Abstract
Species interactions form food webs, impacting community structure and, potentially, ecological dynamics. It is likely that global climatic perturbations that occur over long periods of time have a significant influence on species interaction patterns. Here, we integrate stable isotope analysis and network theory to reconstruct patterns of trophic interactions for six independent mammalian communities that inhabited mammoth steppe environments spanning western Europe to eastern Alaska (Beringia) during the Late Pleistocene. We use a Bayesian mixing model to quantify the contribution of prey to the diets of local predators, and assess how the structure of trophic interactions changed across space and the Last Glacial Maximum (LGM), a global climatic event that severely impacted mammoth steppe communities. We find that large felids had diets that were more constrained than those of co-occurring predators, and largely influenced by an increase in Rangifer abundance after the LGM. Moreover, the structural organization of Beringian and European communities strongly differed: compared with Europe, species interactions in Beringian communities before--and possibly after--the LGM were highly modular. We suggest that this difference in modularity may have been driven by the geographical insularity of Beringian communities.