PS 61-177 - Defining and predicting keystone species in the tangled bank

Thursday, August 10, 2017
Exhibit Hall, Oregon Convention Center
Jonathan Borrelli and Martin Wu, Biology, University of Virginia, Charlottesville, VA

Keystone species, as first described by Robert Paine, are those that have a disproportionate impact on the stability of their community relative to their abundance. Nearly 60 years later the term keystone species has been taken to mean anything from a highly connected species in an ecological network to one that performs a specific function. Broadly conceived, most ecologists accept that a keystone species is one whose removal has a disproportionate impact relative to its abundance in a community. To better quantify what makes a keystone species important in a community I explored multiple dimensions of impact that their removal has on simulated communities. Moreover, I expand the traditional search for keystone species beyond food webs into ecological networks containing multiple interaction types. I simulated the dynamics of model communities to steady state and then systematically removed each species, measuring the resulting change in the equilibrium community. I also sought to determine whether the way a species interacts with others in its community can allow us to predict the impact its removal will have.


I measured several ways species removal impacted their communities, including persistence, local stability, change in biomass, coefficient of variation, and change in diversity. The first principal component of impact corresponded to the change in biomass (trophic), stability (competitive), or diversity and stability (multiple interaction types). This result means that different interaction types lead to different ways species removal impacts the system. For example, change in biomass is smaller in communities with multiple interaction types due to compensation effects. Comparing impacts to other traditional measures of keystoneness such as the interaction degree, predation communities have increased impact with degree, but that pattern disappears when other interaction types are introduced. I used generalized linear models to attempt to predict the impact a species’ removal would have as a function of the way they interact with other species (e.g., proportion of mutualistic or competitive links). However, based on cross-validation the predictive abilities of these models were relatively low.