Ecological networks across interaction types are modular and highly driven by sampling intensity at biogeographical scales

Understanding how the structure of ecological communities varies across biotic and abiotic dimensions is a fundamental goal in ecology. This challenge is now approachable due to the increasing availability of data on community structure across the globe. Ecological communities are often defined with respect to the guilds considered and the interactions they engage in, but it is unclear whether interactions of different types respond similarly to large-scale environmental gradients. Therefore, we don't know whether there exist differences in how the emergent structure of ecological networks varies across biogeographical gradients, depending on their constituent interaction types. Here, using a unique dataset of 952 networks across the globe, we provide a first comparison of network structural metrics and their large-scale variability for five overarching interaction types (feeding, frugivory, herbivory, parasitism, pollination). We show that networks of different types tend to be more modular than expected, but other structural metrics do not deviate from what is expected given the degree distributions of the networks. Our analysis also reveals that network sampling intensity is a particularly relevant factor influencing network degree distribution, and that food webs appear in general more sensitive to environmental factors than other interaction types. By analysing common descriptors from the degree distributions of ecological networks, this study underscores for the first time generalities and differences across different interaction types and their response to environmental, sampling, and anthropic factors.