The diversity of tropical forests is strongly shaped by mutualistic interactions involving plants and frugivores that disperse their seeds. However, it is little known how decreases in the diversity of frugivores can affect seed dispersal patterns, plant community composition and species' coexistence in tropical forest landscapes. Here, we investigated the effects of bird frugivore diversity on seed dispersal of rare plant species and on the magnitude of equalizing effects on the seed rain in open areas within 12 fragmented landscapes in the Brazilian Atlantic Forest. We monitored the production of bird-dispersed seeds and bird abundance in forest fragments, and sampled the seed rain and the activity of birds attracted to experimental tree nuclei established in neighboring pastures. The activity of frugivores in tree nuclei was positively correlated with the diversity of birds recorded in nearby forest fragments, and the seed rain diversity increased with frugivore activity. The proportion of seeds dispersed more frequently than expected by chance in tree nuclei increased linearly with the species' richness of birds. The richness and abundance of active frugivores in deforested areas promoted a seed rain with evenness and diversity up to five times greater than the seed pool available in forest fragments due to the proportional increase in the dispersal of rare plant species and a concomitant proportional decrease in the dispersal of dominant fruiting plants. Furthermore, every additional bird species detected in a site was associated with a 10% increase in the equalizing effect on dispersed seeds’ relative abundance. Our results show that the aggregated behavior of avian frugivore communities on deforested areas results in higher species richness in the seed rain of plant communities and underscore the urgency to reduce bird species' loss and the simplification of their communities in tropical landscapes.

Study sites

The study was conducted in 12 sites of the Paranapanema municipality in São Paulo, southeastern Brazil (23°23' S, 48°43' W, Fig. 1). The sites were located in private farms with cattle pastures and > 30-year-old fragments of secondary semideciduous forests ranged from 12.2 to 98.8 ha. The study area is located within the second most threatened biogeographical zone (<7% forest cover remaining) of the Atlantic Forest (Ribeiro et al. 2009), a global hotspot for biodiversity conservation and tropical rainforest restoration (Laurance 2009, Brancalion et al. 2019). The local climate is Köppen type Cfa, humid subtropical, with hot summer (Alvares et al. 2013). Average annual rainfall, concentrated in the wet summer season (December to March), is 1.407,9 mm, while mean annual temperature is 18 °C (Cielo-Filho et al. 2009).

Tree nuclei in pastures

In December 2016 we established at each site eight 4.5 x 4.5 m plots in cattle pastures to sample bird visits and the seed rain. Plots were established at 10 and 50 m from the nearest forest fragment (Fig. 1). We used plots at these distances because there was not much difference in the parameters of the seed rain between plots at 10 or 50 m from the forest fragments (see Camargo et al. 2020). At each distance class, we established four plots: three with a tree planted and one Control with no tree. Plots were fenced with barbwire to keep cattle off, grasses inside plots were mowed and mechanically controlled throughout the experiment. At the center of each plot, we planted one pioneer tree of reproductive age at least 1.5 m in height to form small tree nuclei on pastures. We planted pioneer trees with different functional characteristics of fruits to maximize interactions with the bird community: Heliocarpus popayanensis Kunth with wind-dispersed seeds, Acnistus arborescens Schltdl. with fleshy fruits composed of 48.3% of carbohydrates, 7.9% of proteins and 0.04% of lipids, and Trema micrantha (L.) Blume with fleshy fruits composed of 48.8% of lipid, 10.7% of proteins and 2.2% of carbohydrates.

Bird activity in tree nuclei

From October 2017 to November 2018 we recorded bird activity in the experimental plots of tree nuclei in the pastures using direct focal observations and video recording. Each plot was observed once a month by one observer (PHSA Camargo) using a pair of binoculars from a distance of 50 m for 20 min, during morning hours (07:30–10:30), totaling 13.3 h of focal observations per site. We complemented these focal observations with 22 camera-traps. Each camera-trap operated 24 h/day for a total of 59,152.9 h. In addition, we also used six GoPro Hero 3 cameras that filmed for about 2 h before batteries were depleted, totaling 193.5 h of filming. All the cameras were rotated, and the sampling efforts of focal observations and filming were equally distributed among the plots and sites to prevent sampling bias. For more details see Camargo et al. (2020).

Seed rain

To sample seed rain we used one 0.25 m² seed trap lined with a 0.2 mm nylon mesh placed at the center of each tree nuclei plot. We applied Formifuu® to the legs of the traps to prevent ant access and covered the traps with a wire screen (2.5 x 2.5 cm mesh) to exclude vertebrate seed predators. Seeds were collected monthly from the traps and counted and identified to the lowest taxonomic level possible in the laboratory with the aid of a dissecting scope, available reference books and reference seed collections for the local flora. We did not include in the analysis the number of grass seeds or those from the pioneer tree species planted in a particular plot (as they may have originated in situ).

Fruit abundance in forest fragments

To quantify the abundances of bird-dispersed fruits in forest fragments, we systematically established 10 5 x 5 m vegetation inventory plots at each forest fragment, totaling 250 m² sampled in each study site (Fig. 1). Overall, we marked a total of 1827 plants including trees, shrubs, vines, herbs, mistletoes; mean ± SD of 152.25 ± 21.09 individuals per site and 7.30 ± 15.96 individuals per species), of which 72.5% (n = 1324) corresponded to 187 bird-dispersed species (38.3 ± 10.2 species per site). From October 2017 to November 2018, we counted monthly the number of ripe fruits of all bird-dispersed species monitored within forest plots. Whenever possible, we directly counted all fruits without extrapolation, but for large trees with a large fruiting crop, we counted fruits in 3-5 branches or portions of the canopy and extrapolated to the rest of the canopy area bearing fruit. For each species, we average the fruit per individual and calculated the fruit density per plot. Then, we extrapolated the average density across plots to obtain the density of fruits in the whole fragment (fruits/hectare). We obtained the abundance of seeds available in the fragment by multiplying the density of fruits by the average number of seeds per fruit of each species.

Bird abundance in forest fragments                                                                                      

To estimate the abundance of birds, we conducted 10-min point counts at the five vegetation sampling plots per forest fragment (Fig. 1). Bird surveys were conducted once a month between October 2017 and November 2018 from 6:00 to 7:30 when all birds visually and acoustically detected within a 50 m radius were recorded. Abundances were calculated by dividing the number of records of each species by the number of sampled point counts and averaged across months.