as found in the fully coupled model. Next, we forced the ocean model with the saved flux fields, but scrambled randomly in time by month. We found that the spectral peak is still present in the scrambled forcing run, indicating that most, if not all, of the enhanced energy at a preferred timescale of 20 years/cycle comes from stochastic resonance mechanisms, and therefore is inherently unpredictable. Further analysis of this data in under way.

Significance
Determining the levels of natural climate variability, and being able to understand the physical processes responsible for this natural noise, are a requirement of any attempt to make an early detection of mankind�s impact on climate.

Publications
T. P. Barnett, D. W. Pierce, M. Latif, D. Dommenget, and R. Saravanan, �Interdecadal interactions between the tropics and midlatitudes in the Pacific basin,� Geophys. Rev. Lett. 26, 615 (1999).

T. P. Barnett, D. W. Pierce, R. Saravanan, N. Schneider, D. Dommenget, and M. Latif, �Origins of the midlatitude Pacific decadal variability,� Geophys. Rev. Lett. 26, 1453 (1999).

D. W. Pierce, T. P. Barnett, and M. Latif, �Connections between the Pacific Ocean tropics and midlatitudes on decadal time scales,� J. Climate 13, 1173 (2000).