U.S. Department of Commerce


Date of this Version



Ecological Modelling 220 (2009) 3147–3162. DOI:10.1016/j.ecolmodel.2009.07.021


Two issues in ecological network theory are: (1) how to construct an ecological network model and (2) how do entire networks (as opposed to individual species) adapt to changing conditions? We present a novel method for constructing an ecological network model for the food web of southeastern Lake Michigan (USA) and we identify changes in key system properties that are large relative to their uncertainty as this ecological network adapts fromone time point to a second time point in response to multiple perturbations. To construct our foodweb for southeastern Lake Michigan,we followed the list of seven recommendations outlined in Cohen et al. [Cohen, J.E., et al., 1993.Improving foodwebs. Ecology 74, 252–258] for improving food webs. We explored two inter-related extensions of hierarchical system theory with our food web; the first one was that subsystems react to perturbations independently in the short-term and the second onewas that a system’s properties change at a slower rate than its subsystems’ properties. We used Shannon’s equations to provide quantitative versions of the basic food web properties: number of prey, number of predators, number of feeding links, and connectance (or density).We then compared these properties between the two time-periods by developing distributions of each property for each time period that took uncertainty about the property into account.We compared these distributions, and concluded that non-overlapping distributions indicated changes in these properties that were large relative to their uncertainty. Two subsystems were identified within our food web system structure (p < 0.001). One subsystem had more non-overlapping distributions in food web properties between Time 1 and Time 2 than the other subsystem. The overall system had all overlapping distributions in food web properties between Time 1 and Time 2. These results supported both extensions of hierarchical systems theory. Interestingly, the subsystemwithmore non-overlapping distributions in foodweb propertieswas the subsystemthat contained primarily benthic taxa, contrary to expectations that the identifiedmajor perturbations (lower phosphorous inputs and invasive species) would more greatly affect the subsystem containing primarily pelagic taxa. Future food-web research shouldemploy rigorous statistical analysis and incorporate uncertainty in food web properties for a better understanding of how ecological networks adapt.