Our study was conducted at the Cascata Experimental Station (EEC) in Pelotas, Rio Grande do Sul, in southern Brazil, 160 ma.s.l. According to the Köppen climate classification system, the region's climate is defined as Cfa, a humid temperate climate with hot summers, without a dry season (Alvares et al., 2013), and with a yearly average temperature of 18.9 ∘C and an average rainfall of 1794.6 mm. The study area has approximately 0.4 ha of fallow fields, previously covered by semi-deciduous seasonal forest, and is near a riparian forest (Fig. 1; IBGE, 2012). The riparian forest is characterized by native species such as Schinus terebinthifolia, Allophylus edulis, Myrsine coriacea, and Gymnanthes klotzschiana and alien species such as Pittosporum undulatum (Australian cheesewood), Pinus sp., Eucalyptus sp., and Hovenia dulcis. The landscape is composed of forest fragments of different sizes and regeneration stages and has been immersed in a consolidated agricultural matrix for more than a century.

We made 12 artificial bamboo perches (4 m tall: 3.5 m above the ground and 0.5 m buried), with a “triple T” shape and three landing surfaces, 1 m long each and arranged at 1.5, 2.7, and 3.3 m from the ground surface (Fig. 2a). To evaluate seed rain, seed traps of 1 m2 were used, made with water-permeable fabric (anti-insect mesh), and placed 0.7 m from the soil (Fig. 2b). The perches (P) and control seed traps (C) were arranged in the fallow area, at three different distances from the edge of the forest fragment: 5 m (P5m and C5m), 25 m (P25m and C25m), and 50 m (P50m and C50m). At each distance, four seed traps were arranged under the perches, and four control seed traps were interspersed with these at 7.5 m from the perches. Furthermore, we placed four forest seed traps (F) at 5 m inside the edge of the forest fragment, at 14 m intervals (Fig. 2c).

We conducted the experiment for 12 months (between 2017 and 2018). We performed the collection, screening, and identification of diaspores every 2 weeks, with a total of 24 samples per trap. For taxonomic identification, the collected material was compared (observation with the naked eye and stereomicroscope) with diaspores from the seed library of the Embrapa Clima Temperado and the Forest Sciences Laboratory of Faculdade de Agronomia at the Universidade Federal de Pelotas (UFPel), with the help of specialists and specialized bibliographic material (Lorenzi, 1992, 1998; Carvalho, 2003; 2006, 2008, 2010, 2014; Frigieri et al., 2016). In addition, we recorded all fruiting plant species dispersed by animals along transects at the edges and in the interior of nearby forest fragments during all four seasons. Plant species were classified based on the taxonomic system APG IV 2016 (Chase et al., 2016).

We observed the birds that used the artificial perches for 20 h per season per year (80 h of sampling effort) in four periods: dawn, late morning, early afternoon, and dusk. The focal observations were conducted with the aid of the Celestron binocular, Ultima model (8 × 42 magnification), at a distance of 20 m from the nearest perch, from a point where all perches were seen at the same time with no interference in bird activity. The visiting species and the length of stay on the perch were recorded for each visitation event. The taxonomic nomenclature of birds was based on the “Annotated checklist of the birds of Brazil by the Brazilian Ornithological Records Committee” (Piacentini et al., 2015).

We classified all seeds into zoochoric and non-zoochoric following the seed dispersal syndromes proposed by Van der Pijl (1982) and into native and alien species based on the “List of Invasive Alien Species of the State of Rio Grande do Sul” (SEMA, 2013) and the IABIN Invasives Information Network. The classification of birds regarding diet was based on Sick (1997) and on relevant bibliography (Francisco and Galetti, 2002; Pizo, 2004; Jesus and Monteiro-Filho, 2007; Tubelis, 2007; Pascotto, 2007; Howe, 2017).

To assess the richness and abundance of seeds, we fitted a generalized linear mixed model (GLMM), using the individual samples of each treatment. As these variables correspond to counts (i.e., number of species and number of seeds, and to account for overdispersion in the models residuals, we fitted negative binomial models as they showed better fits and lower deviance than Poisson models (Zuur et al., 2009). We fitted two different models for each response variable: one model considering the treatment (control seed traps, forest seed traps, and artificial perches) as a fixed variable and one model considering the distance of forest seed traps (5 m) and the distinct distances between perches and the forest fragment as a fixed variable. In order to account for non-independence of repeated samples from the same seed trap, we considered each seed trap as a random effect in the GLMMs. We plotted the values of richness and abundance of seeds in artificial perches, forest seed traps, control seed traps, and the distinct distances of perches (5, 25, and 50 m) in box plots. After, we used multiple comparisons based on a GLMM (negative binomial model, considering seed traps as a random effect) to compare richness and abundance among these treatments. These analyses were conducted in the R environment (R Core Team, 2022) using functions from the lme4 package (Bates et al., 2015).

To evaluate the differences in the composition of seeds deposited in artificial perches and inside the forest fragment, a non-metric multidimensional scaling (NMDS) was performed using the Bray–Curtis quantitative dissimilarity index (IsBC) (Magurran, 2013). The fit of the NMDS ordinations was quantified by a value of stress. In addition, to test for differences between seed species groupings obtained with the NMDS, a multivariate permutational variance analysis (PERMANOVA) was performed using the same similarity measurement and 999 randomizations (PAST 3.20 software; Hammer, 2001).

For the distinct distances between perches and the forest fragment, the diversity profile was calculated using the estimated model through the Chao1 index with 1000 repetitions, in which the overlap between the confidence intervals of the communities indicates the absence of significant difference (Chao and Jost, 2012). The diversity profile is represented by species richness (Q0) without considering their relative abundances, the exponential of Shannon entropy (Q1), Simpson's dominance index (Q2), and the Berger–Parker index (Q3). The more Q increases, the higher the value given to the dominant species (Hill, 1973; Gotelli and Chao, 2013; de Vries et al., 2021). The iNEXT software was used to calculate the diversity profile (Chao et al., 2016).

A total of 24 655 seeds were sampled, distributed in 23 families and 34 species of plants. For 6 taxa, it was possible to reach only the family or genus, in addition to 11 morphospecies, totaling 51 different diaspores (Appendix A). For morphospecies, it was not possible to identify their dispersal syndrome. The most abundant species in the seed rain were Solanum americanum (n=9794), Ficus organensis (2993), Pittosporum undulatum (2835), and Schinus terebinthifolia (2157).

Under the artificial perches (P), 22 892 seeds were deposited (mean ± SD, 1907.25 ± 1347.53) with a richness of 46 species (24.58 ± 4.87); in the control seed traps (C), 256 seeds (21 ± 34.23) of 5 species (1.58 ± 1.08) were deposited; and in the forest seed traps (F), 1507 seeds (374.5 ± 323.04) of 29 species (17 ± 3.36) were deposited. According to GLMM analysis, there was a significant difference in seed abundance and richness between control seed traps (C), forest seed traps, and artificial perches (Table 1 and Fig. 3a and b). For both response variables, higher values occurred in artificial perches, followed by forest seed traps and control seed traps.

Regarding the seed rain, the GLMM indicates significant differences between forest seed traps and perches at different distances to the forest fragment and among perches at all distances (Table 2 and Fig. 3c). The largest seed deposition occurred at a distance of 50 m (3060 ± 1575) from the forest fragment, followed by 25 m (1897 ± 813) and 5 m (765 ± 86). Considering richness among artificial perches at different distances and seed traps in the forest fragment, the GLMM showed a significant difference between forest seed traps and perches at 25 and 50 m but not with perches at 5 m (Table 2). Richness among artificial perches was significantly different between perches at 5 m and perches at 25 and 50 m, which showed no significant differences (Fig. 3d). Perches at 25 m from the forest edge had the highest number of species (27.50 ± 4.34), followed by perches at 50 m (27.25 ± 1.70) and by perches at 5 m (19 ± 1.41).

The NMDS indicates segregation in the composition of seed species between the forest seed traps and the different distances of artificial perches in the fallow area and that perches at 25 and 50 m have greater similarity in the seed species deposited under these structures (Fig. 4). The overall result of PERMANOVA shows differences (pseudo-F3,12=4.797; p=0.001) in the similarity of seeds between the forest seed species and artificial perches. In this way, the forest seed traps had a distinct species composition when compared to all perches' distances, and among perches there were significant differences between perches at 5 m when compared with 25 and 50 m, which were similar to each other (Table 3).

There was no significant difference in the diversity profile between the estimated richness (q=0) due to the overlap of the intervals. The forest seed traps presented higher equitability and lower dominance (q1, q2, and q3), indicating a greater diversity of the plant community in the forest area compared to the other treatments (Fig. 5), which were similar between the different distances. The high dominance of some species in the seed rain was observed in artificial perches, in which 79 % of the seeds were restricted to five species in the perches at 50 m (Solanum americanum, Ficus organensis, Pittosporum undulatum, Myrsine spp. 1, and Schinus terebinthifolia), and 71 % of the seeds came from only three species, a fact that is repeated for the perches at 5 m (Solanum americanum, Schinus terebinthifolia, and Pittosporum undulatum). The most abundant species for the perches at 25 m were Solanum americanum, Pittosporum undulatum, and Schinus terebinthifolia.

There were two invasive plant species in the experiment, namely Pittosporum undulatum (2835 seeds) and Lonicera japonica (61 seeds). Pittosporum undulatum represented 11.49 % of the total seed rain in the experiment, occurring at all distances and structures, with higher deposition on perches at 50 m. As a climber plant, Lonicera japonica occurred predominantly in the forest (96 %) and was absent in control seed traps.

We observed 24 species from 12 bird families using artificial perches during the experiment, and several eating habits were recognized (Appendix B). Of this total, 10 are potentially seed dispersers (Table 4). Of the 80 h of focal observation in artificial perches, 45.85 % of the time at least one bird was using the structures, corresponding to 36 h and 41 min of use. Of this total, the species Tyrannus savanna, T. melancholicus, Zonotrichia capensis, Pitangus sulphuratus, and Empidonomus varius represented 97 % of the length of stay on the structures, with Pitangus sulphuratus being the only species that used perches throughout the year.

The Tyrannidae family was most frequently observed during the experiment. Five of the six species that used the structures are considered seed dispersers. Tyrannus savanna and T. melancholicus occurred in the experiment only in spring and summer as they are migratory birds to the region (Timm and Timm, 2016).

We showed that artificial perches are efficient structures for attracting seed dispersers, increasing the richness and abundance of seed species (≈ 90× more seeds in artificial perches) at least up to a distance of 50 m from the forest fragment. The diversity of seeds is higher inside the forest fragment, and the species composition in the seed rain differed between the forest fragment and perches but was similar between perches at 25 and 50 m from the edge of the forest.

The farther the artificial perches were from the forest fragment edge, the more abundant and richer in species the seed rain was. These increasing values of richness and abundance of seeds may be related to the landscape features that provide several feeding resources to the birds with generalist habits that used the perches to move across the landscape. Distinct landscape elements, such as isolated trees, riparian forests, and forest patches in different successional stages, can enhance the dispersal of seeds (Zwiener et al., 2014). During the study, we observed the movement of birds to the perches, mainly crossing the open areas instead of using the closer forest fragment, although some birds accessed the fragment after having used the artificial perches. This pattern probably contributed to the greater abundance of seeds in the more distant perches embedded in the open matrix of the studied area. In addition, the availability of landing structures in fields favors the attraction of birds and the increase in seed deposition on the site, promoting a greater seed rain in these places than in open areas because these structures provide shelter, rest, and foraging for birds (Holl, 1998; Graham and Page, 2012; Alencar and Guilherme, 2020). Thus, perches function as stepping stones, connecting close fragments that favor gene flow (Sant'anna, 2011; Pustkowiak et al., 2021).

Nevertheless, the efficiency of seed rain at distinct distances from the forest edge encompasses several factors like (i) fruit and seed availability, (ii) the presence of forest fragments and isolated trees in the landscape, (iii) which seed dispersers there are in the region, (iv) the distribution of perches in the landscape, and (v) visual acuity of the dispersers (Brancalion et al., 2015; Almeida et al., 2016; Carlo and Morales, 2016). Our results corroborate studies that evaluate the use of perches across different distances (Alencar and Guilherme, 2020) but contrast with other studies that did find differences between perches near or far from forest fragments (Dias et al., 2014; Iguatemy et al., 2020). One study (Zwiener et al., 2014) found differences for plant species richness but not for abundance among different perch distances. This shows that there is no marked pattern for this kind of study; results instead depend on a set of factors, like spatial scale evaluated, type of ecosystem, and the biotic and abiotic variables analyzed. However, the use of artificial perches is effective to overcome the barrier of seed arrival in open habitats compared to studies without structures to attract fauna (Cubiña and Aide, 2001; Piotto et al., 2019).

When we consider the equitability of the species, beyond richness and abundance, the artificial perches at farther distances presented high dominance despite presenting a higher number of seeds and species reflecting a low diversity compared to seed traps in the forest fragment, which shows a greater diversity. The forest fragment presents a more heterogeneous structure, with several ecological niches and with greater diversity of zoochoric birds, which do not use artificial perches but fulfill their role of dispersers within the forest fragment. In tropical forests, up to 90 % of plants depend on birds to be dispersed, thus forming diverse and complex seed dispersal networks characterized by a large number of bird species and interactions with plants (Emer et al., 2020). Artificial perches may act as a filter for the arrival of diaspores because the shape, ramifications, and heights of these structures will influence differently their use by distinct bird species that disperse small seeds (McClanahan and Wolfe, 1987; Holl, 1998). Another factor that limits the efficiency of this technique is seed predation by rodents, insects, and pathogens, decreasing the rates of seedling emergence in open areas, as well as soil compaction and competition with grasses to establish saplings (Holl et al., 2000; Almeida et al., 2016).

On the other hand, the low richness and high dominance by generalist birds, which were observed using artificial perches, are relatively expectable because few species can disperse seeds in both open and forested areas (Pizo and Santos, 2011). In addition, the high representativeness of some plant species in artificial perches may be due to the large number of fruits and seeds that these plants provide, their small fruit size, and visibility, besides harvest season and the time in which they are available for the bird fauna (Jesus and Monteiro-Filho, 2007; Pascotto, 2007). This is the case for Schinus terebinthifolia, Ficus organensis, Pittosporum undulatum, and Solanum americanum, which are abundant and common species in the landscape and at the edge of the fragments and present long fruiting periods, being important sources of food resources in times of scarcity, such as in winter (Jesus and Monteiro-Filho, 2007; Vissoto et al., 2019).

This dominance of dispersing birds and seeds contributed to the formation of distinct plant communities in the study area. The forest seed trap community possibly differed from the others due to the composition of plants and dispersers being different from artificial perches in the open area. Forest specialist birds rarely leave forest fragments, and the few that go into open areas do it for short periods of time (Da Silva et al., 1996). No forest specialist birds were observed using the perches in the open area, probably because the birds prefer natural perches, using the edge trees and depositing their seeds in these places. This fact was corroborated in our study due to differences in richness and abundance and due to the formation of different seed species communities. However, the forest seed community had greater diversity.

Seed rain abundance and richness show the same general pattern, with more distant perches having higher values than perches at 5 m and forest seed traps. This pattern is even more marked for the abundance of seeds because all distances were different to each other, but in this case artificial perches at 50 m had more deposited seeds. The different seed compositions between the perches near and far from the forest fragment depend on several factors, such as the surrounding landscape with the source of propagules and resources for birds species (both in quality and in quantity), the dispersing birds, and the distance from artificial perches perceived by birds (Dias et al., 2014; Zwiener et al., 2014; Alencar and Guilherme, 2020; Iguatemy et al., 2020).

We emphasize the large seed deposition of Pittosporum undulatum, an invasive alien species of Australian origin (Goodland and Healey, 1996; IABIN-I3N, 2019), demonstrating a negative aspect of the use of artificial perches in landscapes with the presence of invasive plant species, since there is no selectivity of seeds deposited by the avifauna. The danger of biological invasion in restoration programs is worrisome, as they prevent ecological succession and tend to be highly aggressive (Tomazi et al., 2010). Pittosporum undulatum is an abundant species in the studied region with a density of 1492 individualsha-1 in nearby forest fragments (Karam and Cardoso, 2010), besides several isolated individuals in the landscape. This expressive bioinvasion is reflected in the reproduction ability of the species with about 20 and 40 seeds per fruit (Goodland and Healey, 1996) and with a long fructification period, mainly in periods of scarcity of other fruit resources (the seeds were recorded across 8 consecutive months, March to October, in the artificial perches). Pitangus sulphuratus, an abundant and generalist bird species found in the region, is an important agent in the dispersion and invasion of Pittosporum undulatum, having a positive effect on the germination of this species after the defecation (Freitas et al., 2020). In addition, the genus Turdus is considered another disperser of Pittosporum undulatum in the Atlantic Forest (Campagnoli et al., 2016). Thus, some birds prefer to feed on exotic species instead of native species (Mokotjomela et al., 2013; Maruyama et al., 2016).

Other studies using artificial perches also found exotic plant species in the seed rain, as did Marcuzzo et al. (2013) in the restoration of urban fragments, with the presence of Melia azedarach and Psidium guajava being attributed to the high presence of these species in urban afforestation. The deposition of Morus nigra and P. guajava was observed under artificial perches in an ecological restoration project in the Atlantic Forest (Almeida et al., 2016). Due to the high abundance of Pittosporum undulatum seeds (the third largest in the present study), care should be taken in the use of artificial perches as an ecological restoration technique in this region. Artificial perches can be replaced by other techniques (i.e., plant mat transplantation, islets of seedlings, direct seeding, and topsoil seed banks of preserved forest fragments) that will not promote the dispersion of invasive species; otherwise the situation requires constant monitoring and control of seeds or seedlings of this species and other invasive alien species that may arise during restoration.

All species observed using artificial perches have generalist habits for both habitat and feed (Sick, 1997; Francisco and Galetti, 2002; Pizo, 2004; Jesus and Monteiro-Filho, 2007; Howe, 2017). This group of birds has great importance in ecological restoration, especially in the dispersion of seeds over long distances, because the birds have a great capacity to explore many resources (González-Castro et al., 2019; Camargo et al., 2020; Campos-Silva and Piratelli, 2021).

Among generalist birds, the Tyrannidae family stands out, representing almost the total length of permanence in the artificial perches, using these structures for rest and foraging and having an overview of the area to search for food. Tyranids are able to eat fruits, poultry, and also insects, foraging in forest edges and more open areas and being able to disperse seeds from forest fragments into open habitats (Guedes et al., 1997; Timm and Timm, 2016). Thus, it is possible that they had contributed with seeds from other forest fragments and isolated trees in the landscape. This behavior was observed, mainly, for T. savanna, T. melancholicus, Pitangus sulphuratus, and E. varius, which captured insects in mid-flight and returned to the perches, defecating under the structures during rest (personal observation). Several studies using artificial perches have reported the presence of generalist species and have attributed the high rate of seed deposition to the family Tyrannidae (Guedes et al., 1997; Bocchese et al., 2008; Athiê and Dias, 2016). According to Camargo et al. (2020), the species Pitangus sulphuratus and T. melancholicus accounted for almost half of all bird activity in the ecological restoration experiment.

In this work we aimed to understand the effectiveness of artificial perches as a restoration technique in assisting the dispersal of seeds by birds into degraded areas. Future studies will further benefit from carrying out bird census inside the forest fragment to compare with the community of birds using artificial perches, as well as assessing the use of perches at different distance by distinct bird species. Furthermore, following germination and seedling recruitment success, as well as their relationship with seed predation in open areas (Bocchese et al., 2008), will allow a better assessment of the restoration success of artificial perches. Finally, using more than one nucleation technique and increasing replication will also allow us to reach more robust results.