Identifying keystone plant resources (KPR) is a contentious issue in ecology and conservation. Despite recent advances provided by mutualistic networks, we still lack studies addressing large-scale identification of keystone plants. We developed a novel quantitative framework for the large-scale identification of KPR that combines centrality and effects of simulated removals on networks properties. We built a database with 38 fruit-frugivore networks comprising 6180 pairwise interactions from Neotropical forest and non-forest ecosystems ranging from sea level to 2,500 m a.s.l.. Effects of random removal simulations and removal of keystone taxa candidates on nestedness, modularity and niche overlap were assessed. Furthermore, we tested whether changes in network descriptors responded to both latitudinal and elevational gradients. From a total of 373 genera on networks, only species in six genera were ranked as potential keystones. The simulated removal of species in Miconia (Melastomataceae), Cecropia (Urticaceae) and Byrsonima (Malpighiaceae) led to significant changes in nestedness, modularity and niche overlap, suggesting that these plants play important roles in maintaining the structure of Neotropical fruit-frugivore interaction networks, regardless of latitude or elevation. Our results confirm the keystone role of previously-recognized taxa, recognize overlooked ones, providing support for their role at a biogeographical scale, and partially challenge taxa traditionally identified as keystone resources for frugivores. Our study has implications for conservation and restoration of Neotropical ecosystems and provides a new framework for large-scale identification of keystones in any type of ecological networks.

We searched papers published from 1945 to 2017 in the Web of Science, Scopus and SciELO databases, using the following terms and their combinations in the title, abstract and keywords: ecological networks, fruit-frugivore, frugivore community, frugivory, Neotropics, plant-seed disperser interaction, seed dispersal. We also checked the Interaction Web Database (www.nceas.ucsb.edu/interactionweb/), the Web of Life Ecological Networks Database (http://www.web-of-life.es/) and the Atlantic-Frugivory dataset (Bello et al., 2017) to make sure we did not miss any study. Finally, we searched the references listed in the studies surveyed to build an extensive and updated database on Neotropical fruit-frugivore interactions.

To be included in our database, a study must have reported interaction matrices between plant and bird species at the community-level, with sampling encompassing climatic seasonality for at least one year, to capture the seasonal variations in fruit-frugivore interactions (Ramos-Robles, Andresen & Díaz-Castelazo, 2016). Whenever the data were not clear, neither readily available, we asked the correspondent author for permission to use the original dataset. Some studies were then excluded from our database, because the authors did not reply or provide data.

We organized the interaction datasets obtained for each site as interaction matrices, in which the plant species are represented in the rows and the bird species in the columns (nodes), and the interaction between them computed in the cells (links) (Bascompte et al., 2003). Since fruit-frugivore interaction datasets vary strongly in sampling effort and criteria for accounting interaction frequency (e.g. number of seeds present on feces; number of fruits consumed; number of visits; stomachal content), we transformed all data into binary matrices, where the presence of fruit-frugivore interaction is represented as 1 and the absence as 0 and, therefore, ensuring that all networks report the same type of biological information (Bascompte et al., 2003; Jordano, 2016). Our literature survey resulted in 29 studies that matched our inclusion criteria, resulting in 38 fruit-frugivore networks (Appendix S1). These networks encompassed 31 forests and seven non-forests ecosystems, extending from 19º36' N to 27º30' S, with elevation ranging from 30 to 2,500 m a.s.l. (Figure S1; Table S1).

When using this data please cite our publication 'Searching for keystone plant resources in fruit-frugivore interaction networks across the Neotropics' athored by Messeder, Guerra, Dáttilo & Silveira 2020 published in Biotropica. Additionally, please also cite the original publications which the data were derived.

Please note that data from Hawes & Peres 2014 is available upon request to the original authors.