In most tropical ecosystems, birds play a crucial role as primary dispersers (phase 1) by removing seeds from the plant crown. Following primary dispersal, ground-dwelling animals, notably ants, often disperse seeds secondarily (phase 2). However, the relative contributions of each phase to seed dispersal effectiveness until plant establishment remains unclear. We combined observational and manipulative experiments to quantify the contributions of birds and ants to seed dispersal effectiveness of the ornithochoric tree Xylopia sericea in the Brazilian Atlantic Rainforest. Birds dispersed 17.2% of diaspores, with a dispersal distance (6.6 m ± 6.7 m) twenty times greater than that of ants (0.3 m ± 0.5 m). Notably, birds often dispersed seeds beyond the parental canopy, where seedling survival is higher. Ants were quantitatively more important to seed dispersal, with relative diaspore removal 221.3% higher than birds. Despite the limited dispersal-distance by ants, proximity to ant nests increased seedling survival. Phase 1 led to the establishment of 16.26 seedlings per 100 diaspores produced, and phase 2 increased the number of seedlings established by 28.6% (N = 20.91). Ants enhanced seed recruitment and improved seedling survival both away and near the parental plant, likely due to ant nests providing favorable microsites for establishment. Conversely, plants distant from ant nests performed better when away from the parental plant, underscoring the importance of birds in seed dispersal. Birds promote long-distance dispersal and ants redistribute diaspores to microsites that increases establishment, so the dispersal of X. sericea is highly dependent on both dispersers.

Information from our manuscript:

Study site

The study was conducted in Mata do Paraíso, a forest fragment located in Viçosa, Minas Gerais state, Brazil (20°48'8.79" S, 42°51'30.73" W). This fragment (ca. 300 ha) is a secondary seasonal semideciduous tropical forest in the Atlantic rainforest, with early regeneration areas located along forest edges in dystrophic soils, and advanced regeneration areas form the forest nucleus in eutrophic soils (Meira-Neto & Martins, 2002; Pinto et al., 2007). The climate is CWA subtropical according to Köppen, with rainy summers and dry winters (Alvares et al., 2013). The mean annual precipitation is around 1300-1400 mm, and the mean annual temperature is 19°C (Silva et al., 2010).

Plant species

We used Xylopia sericea A.St.-Hil. (Annonaceae) (hereafter Xylopia) in all experiments. Xylopia occurs mainly in the forest edges and has a clumped distribution (Pinto et al., 2007). The species is an evergreen, xerophytic pioneer native tree, typical of nutrient-poor soils, with six to eight meters in height (Lorenzi, 2002). The average crown area of the plants we studied was 6.75 ± 2.43 m²/plant (mean ± SD, N = 7). The fruits are dehiscent, apocarpous, composed of units called monocarps (Lobão et al., 2005; Figure S1), and fruit ripening occurs between February and April (personal observation). Usually, each monocarp contains one to four oleaginous seeds, each with a grayish coat and two fleshy arils on top (Castellani et al., 2001; Lorenzi, 2002). The seeds present physiological dormancy (Ferreira et al., 2019) and low germination rates, reaching up to 28% in ex situ conditions (Andersen, 1986).

Diaspores quantification

To estimate diaspore production and the proportion of diaspores removed and dropped beneath the plant crown by birds, we put seed traps (0.16 m², mesh 0.2 mm) beneath the crown of seven individuals of Xylopia, at least five meters apart from conspecific fruiting individuals during the 2016 fruit season. The crown area covered by seed traps varied between 14% and 18% (15.14% ± 1.27%).

The seed traps were checked once a week throughout the fruiting period. Their contents were counted and classified into four categories (adapted from Christianini & Oliveira 2010): (1) mature diaspore (with grayish coat and aril attached to the monocarps); (2) immature diaspore (monocarps still closed; monocarps were opened and diaspores counted); (3) preyed diaspore (with evidence of pre-dispersal predation, e.g. fruit broken and/or pierced by insects); and (4) dropped by primary disperser (with evidence of birds handling, e.g. embedded in feces or regurgitations). The latter category can be overestimated, as the diaspores may have been brought from other plants by primary seed dispersers. To estimate the total diaspore production per tree, we counted the number of collected monocarps for each tree. When the diaspore decouples from the monocarp, it leaves a scar on the inner wall. Thus, the number of monocarps in the seed traps allowed us to estimate the number of diaspores produced by the tree by recording the number of scars and diaspores still attached to monocarps. To estimate the bird diaspore removal from the crown of each tree, we summed the number of diaspores of categories 1 to 4, and the result was subtracted from the estimated total production of each tree (adapted from Christianini & Oliveira 2010).

Primary seed disperser interactions on the crown

To investigate birds interacting with Xylopia diaspores, we made 69 h of focal observations using binoculars, between 06 h and 18 h, on 15 individual Xylopia trees during the 2016 and 2017 fruiting seasons. We recorded bird species, numbers of diaspores manipulated, duration of visit, feeding behavior, and estimated post-feeding flight distance (a proxy for minimum seed dispersal distance; Jordano & Schupp, 2000) for all birds that interacted with diaspores. This protocol has limitations, as the gut retention time is longer than the interval between perches. However, this generated minimum estimates that allowed us to compare seed dispersal distances of phase 1 with phase 2 (Camargo et al., 2016). We also noted occasional interactions between birds and Xylopia observed along the study area to increase the sample size (Christianini & Oliveira, 2009).

Secondary seed dispersers and interaction on the ground

To record ant interactions with Xylopia diaspores, we established ten 200 m transects between 50 to 500 m apart, each containing 20 on-ground sampling stations positioned ten meters apart during the 2016 and 2017 fruiting seasons. In each sampling station, we placed one diaspore with whole aril and one with manually removed aril, simulating diaspores defleshed by birds, both marked with a dot of enamel paint (Tamiya®). We set up the sampling stations at 08 h and 17 h, monitoring them continuously for two hours, during which we recorded behavior and collected voucher specimens of all ants observed interacting with diaspores. We measured the removal distance by following ants carrying diaspores to their nests or until they disappeared in the litter (Christianini & Oliveira, 2010). We recognize that this method may underestimate removal distances since we cannot measure the distance of all removals to the final destination.

To assess the relative contribution of ants and vertebrates on the fate of diaspores dropped beneath Xylopia trees, we conducted a paired selective exclusion experiment beneath 30 fruiting plant individuals at least 20 m apart. For this, we used two paired-treatments, one with cages to exclude vertebrates (caged treatment) and another with free access to both vertebrates and invertebrates (open treatment). In both, we used two diaspores marked with enamel paint, one with whole aril and one with aril manually removed (the latter was used to assess the removal rate by seed predators; Pan et al., 2016). For the caged treatment, we placed a cage (15 × 15 × 10 cm) covered in the sides and top with a plastic net (mesh 1.5 cm) on the ground. The open treatment was established 20 cm next to the caged treatment (Camargo et al., 2016). We radially placed five replicates of this paired experiment beneath each Xylopia crown at 08 h and returned after 24 h to count how many diaspores were removed per treatment (Camargo et al., 2016; Christianini & Oliveira, 2010).

To simulate bird seed dispersal and evaluate spatial variability in diaspore removal, we replicate this experiment beneath 30 non-Xylopia trees at least 20 m away from any fruiting Xylopia (Camargo et al., 2016). We considered the difference mean between diaspore removal in open and caged treatments as preyed upon, as there is a high probability that small-sized diaspores, such as those from Xylopia, were preyed upon rather than dispersed by rodents (Vieira et al., 2003).

The removal of diaspores was compared among treatments with a factorial ANOVA, where the dependent variable was the number of diaspores removed per plant (considering the five replications for each plant). The independent variables were: treatment (caged or open), location (below or outside the Xylopia crown) and diaspore type (whole or removed aril). Each fruiting plant and its non-conspecific pair was used as a block (random effect; Camargo et al., 2016). We performed the factorial ANOVA and checked assumptions in R software (likewise the next analyses detailed below, R Core Team 2023).

To evaluate the ant’s role in the redistribution of the seed rain generated by birds, we offered Xylopia diaspores to two captive toco toucans (Ramphastos toco Statius Muller, 1776), a common frugivore in the area, and collected their feces with diaspores 12 h later. On the next day, we put portions of toucan feces containing one diaspore each into the forest ground on a piece of paper (N = 35). The portions were arranged ten meters apart in a linear transect and were covered by a cage to prevent access by vertebrates. The experiment was inspected for two hours (between 07 h and 09 h), and we sampled all ants interacting with the feces and recorded their behaviors. After these two hours, we left the experiment in the field and returned after 24 hours to record the number of diaspores removed (Camargo et al., 2016).

Assessing the viability of manipulated diaspores

We conducted a seed viability test with tetrazolium salt (2, 3, 5 triphenyl tetrazolium chloride; Brasil, 2009) to assess the proportion of Xylopia embryo survival after bird and ant handling. First, we collected 300 Xylopia ripe diaspores from ten different individuals (30 diaspores per tree), then we took a sub-sample of undamaged seeds (N = 180) that was divided into four treatments (N = 45 for each): (1) diaspores without manipulation; (2) diaspores with aril manually removed; (3) diaspores passed through the bird’s gut; (4) diaspores manipulated by ants. For treatments 3 and 4, we offered diaspores to two toco toucans and six laboratory colonies of Acromyrmex subterraneus (Forel, 1893), respectively, and collected the diaspores after 12 h. The colonies of A. subterraneus were maintained in the Federal University of Viçosa. We compared the frequencies of viable and unviable diaspores in each treatment running Chi-Square tests.

Distribution and survival of diaspores

To quantify ant’s contribution to Xylopia recruitment, we followed 27 ant nests (one Camponotus rufipes, one Atta sexdens, five Ectatomma spp., eight Pachycondyla striata, twelve Pheidole spp.), and measured natural germination and survival around them monthly from July 2016 to April 2017. We established a circular plot (radius = 0.5 m) centered on each ant nest entrance (nest treatment) and established a paired-control with the same dimensions and without ant nests nearby in a random location between 2-5 meters away (nest-free treatment; adapted from Camargo et al., 2016). We noted the distance of both plots to the nearest adult Xylopia and if the plots were beneath the crown of a Xylopia. To increase sampling, we marked seedlings and followed up their survival in three transects (100  2 m and 200 m apart) for the entire study period. For each seedling, we recorded the distance to both the nearest ant nest and adult Xylopia, and the location beneath or out of the conspecific crown. We defined seedlings as individuals approximately 5 cm in height, with two to four leaves. Ant nests were located by attracting ants with sardine baits near the seedling stems and following them to their nests. Seedlings within 0.5 m were considered proximate to nests (Böhning-Gaese et al., 1999; Camargo et al., 2016). All seedlings and ant nests were marked with flags.

To compare the seedling abundances near and far from ant nests, we built a generalized linear mixed model (GLMM), using the seedlings number in plots as the response variable, the plot category (nest vs. nest-free treatment) as the fixed predictor variable and the pair identity as the random effect variable. We used a negative binomial error distribution from the lme4 package (Bates et al., 2015) and analyzed the residuals to check assumptions. Additionally, to test whether nearby adult Xylopia influenced seedling survival, we used two logistic regressions: one with all seedlings followed (near and far from ant nest), and another excluding seedlings near ant nests to exclude the potential nest effects on seedling survival. We used seedling survival as the response variable for both analyses and distance to adult Xylopia as the predictor variable. We also compared seedling survival near and far from ant nests. Then, we filtered the dataset to analyze seedling survival below and outside the Xylopia crown. Finally, we removed all seedlings near the nests and analyzed the survival of seedlings below and outside the Xylopia crown without the effect of ant nests. For each analysis, we performed a Kaplan-Meier survival estimator and compared the survival curves with log-rank tests using the Survival package (Therneau & Lumley, 2015).

Bird and ant contribution to the seed dispersal effectiveness

We estimated the number of diaspores removed by primary dispersers per plant by subtracting the total number of diaspores found in seed traps (categories 1-4, subtopic Diaspores quantification) from the total diaspore production estimates. To quantify the relative contribution of secondary dispersers per plant, we summed: (I) the average proportion of diaspores removed by ants in exclusion experiments (caged treatment) multiplied by the proportion of viable diaspores dropped below each plant, and (II) the proportion of diaspores removed by ants from feces during removal experiments multiplied by the proportion of diaspores removed by primary dispersers. To compare seed removal by each disperser group (birds or ants) per tree, we used a paired t-test.

We built a transition probability flowchart from seed to ten-month-old seedlings incorporating combined quantity (described above) and quality data at each phase of the dispersal cycle and predation probability (Camargo et al., 2016; Culot et al., 2015). To calculate the qualitative component of each dispersal phase, we divided it into two subcomponents: (III) the likelihood of diaspore viability indicated by the tetrazolium test after handling by dispersers and, (IV) the survival probability of seedlings obtained through survival analyses, as a function of seed deposition locations. The diaspore predation probability was estimated as the average difference between diaspores removed in paired caged and open treatments, both below and away from the Xylopia crown. The probability of diaspore predation by rodents in bird droppings was estimated as the mean difference in diaspore removal between caged and open treatments in exclusion experiments conducted away from the Xylopia canopy (Camargo et al., 2016).

We estimated the contributions of phase 1 to SDE by estimating the number of seedlings produced after the follow-up period without the occurrence of phase 2 of dispersal (assuming the probability of occurrence of phase 2 is equal to 0). The contribution of phase 2 to SDE was obtained by subtracting the total number of seedlings produced by phases 1 and 2 from the total number of seedlings produced only by phase 1 (Culot et al., 2015).