Biophysical uncertainty in the Bering Sea as a function of space, time, and trophic scale

Albert J Hermann, University of Washington, Joint Institute for the Study of the Atmosphere and Oceans, Seattle, WA, United States, Georgina A Gibson, Univ Alaska Fairbanks, Fairbanks, AK, United States and Kerim Aydin, NOAA/Alaska Fisheries Science Center, WA, United States
Predictive skill is a strong function of scale. With notable exceptions (e.g. local tides), we typically anticipate better skill at predicting the broad features of the ocean, as compared with its spatial, temporal, or biological details. Indeed, for some fisheries management applications it is precisely those broad features which are of greatest interest. While field measurements of the ocean across a wide range of space and time scales are limited, we have ready access to multi-scale synthetic datasets from dynamical biophysical models. As in weather forecasting, we can utilize multiple forecast realizations of a particular domain to get a sense of the uncertainty (the lack of predictability) of projected futures at various scales. Here we examine multiple downscaling realizations of biophysical dynamics for the Bering Sea, to estimate how the "cone of uncertainty" - that is, the growth in variance among projected futures - is affected by spatial, temporal, and trophic aggregation. Specifically we examine three multi-decadal projections of the Bering Sea, each driven by a separate global CMIP model projection, and three seasonal projections of the Bering Sea, each driven by a separate global realization of the NCEP Climate Forecast System. Finally, we consider how EOF analysis might be utilized to derive the most predictable aggregates for a region.