Experimentally reducing species abundance indirectly affects food web structure and robustness.

Studies on the robustness of ecological communities suggest that the loss or reduction in abundance of individual species can lead to secondary and cascading extinctions. However, most such studies have been simulation-based analyses of the effect of primary extinction on food web structure. In a field experiment we tested the direct and indirect effects of reducing the abundance of a common species, focusing on the diverse and self-contained assemblage of arthropods associated with an abundant Brazilian shrub, Baccharis dracunculifolia D.C. (Asteraceae). Over a 5-month period we experimentally reduced the abundance of Baccharopelma dracunculifoliae (Sternorrhyncha: Psyllidae), the commonest galling species associated with B. dracunculifolia, in 15 replicate plots paired with 15 control plots. We investigated direct effects of the manipulation on parasitoids attacking B. dracunculifoliae, as well as indirect effects (mediated via a third species or through the environment) on 10 other galler species and 50 associated parasitoid species. The experimental manipulation significantly increased parasitism on B. dracunculifoliae in the treatment plots, but did not significantly alter either the species richness or abundance of other galler species. Compared to control plots, food webs in manipulated plots had significantly lower values of weighted connectance, interaction evenness and robustness (measured as simulated tolerance to secondary extinction), even when B. dracunculifoliae was excluded from calculations. Parasitoid species were almost entirely specialized to individual galler species, so the observed effects of the manipulation on food web structure could not have propagated via the documented trophic links. Instead, they must have spread either through trophic links not included in the webs (e.g. shared predators) or non-trophically (e.g. through changes in habitat availability). Our results highlight that the inclusion of both trophic and non-trophic direct and indirect interactions is essential to understand the structure and dynamics of even apparently discrete ecological communities.