Discussion
The role of seasonality in shaping species interaction networks remains largely unexplored. While few studies have shown seasonality in small-size organismal bipartite networks (e.g., plant-pollinator networks), limited evidence exists on how seasonality shapes multiplex networks across communities of larger species. Despite increasing evidence that topological changes arise due to rewiring and species turnover (Alarcón et al. 2008; Petanidou et al. 2008; Olesen et al. 2011; Rasmussen et al. 2013; Lopez et al. 2017), as of yet, identifying which process is dominant and more critically, quantifying the contributions of each process, is rarely done across seasons and for multiplex networks. Here, we provide a study on seasonal multiplex networks that shows strong evidence of seasonal change using both modularity scores and interaction turnover, and further show that seasonal rewiring is the main driving process of community changes in stream fish communities.
Given the long appreciated influence of temporal oscillations in theoretical ecology (e.g., Hutchinson 1961), as well as the strong evidence of seasonality in stream fish networks (e.g., Thompson & Townsend 1999), unsurprisingly, we found evidence that seasonality influenced our network structure. In particular, the arrangement of species into close knit groups (i.e. modules), was higher in the Spring than the Fall underscoring that species are assembling differently across seasons. Specifically, in the Spring, species on average confine their interactions more often to a subset of species (i.e. those in their module) as compared to species in the Fall. This may signify that communities in the Spring are more robust to perturbations as the effects of these perturbations may be confined to specific modules rather than spreading to the entire community (Gilarranz et al.2017). The differences in network structure between seasons is also exemplified in the degree of its interaction turnover across seasons. As a score of zero indicates networks are identical and a score of one indicates that networks have no common interactions, our average score of \({\overset{\overline{}}{\beta}}_{\text{WN}}=0.60\), indicates a relatively high differentiation between our seasonal networks. Taken together, both modularity and total interaction turnover metrics provides strong evidence that network topology is indeed changing across seasons.
Beyond classifying network change, identifying the primary drivers of species interactions is essential for predicting community structure (McLeod et al. 2020). In our study, we found that seasonal topological changes to our inferred network were primarily driven by interaction rewiring (c.87%) with a small contribution by species turnover (c.13%). Moreover, the contribution of species turnover to seasonal topological change (βSTWN ) was always smaller than the overall number of species that contributed to turnover;\(7/32\approx 0.22\). Generally, this low value for species turnover indicates that species only observed in a single season are not highly connected to other species. Our results contribute to the small but growing recognition that rewiring occurs at seasonal scales, and that a significant portion of interactions in an ecological network are transient (Carnicer et al. 2009; CaraDonna et al. 2017; Lopez et al. 2017).
Despite highlighting the need to resolve networks along a temporal gradient, our results regarding the potential dominance of rewiring across seasonal networks also provides a general prediction for how these seasonal communities may respond to disturbances. If species subject to seasonality are more strongly driven by rewiring, we may also expect these species to be more robust when subject to other types of disturbances (CaraDonna et al. 2017). For example, Kaiser-Bunburyet al. (2010) found that in plant-pollinator networks, rewiring increased community robustness when faced with community species loss, and Saavedra et al. (2016) found that seasonal interactions play a key role in maintaining the homeostatic state of ecological communities. Indeed, it would be of interest to conservation managers to determine if their systems are robust to future perturbations, given that their system also undergoes seasonal rewiring. However, while in general we expect rewiring to have a stabilizing effect, rewiring has been shown in some cases to have a negative effect on the persistence of both natural and computer generated food webs (Gilljam et al.2015). Hence, future studies should explore whether stability due to rewiring holds for different disturbance types, and under which conditions it switches from a stabilizing to a destabilizing effect.