Resource Ecology & Ecosystem Modeling (REEM) Program
Multispecies and Ecosystem Modeling
Figure 1. Network representation of the Gulf of Alaska groundfish food web in 2001.
In a recently accepted publication in the Canadian Journal of Aquatic and Fisheries Sciences, Drs. Sarah Gaichas (Resource Ecology & Ecosystem Modeling (REEM) Program) and Robert Francis (University of Washington) reviewed key concepts from graph theory and network analysis that have not traditionally been used in fisheries applications. They then applied these concepts to the food web of the Gulf of Alaska marine ecosystem to classify its structural properties, which suggest how the ecosystem as a whole may respond to heavy fishing pressure on its components.
Three conceptual models of network structure -- random, small world, and scale free -- each have different implications for system behavior and tolerance to perturbations. Gaichas and Francis constructed two food web network models using detailed quantitative information on the stomach contents of 57 predator (fish) species collected during trawl surveys of the Gulf of Alaska between 1981 and 2002. (Figure 1 shows one of the food webs, constructed using data from 2001.)
The resulting food webs displayed both small-world and scale-free network properties, suggesting that impacts on one species might spread to many through short interaction chains, and that while most food web connections are not critical, a small set of fished species support critical structural connections. Gaichas and Francis concluded that ecosystem-based fishery management should, therefore, first focus on protecting the highly connected species in the network to avoid structural impacts of fishing on the food web.
By Sarah Gaichas
BSIERP Modeling Component FEAST
Dr. Kerim Aydin, REEM Program, worked on substantial development of the central modeling component of the North Pacific Research Board (NPRB) Bering Sea Integrated Research Program (BSIERP). The model, FEAST (forage and euphausiid abundance in space and time) couples fish species, particularly forage species, to nutrients and plankton production on the grid of a regional oceanographic model (ROMS).
With respect to motile predators (fish and higher trophic levels), FEAST is designed around the "landscape approach" for modeling fish foraging, mortality, and growth. The landscape approach (also known as the dynamic habitat approach) treats the space of a model as a series of layers, each layer defining a different spatial (dynamic) quantification of habitat. For example, a temperature layer, a prey density layer, a prey size layer, and a mortality layer may be used, quantifying any given point for its "growth" or "predation" potential thus determining growth and survival (dynamic state variables) of the fish.
The landscape approach has been successful, for example, at predicting the distance at which fish congregate around a front; in a front between warm and cold water, warm or cold adapted fish will approach the front from either side, stopping where gain from frontal concentrations of prey are cancelled out by thermal stress. This can be a powerful tool for modeling dynamic climate scenarios in which fronts shift, break down, or otherwise change over time.
