Seed dispersal networks
During the fruiting season (January–April) we identified all fleshy-fruited plant species in each site. To quantify seed dispersal by birds, we observed one hour per plant and recorded the number of fruits consumed by birds. We discarded the records of seed predators and pulp pickers. Censuses were performed from 6:30 to 10:30 h. In addition, to evaluate seed dispersal by the endemic marsupial, we collected and analyzed their feces. The feces were collected with a mesh on the floor of Tomahawk traps (10 cm × 10 cm × 30 cm). We placed 17 Tomahawk traps at a height of 1-2 m above the ground on shrubs or trees. Traps were arranged in a star grid and separated from each other by 15 m to cover an effective area of one hectare. We controlled the traps daily at dawn during four consecutive days in February and March. To determine the number of fruits dispersed by D. gliroides , we divided the number of seeds of each plant species found in the sample (feces) by the average number of seeds per fruit of the same species. Moreover, to accumulate more observation hours per plant species we used two infrared camera traps (Bushnell trophy cam) per site, programmed to record during day and night. Cameras were placed focusing branches bearing fruits of different plant species and relocated periodically to obtain records of all species present in the site. From the videos, we documented the same variables recorded with direct observations, considering a dispersal event when we visually recorded the consumption of fruit by the seed disperser. Overall, 132 observation periods were conducted for all sites and plant species in 2017–2018, and 158 observation periods in 2018–2019. These numbers varied across species and sites because species differed in their abundance, duration of fruiting period and fruit production (Table S2).