Ma. Leibold, A GRAPHICAL MODEL OF KEYSTONE PREDATORS IN FOOD WEBS - TROPHIC REGULATION OF ABUNDANCE, INCIDENCE, AND DIVERSITY PATTERNS IN COMMUNITIES, The American naturalist, 147(5), 1996, pp. 784-812
I analyze a model of species interactions involving species that compe
te for a single resource and share a common ''keystone predator'' to s
tudy the ''bottom-up'' effects of productivity (potential carrying cap
acity of the resource) and the ''top-down'' effects of factors that af
fect the death rate of the predator on diversity, abundance, and distr
ibution patterns in the resulting assemblages. The model predicts that
coexistence of such species will occur at intermediate productivity (
and at intermediate death rates on the top predator) and that superior
resource exploiters will dominate at low productivity (and high preda
tor death rates), whereas predator-resistant forms will dominate at hi
gh productivity (and low predator death rates). In this model, predato
r and resource densities are ''buffered'' against variation in product
ivity, but the densities of species at the intermediate trophic level
are not. Given a large ''pool'' of potential such species, the model p
redicts a replacement series involving multiple pairs of species rangi
ng from good resource exploiters to predator-resistant forms as produc
tivity increases or predator death rates decrease. In such a case, coe
xistence is most likely among the more similar forms, but abundances o
f similar species are predicted to be negatively correlated. Furthermo
re, the overall density of organisms at all trophic levels is correlat
ed with variation in productivity. Spatiotemporal heterogeneity modifi
es these predictions by allowing more than two species to coexist in t
he presence of a single resource and a single predator and by permitti
ng positive covariation in the abundances of coexisting competitors as
well. These models show the critical role of species compositional tu
rnover in determining food web responses to bottom-up and top-down reg
ulation by productivity and variation in predator death rates. The mod
els also predict unimodal diversity versus productivity curves that de
pend in part on the degree of prey specificity by the top predator.