Mutualistic network models are used to understand the structure, stability, and assembly of mutualistic communities. In doing so, these models offer a framework for predicting network disassembly—an important tool as the impacts of mutualist declines and extinctions reverberate through ecosystems worldwide. Existing models offer strong predictions that coextinction will be widespread. However, the mechanisms of network (dis)assembly assumed in these models are untested empirically. This talk focuses on two questions whose empirical answers are critical for improving predictions of network disassembly: Do interactions change in mutualistic networks to compensate for extinct interactions? Which species can cope without their mutualists? We use an accidental experiment involving the seed dispersal mutualism in the Mariana Island chain. Comparing relatively intact and disrupted seed dispersal networks on two islands, we determine the mechanism and strength of compensatory interaction changes. We test the predictions of network models that, after the disruption of certain mutualistic interactions, remaining interactions strengthen and new interactions develop (network “re-wiring”) to achieve strong compensation. Through field experiments, we quantify how dependent plants are on their seed dispersers. We test the prediction that species are equally dependent on the mutualism for survival, making species with few mutualists most vulnerable to coextinction.
Results/Conclusions
We found that seed dispersal interactions have not compensated for the loss of seed disperser diversity. In the disrupted ecosystem, interaction frequencies were far below those observed in the intact network for most species, with the plant species with fewest partners most severely undercompensated. Thus mutualistic network disruption favors species with generalized interactions to the detriment of species with specialized interactions. We did observe network “re-wiring”, but it was typically anti-compensating, with interactions lost in the disrupted network despite the persistence of both partners. Also contrary to current assumptions, we found that behavioral changes—rather than abundance changes—caused larger differences in interaction frequency. In sum, this work suggests that mutualist declines and extinctions may have far greater cascading impacts than currently predicted. However, our analysis of the demographic dependence of plants on seed dispersers showed that species with few partners—which are also most likely to lose their interactions—depend least on seed dispersal for survival. Rather than confirming the predictions of network models that structure and compensation are critical for minimizing coextinction, this empirical work shows that species have evolved to be highly resilient to coextinction by balancing their dependence on mutualism with their risk of losing mutualists.