The study of ‘ecological multilayer networks’ is quickly developing. Prominent examples include networks with multiple interaction types, networks that interconnect with other networks, and spatial/temporal networks. Recent advances in network science, in concert with the growing availability of ecological data sets, provide a fertile ground for extending the scope of current ecological network theory. The variety of ecological multilayer networks can introduce however problems with network definitions and applications. For example, what is the difference between studying a collection of multiple but independent networks and networks that are explicitly connected to each other? What is the difference between a multiplex network and a network of networks? Such questions highlight the need for a uniform mathematical framework for defining network structure, enabling cross-study comparisons and a smooth interface with network science. In this talk, we first present general definitions for ecological multilayer networks, and then illustrate with two examples the analysis of the structure and stability of empirical networks.
Multilayer networks encompass two or more ‘layers’ (representing, for example, different types of interactions) and two distinct types of edges: intralayer edges, which represent ‘classical’ ecological interactions; and interlayer edges, which connect nodes in different layers. We first study temporal community structure in a host–parasite network observed during six consecutive summers (layers) in Siberia. We use multilayer community detection to partition the network into six temporal modules, and observe that this structure depends on the temporal changes in species abundance, which are encoded in the interlayer edges. About 47% of the hosts and 35% of the parasites change their module affiliation at least once, most likely reflecting a time-dependent distribution of parasites on hosts. In the second example, we analyze the robustness of a network with two layers: plant–flower-visitors and plant–leafminer parasitoids, which are interconnected via the same set of plants. We examine the robustness of this network based on the proportion of surviving leafminer parasitoids, as a function of the removal of flower-visitors or plants. We show that leafminer parasitoid extinction is slower when flower-visitors are removed, highlighting the importance of considering percolation of extinction via interlayer edges. We also discuss ways in which the robustness analysis can be extended.