ABSTRACT
The pressing need to conserve and restore habitats in the face of ongoing species loss [1, 2] requires a better understanding of what happens to communities when species are lost or reinstated [3, 4]. Theoretical models show that communities are relatively insensitive to species loss [5, 6]; however, they disagree with field manipulations showing a cascade of extinctions [7, 8] and have seldom been tested under field conditions (e.g., [9]). We experimentally removed the most abundant seed-dispersing ant species from seed dispersal networks in a Mediterranean landscape, replicating the experiment in three types of habitat, and then compared these communities to un-manipulated control communities. Removal did not result in large-scale changes in network structure. It revealed extensive structural plasticity of the remaining community, which rearranged itself through rewiring, while maintaining its functionality. The remaining ant species widened their diet breadth in a way that maintained seed dispersal, despite the identity of many interactions changing. The species interaction strength decreased; thus, the importance of each ant species for seed dispersal became more homogeneous, thereby reducing the dependence of seed species on one dominant ant species. Compared to the experimental results, a simulation model that included rewiring considerably overestimated the effect of species loss on network robustness. If community-level species loss models are to be of practical use in ecology or conservation, they need to include behavioral and population responses, and they need to be routinely tested under field conditions; doing this would be to the advantage of both empiricists and theoreticians.
