Data for: Fundamental interaction niches: towards a functional understanding of ecological networks’ resilience

Global change will create new species interactions and alter or eliminate existing ones, a process known as interaction rewiring. This rewiring can significantly affect how ecosystems function. To better predict the future structure of ecological networks, it is crucial to assess their ability to adapt to changes. Here, we introduce two concepts: ‘rewiring capacity’ of a single species (the multidimensional trait space of all its potential interaction partners within a region) and ‘rewiring potential’ of a local community (the total trait space covered by interaction partners of the species at the target trophic level locally). To quantify the rewiring capacity and potential, we apply existing methods for determining species' functional interaction niches in a novel way to quantify species’ and communities’ ability to form new interactions and the functional resilience of interaction networks to global change. To illustrate the applicability of these concepts, we assessed the rewiring capacity and potential of interactions between 1,002 flowering plant species and 318 hummingbird species across the Americas. The rewiring capacity and potential metrics offer a new way to understand and quantify ecosystem resilience, allowing us to map how ecological networks respond to global change.

Methods
The dataset includes a commented code file and filtered and ready-to-use data of plant-hummingbird interactions and traits for running the case study analyses.

We obtained data from 79 complete bipartite plant-hummingbird networks from Dalsgaard et al. (2021) with 1002 plant species, 172 hummingbird species, and 4155 observed pairwise interactions in the Americas. In addition, 10,814 absences of pairwise interactions were inferred from the complete networks if an interaction was not observed between a pair of plant and hummingbird species that occurred in a study site. We transformed the number of observed flower visits to binary presences and absences of interactions to allow the fitting of a probabilistic model. To extend beyond the observed hummingbird species in the 79 networks, we included 146 additional hummingbird species that consume nectar as their main food source (>= 90% in EltonTraits; Wilman et al. 2014) and are included in the checklist of South, Central, or North America (Lepage 2023). We obtained data on the following plant species’ traits: corolla length, floral color following the pollination syndrome concept, nectar concentration, maximum height of the plant, mean seed length, and maximum seed mass (Weigelt et al. 2020; Vollstädt et al. 2025). We use this broad array of plant traits to increase model performance. For those plant species with missing species-level trait data, we imputed traits using an ancestral state reconstruction approach. For this, we constructed a phylogeny using V.Phylomaker (Jin & Qian 2019) and the Smith & Brown (2018) phylogeny as a backbone. We obtained data on the following hummingbird species’ traits: beak length, depth, and curvature, tarsus length, hand-wing index, tail length, body mass, and primary habitat association (Tobias et al. 2022; Dalsgaard et al. 2021). We obtained plant species range maps from the BIEN database (Enquist et al. 2016) or range estimates based on GBIF observations. We obtained hummingbird ranges for both species in the original networks and the additional species from BirdLife International (2022).