Geoscience Reference
In-Depth Information
loses all of its prey items except itself, are deemed to have gone secondarily
extinct and are then removed from the web (
Dunne et al., 2002
). The
simulation next checks to see if any further extinctions occur as a conse-
quence of the loss of those species which went secondarily extinct. Once the
cascade ends, another species is selected for removal (using criteria described
below) and the process repeats itself until the web is reduced to half its
original species richness. Given this algorithm, basal species may experience
primary removals but not secondary extinctions.
The propensity of the web to suffer secondary extinctions (its robustness)
is then quantified as the fraction of species that had to be removed in order to
result in a loss of at least 50% of the species (i.e. primary species removals
plus secondary extinctions). A value of robustness for the web was calculated
in the following way:
N
1
Robustness
ΒΌ
1
2
S
1
where N is the number of removals and S is the original species richness. The
value of robustness can range from 0 where the web collapsed to half its
original species richness following the first removal to 1 in the case where
there are no secondary extinctions.
The sequential orders of the species removed (the primary extinctions)
are based on specific species traits. In this study, we based the orders on three
traits: a species' generality (the number of its prey species), vulnerability (the
number of its predator species) and average body mass, removing species in
both increasing and decreasing order of each trait. There was also a random
order (1000 implementations of which were run), which served as a reference
point. Consequently, in total, we had seven distinct extinction orders. Those
orders based on a species' links updated the sequence following each round
of extinction to take into account links lost in the previous round. If trait
values were tied (e.g. if two species had the same number of prey), then the
one to be removed was chosen at random. All computer simulations were
carried out using R (Code: Thierry unpublished).
III. RESULTS
A. The Weddell Sea Food Web Data Set
The Weddell Sea food web dataset consisted of 488 species (out of which 420
species are consumers, see
Appendix
), and 16,200 feeding links were docu-
mented (
Figure 2
). This included all the food web data available for the high
