The domain for the physical model is the eastern Bering Sea shelf, extending out into the Bering Sea basin, and including at a minimum, the Bering Strait, Unimak Pass, and Amukta Pass. The inflow from Amchitka Pass must also be considered. Grid spacing should be approximately 5 to 50 km, varying by location. Finest resolution should be employed near the Pribilof Islands, because of their special biological significance, and near Unimak Pass, to better capture the detailed structure of that inflow. Fine vertical spacing would be needed near the sea surface.

Knowledge of the temperature, salinity and velocity fields is crucial for understanding multiple trophic levels. Primary production depends on the interplay of light, predation, advection and vertical mixing, and varies widely in space. Both vertical and horizontal advection and diffusion supply nutrients, whereas excessive vertical mixing can deprive phytoplankton of adequate light.

Secondary production is in turn strongly dependent on the magnitude of this primary production. Species at higher trophic levels, such as pollock, can be strongly affected by the circulation field as spawned individuals are advected to food-rich or food-poor environments. Hydrographic features such as the cold pool may act to segregate predators from their prey (e.g., adult pollock from juveniles).

The physical models described above can be coupled with a suite of biological, biophysical and ecosystems models. Development of biological models should occur concurrently with development of the physical model. Four types of biological or biophysical models are recommended (See below). Linking outputs from each of these models will allow the examination of ecosystem level questions regarding top down or bottom up controls in determining pelagic production in the Bering Sea.

1. A spatially-explicit, individual-based model (IBM) of the early life stages of key fish species (e.g., walleye pollock and/or sockeye salmon smolts). This model should be coupled with a physical model, and eventually, with both lower and upper trophic level models (see below). Such a model could be used to examine hypotheses relating biotic (e.g., zooplankton prey production, predation, cannibalism) and abiotic (e.g., temperature, salinity, advective patterns, the role of the cold pool and fronts, climate change) factors to the population dynamics of key species.

2. A lower trophic model, a spatially-explicit Nutrient-Phytoplankton-Zooplankton (NPZ) model, should be developed to investigate questions of primary and secondary production as they relate to physical forcing and control by upper trophic level processes. As mentioned above, this model should also be coupled with the IBM.

3. A spatially resolved multispecies model of upper trophic levels, which focuses on adult pollock and sockeye salmon, their predators and competitors. This model would be useful in investigating hypotheses regarding bottom-up and top-down control of ecosystems processes, including the central issue of cannibalism as a major controlling factor in pollock population dynamics. This model may use physical forcing by the hydrodynamic model, lower trophic level input from the NPZ model, and may be coupled with the IBM in order to track the early life stages of key fish species.

4. An aggregated ecosystems model, which includes basic trophic components and a simple spatial structure. This simplified model would be intended primarily as an exploratory tool. Questions about the factors that might lead to shifts in the broad ecosystem structure of the Bering Sea, about net production ("carrying capacity") of the systems, and about the effects of climate change on ecosystem structure could be addressed with this type of model.