Seed dispersal constitutes a pivotal process in an increasingly fragmented world, promoting population connectivity, colonization and range shifts in plants. Unveiling how multiple frugivore species disperse seeds through fragmented landscapes, operating as mobile links, has remained elusive owing to methodological constraints for monitoring seed dispersal events. We combine for the first time DNA barcoding and DNA microsatellites to identify, respectively, the frugivore species and the source trees of animal-dispersed seeds in forest and matrix of a fragmented landscape. We found a high functional complementarity among frugivores in terms of seed deposition at different habitats (forest vs. matrix), perches (isolated trees vs. electricity pylons) and matrix sectors (close vs. far from the forest edge), cross-habitat seed fluxes, dispersal distances, and canopy-cover dependency. Seed rain at the landscape-scale, from forest to distant matrix sectors, was characterized by turnovers in the contribution of frugivores and source-tree habitats: open-habitat frugivores replaced forest-dependent frugivores, whereas matrix trees replaced forest trees. As a result of such turnovers, the magnitude of seed rain was evenly distributed between habitats and landscape sectors. We thus uncover key mechanisms behind ‘biodiversity–ecosystem function’ relationships, in this case, the relationship between frugivore diversity and landscape-scale seed dispersal. Our results reveal the importance of open-habitat frugivores, isolated fruiting trees, and anthropogenic perching sites (infrastructures) in generating seed dispersal events far from the remnant forest, highlighting their potential to drive regeneration dynamics through the matrix. This study helps to broaden the ‘mobile link’ concept in seed dispersal studies by providing a comprehensive and integrative view of the way in which multiple frugivore species disseminate seeds through real-world landscapes.

barcoding_sequencesDNA-barcoding sequences obtained from the surface of the seeds sampled. A column specifies the length of the resulting sequences.coordinates_UTM30S_genotyped_treesUTM30S coordinates of the genotyped wild olive trees in forest and matrix.fully_identified_dispersal_eventsDataset of the fully characterised seed dispersal events, belonging to seeds for which we identified both the frugivore species responsible of dispersal and the source tree. This dataset also includes the dispersal distance (m) and the canopy cover (%) within a 25-m buffer along each dispersal event.genotypes_seeds_treesDataset with the microsatellite genotypes in 11 loci analyzed in dispersed seeds (endocarps), wild olive trees (leaves), and crop olive trees (leaves).seed_rain_densityData on seed density measured during the fruiting seasons 2013–2014 and 2014–2015 in seed traps, in the base of electricity pylons and along transects. This dataset contains information of the distance from the forest edge of each sampling site (0: forest; 1: 0–50 m; 2: 50–100 m; 3: 100–150 m; 4: 150–200 m; 5: > 200 m). This dataset does not include those seed traps placed beneath fruiting wild olive trees within the forest (n = 5) in order to account for actual dispersal events, that is, involving horizontal movement away from the canopies of source trees.seeds_sampledData on the seeds sampled for barcoding and microsatellite analyses. Codes for columns of disperser and source-tree identification are: 1: yes; 0: no; NA: not applicable. This dataset contains information of the distance from the forest edge of each sampling site (0: forest; 1: 0–50 m; 2: 50–100 m; 3: 100–150 m; 4: 150–200 m; 5: > 200 m).README_Gonzalez-Varo_etal_2017_MolecularEcologyInformation on the content of each data file.