From Bites to Bytes: The Use of Fish Food Habits Data in Ecosystem Modeling

Over the last 30 years, the Alaska Fisheries Science Center has collected fish food habits from over 400,000 fish for the purpose of conducting ecosystem modeling.  The resulting models have included bulk biomass models of Laevastu and Livingston, Multispecies Virtual Population Analysis, Ecopath and Ecosim, bioenergetics models, network models, multispecies statistical models, and high-resolution spatial process models.  Each of these models has grappled with different assumptions in going from raw diet data to modeled feeding rates, whether these are assumptions about functional response, ration, or how uncertainty in the diet data translates to uncertainty in ecosystem projections.  Here, I review statistical techniques and projections made with Alaskan ecosystem models over the last 30 years in reference to (1) how our view of the ecosystem changed as more years of diet data became available; (2) how this has impacted our sampling design and implementation of stomach sampling given management and resource constraints; and (3) how we’ve approached statistical analyses and resulting model uncertainties.   What are realistic expectations when designing a food habits collection for “fisheries management” purposes?