We use food webs generated by a model to investigate the effects of deleting
species on other species in the web and on the web as a whole. The model
incorporates a realistic population dynamics, adaptive foragers and other
features which allow for the construction of model webs which resemble
empirical food webs. A large number of simulations were carried out to produce
a substantial number of model webs on which deletion experiments could be
performed. We deleted each species in four hundred distinct model webs and
determined, on average, how many species were eliminated from the web as a
result. Typically only a small number of species became extinct; in no instance
was the web close to collapse. Next, we examined how the the probability of
extinction of a species depended on its relationship with the deleted species.
This involved the exploration of the concept of indirect predator and prey
species and the extent that the probability of extinction depended on the
trophic level of the two species. The effect of deletions on the web itself was
studied by searching for keystone species, whose removal caused a major
restructuring of the community, and also by looking at the correlation between
a number of food web properties (number of species, linkage density, fraction
of omnivores, degree of cycling and redundancy) and the stability of the web to
deletions. With the exception of redundancy, we found little or no correlation.
In particular, we found no evidence that complexity in terms of increased
species number or links per species is destabilising.