Conclusion
The approach we take for network reconstruction highlights the utility
of non-traditional methods (e.g., species co-occurrence data) to infer
interactions and thus community structure. Moreover, through our
indirect methods (i.e. abundance measurements over time), we captured
potential interactions other than those that are amenable to direct
observation. Although our network reconstruction is not without its
uncertainties, we demonstrate how EMtree methods can be used to
elucidate multiplex network structure. Overall, we find strong evidence
that differences in our seasonal networks appear to be driven mainly by
rewiring as compared to species turnover. Additionally, while there is
recognition that traits are important factors of community assembly
(e.g., Kneitel & Chase 2004), our findings that maximum length and
piscivore status contributes to a species’ number of rewirings provide
evidence that traits may influence how temporal interaction networks
change. Finally, our study highlights the need to consider communities
as evolving through time. Since seasonal change is capable of
dramatically altering network topology, failing to capture temporal
heterogeneities may cause us to mischaracterize true community
structure.