Skip to main content
Dryad logo

Temporal consistency in interactions among birds, ants, and plants in a neotropical savanna


Campagnoli, Mariana (2022), Temporal consistency in interactions among birds, ants, and plants in a neotropical savanna, Dryad, Dataset,


Interactions among animals and plants are key to understanding seed dispersal, plant regeneration and plant community patterns. These interactions can be dynamic, with changes of species and functional roles across space and time. Despite fluctuations in species abundances and resources over time being the rule in natural communities, most studies approach plant-frugivore interactions as temporally static. We documented changes in bird and ant species in interactions with plants producing carbohydrate- (Miconia rubiginosa) or lipid-rich diaspores (Xylopia aromatica) comparing two fruiting seasons 15 years apart in a savanna in southeastern Brazil. We tested if abundance and body size of dispersal vectors (birds and ants) affected the frequency of interactions and the quantitative effectiveness of seed dispersal (QTC). Rich bird (26 species) and ant (18 genera) assemblages interacted with the diaspores. A core of bird and ant taxa was temporally consistent, and responsible for most seed removal across the two years for both plants. Temporal transience was more common for sporadic partners of interactions, and for species with similar functional roles. Abundance and body size of birds affected the interspecific contributions to QTC. Large-bodied birds dispersed large quantities of seeds of our relatively small-seeded plants, even if their visits were sporadic. For ants, variations in temperature and rainfall across time seems more important in driving variations in the contributions to QTC. The combined effect of birds and ants in the same system help to stabilize the temporal fluctuations in the amount of seeds dispersed. However, ants alone are unlikely to replace the functional roles of vertebrate dispersers. Even though species in the assemblage change either their quantitative contribution to seed dispersal or their presence or absence in the interaction network, species persisting in interaction over time are the most important quantitative partners for plants.


Crop size and immediate seed fate

To estimate seed production, seed fall and seed removal by birds and ants in 2004 and 2019, we used the experimental design described in detail in Christianini and Oliveira (2009, 2010). Briefly, we estimated crop size in Miconia by direct counts of seed numbers in parts of the crown and further extrapolations to the whole plant. For Xylopia we estimated crop size with the aid of seed traps placed beneath parental plants. Traps harvested fallen follicles. We counted the number of scars that seeds left impressed in the inner wall of fallen follicles to obtain an estimate of crop size (Appendix 2) (see Christianini and Oliveira 2010). Although we may have underestimated total crop size in Xylopia, we choose this method because fruits were not uniformly distributed in plant canopies and there was great variation in the number of seeds per fruit. Therefore, counts of fruits in parts of the crown and further extrapolations to the whole plant would have produced no reliable estimates of seed crops. To allow comparisons among plants, seed traps covered a similar percentage (ca. 15%) of the area beneath the crown of each tree (Christianini and Oliveira 2010). For both plants we used seed traps placed beneath 5-10 individuals in each year to estimate seed fall beneath parental plant canopies and seed dispersal rates (Table 1). We removed and sorted seeds inside the traps biweekly. For Miconia, we obtained the number of seeds wasted beneath the canopy by dividing the total number of seeds harvested from seed traps by the fraction of canopy area sampled with the traps. To estimate the amount of seeds dispersed by birds, we subtracted from the total crop size the number of seeds wasted beneath canopy. For Xylopia, the number of seeds removed from the canopy by birds was obtained for each individual plant by subtracting the estimated crop size (based on seed traps) by the number of wasted seeds sampled in traps.

To estimate ant rescue of seeds fallen to the ground, we used removal experiments with the aid of selective exclosures to separate the impact from vertebrates and invertebrates on seed removal (see details in Christianini and Oliveira 2009, 2010). Briefly, we set up from 30 to 42 removal stations each year beneath parental plants located >20m from each other. Each station consisted of two paired treatments: an open treatment (control) and a selective exclosure treatment with a fenced cage allowing only invertebrates to access the seeds. We added ten fruits of Miconia or five seeds of Xylopia per treatment. After 24hs we checked the removal stations recording how many seeds were removed. To obtain an estimate of the proportion of crop size that could be rescued by ants in each year we multiplied the proportion of seeds dropped under the canopy for each individual plant by the mean proportion of seeds removed from caged treatments in each year. We occasionally observed unknown species of crickets, flies and cockroaches at fallen seeds, but we did not witness removal performed by invertebrates other than ants. Therefore, we assumed ants are responsible for all removals under exclosure cages


Plant-frugivore interactions and animal traits

To identify bird dispersers and estimate their contribution to quantitative effectiveness (QTC) we performed focal observations at several Miconia and Xylopia individuals in 2004 (totaling 200.8 tree observation hours), and in 2019 (584.0 tree observation hours) (Table 1; part of these data were presented in Christianini and Oliveira 2009, 2010). For each bird visitor we recorded the species, number of interactions with seeds and handling behavior. Seeds swallowed or carried away by birds were considered dispersed, since local birds that perform these behaviors usually do not destroy seeds in the gizzard.

To investigate the identity and contribution of ant dispersers to QTC, in each year we sampled interactions of ants with seeds on the ground. We placed seeds of Xylopia and Miconia over white filter paper (4 x 4 cm), exposed in five stations (10 m from each other) along each of 8 transects (100 m from each other). Filter paper was used to facilitate visualization on the leaf litter and had no detectable effect on ant behavior. Seeds were checked each 15 minutes during 2hs. We recorded the ants and their behavior towards the seeds (i.e., whether the seeds were removed from the depot or cleaned on the spot and left in place). Observations were made during the day and night, so we were able to sample ants active during both periods. We also recorded interactions ad libitum to increase cover of bird and ant taxa in interactions.

To investigate the influence of animal abundance on QTC, we quantified bird abundances in 2019 based on records obtained in nine point counts lasting 10 min each. Point counts were performed near focal fruiting plants and repeated three times each through the fruiting period. All birds heard or seen (irrespective of distance) were recorded, with special care to avoid double counting (Bibby et al. 1998). For each bird species we used the mean number of records as a single estimate of abundance. Only species recorded in interaction with seeds were considered. For ants, we quantified local abundance as the proportion of 64 pitfall traps that trapped a given ant taxa based on data from Salles et al. (2018) obtained at the same study site. To investigate how the level of bird frugivory relates to visitation rates and variations in QTC, we used data on percentage of fruit in the diet and body masses from Wilman et al. (2014).  For ants we followed the classification of feeding behavior from Christianini et al. (2012) and head length (which is correlated with body size) from Parr et al. (2017). The percentage of fruit in the diet, abundance and body size were incorporated as explanatory variables of QTC in statistical models (see below).


Quantitative effectiveness (QTC)

To compare QTC among dispersers and years we calculated for each plant and year the number of seeds produced by plants (for birds) and number of seeds reaching the ground (for ants), and the mean proportion of seeds removed by birds and ants, and plotted quantitative effectiveness landscapes (Jordano 2017, Schupp et al. 2017). Each dot (data) represented a bird species or ant genus. The contribution by each animal taxa to QTC in a year was given by the number of visits divided by the number of sampled hours (i.e., focal observations for birds and transects for ants), multiplied by the mean number of seeds removed per visit, resulting in a metric of seeds removed per hour for that year and plant species. Seeds that were dropped beneath the parental plant by birds or cleaned at the spot by ants were not considered removed, and we set QTC as zero for taxa performing exclusively these behaviors. We then plotted graphs to compare the total QTC provided by birds solely (QTC1), ants solely (QTC2), and birds and ants together (total QTCt). To test how adding seed removal by ants affected the total QTC received by plants, we compared landscape graphs considering QTC provided by birds only (QTC1) to those performed by birds followed by ants (QTCt). QTCt performed by birds and ants together were obtained by adding the number of bird and ant visits/h, times the mean number of seeds dispersed per visit (by each bird species) added to the total number of seeds they dropped under the canopy times the mean proportion of seeds removed by all ant genera, based on data collected during transects, resulting in an estimate of seeds removed by birds and ants.


FAPESP, Award: 02/12895-8

CAPES, Award: 001