Understanding which structures and processes contribute to biodiversity persistence in ecological communities is a major issue for both ecological theory and ecosystem conservation. In natural communities, the organization of interactions between species often presents non-random patterns at the origin of complex network architectures. There is growing evidence that some of these non-random patterns can enhance the long-term species coexistence and stability. Moreover, few recent studies have highlighted structural differences between networks depending on the type of ecological interaction involved, i.e. mutualistic or trophic. These findings raise the following questions: (i) To what extent the structure of ecological networks differs depending on the type of interaction considered? (ii) How are community persistence and stability affected by these different structures? (iii) Does the type of ecological interaction involved affect the relationship between network structure and its persistence and stability? We address these questions using both a comparative network approach on 57 empirical datasets describing plant-pollinator and plant-herbivore communities, and a theoretical approach with two dynamical models of mutualistic and trophic network.

Results/Conclusions

Our comparative approach on empirical datasets shows that trophic and mutualistic networks actually differ in their structure. Pollination networks appear more connected, more nested and less modular than herbivory networks. Our theoretical analysis shows that in mutualistic networks, both low modularity and high connectance promote stability. The complete opposite pattern was observed in trophic networks where low nestedness and low connectance enhance stability. Moreover, in both cases most of the effects of connectance are indirectly mediated through related changes in network modularity and nestedness. These results highlight the importance of the complex architecture of networks on long-term diversity persistence and strongly comfort the relevance of modularity and nestedness analysis for community stability. Strikingly, networks characteristics which appear to promote stability in each type of interaction model are also the ones which are mostly observed in each type of empirical interaction network. Such congruence between our empirical and theoretical approaches suggests that in addition to evolutionary processes, ecological factors may contribute to the observed structure of interaction networks.