Mutualistic interactions repeatedly preserved across fragmented landscapes can scale-up to form a spatial metanetwork describing the distribution of interactions across patches. We explored the structure of a bird seed-dispersal (BSD) metanetwork in 16 Neotropical forest fragments to test whether a distinct subset of BSD-interactions may mediate landscape functional connectivity. The metanetwork is interaction-rich, modular and poorly connected, showing high beta-diversity and turnover of species and interactions. Interactions involving large-sized species were lost in fragments< 10 000 ha, indicating a strong filtering by habitat fragmentation on the functional diversity of BSD-interactions. Persistent interactions were performed by small-seeded, fast growing plant species and by generalist, small-bodied bird species able to cross the fragmented landscape. This reduced subset of interactions forms the metanetwork components persisting to defaunation and fragmentation, and may generate long-term deficits of carbon storage while delaying forest regeneration at the landscape level.

Abstract. The data set provided here includes 8,320 frugivory interactions (records of pairwise interactions between plant and frugivore species) reported for the Atlantic Forest. The data set includes interactions between 331 vertebrate species (232 birds, 90 mammals, 5 fishes, 1 amphibian, and 3 reptiles) and 788 plant species. We also present information on traits directly related to the frugivory process (endozoochory), such as the size of fruits and seeds and the body mass and gape size of frugivores. Data were extracted from 166 published and unpublished sources spanning from 1961 to 2016. While this is probably the most comprehensive data set available for a tropical ecosystem, it is arguably taxonomically and geographically biased. The plant families better represented are Melastomataceae, Myrtaceae, Moraceae, Urticaceae, and Solanaceae. Myrsine coriacea, Alchornea glandulosa, Cecropia pachystachya, and Trema micrantha are the plant species with the most animal dispersers (83, 76, 76, and 74 species, respectively). Among the animal taxa, the highest number of interactions is reported for birds (3,883) followed by mammals (1,315). The woolly spider monkey or muriqui, Brachyteles arachnoides, and Rufous-bellied Thrush, Turdus rufiventris, are the frugivores with the most diverse fruit diets (137 and 121 plants species, respectively). The most important general patterns that we note are that larger seeded plant species (>12 mm) are mainly eaten by terrestrial mammals (rodents, ungulates, primates, and carnivores) and that birds are the main consumers of fruits with a high concentration of lipids. Our data set is geographically biased, with most interactions recorded for the southeast Atlantic Forest.

Biological invasions are a threat to the maintenance of ecological processes, including pollination. Plant-flower visitor networks are traditionally used as a surrogated for pollination at the community level, despite they do not represent the pollination process, which takes place at the stigma of plants where pollen grains are deposited. Here we investigated whether the invasion of the alien plant Impatiens glandulifera (Balsaminaceae) affects pollen transfer at the community level. We asked whether more alien pollen is deposited on the stigmas of plants on invaded sites, whether deposition is affected by stigma type (dry, semidry and wet) and whether the invasion of I. glandulifera changes the structure of the resulting pollen transfer networks. We sampled stigmas of plants on 10 sites invaded by I. glandulifera (hereafter, balsam) and 10 non-invaded control sites. All 20 networks had interactions with balsam pollen, although significantly more balsam pollen was found on plants with dry stigmas in invaded areas. Balsam pollen deposition was restricted to a small subset of plant species, which is surprising because pollinators are known to carry high loads of balsam pollen. Balsam invasion did not affect the loading of native pollen, nor did it affect pollen transfer network properties; networks were modular and poorly nested, both of which are likely to be related to the specificity of pollen transfer interactions. Our results indicate that pollination networks become more specialized when moving from the flower visitation to the level of pollen transfer networks. Therefore, caution is needed when inferring pollination from patterns of insect visitation or insect pollen loads as the relationship between these and pollen deposition is not straightforward.