Predicting sediment organic carbon and related food web types from a physical oceanographic model on a subarctic shelf

In changing environments, conservation planning for bottom-feeding marine predators requires estimating the present and future spatial patterns of benthic communities. In the northern Bering Sea, we used the Regional Ocean Modeling System (ROMS) to hindcast near-bottom flows that redistribute settled phytodetritus and organic sediments, which in turn strongly affect the dispersion of 3 food web types that differentially favor spectacled eiders Somateria fischeri, walruses Odobenus rosmarus, or gray whales Eschrichtius robustus. Using data collected between 1994 and 2010, we interpolated spatial patterns of sediment organic carbon from field samples and correlated them with water depths and modeled flow velocities, temperatures, and salinities. In the deeper (mean 63 m) southern study area with weak net flows, hindcast near-bottom currents had negligible effects on patterns of sediment longer-term organic carbon (LTOC); instead, regional depth gradients and local bathymetry were the best predictors (r(2) = 0.72-0.85 among 7 years). In that area, climatic variations in total primary production would affect the areal extent of different LTOC levels, but not the core locations of persistent patches defined by depth. In the shallower (mean 39 m) northern study area with much faster flows, seafloor depth had negligible effects, and patterns of LTOC depended mainly on currents (r(2) = 0.48-0.55 over 2 years) that are subject to climatic changes in winds. Based on ranges of LTOC for different food web types, substantial portions of both areas must be conserved to ensure annual availability of all 3 types. Regional ocean circulation models driven by downscaled climate models provide important opportunities for projecting spatial patterns of key benthic habitats in this region.