Bird-plant interaction networks in native forests and eucalyptus plantations within a protected area

1 Universidade Federal de Ouro Preto (UFOP), Instituto de Ciências Exatas e Biológicas (ICEB), Departamento de Biodiversidade, Evolução e Meio Ambiente (DEBIO), Laboratório de Zoologia dos Vertebrados. Ouro Preto, MG, Brasil. 2 ORCID: http://orcid.org/0000-0001-5883-097X. E-mail: cristian_v_b_y@hotmail.com 3 ORCID: http://orcid.org/0000-0002-0992-5515. E-mail: marcelabiol@hotmail.com 4 ORCID: http://orcid.org/0000-0003-0256-9017. E-mail: cristianoroxette@yahoo.com (corresponding author)


INTRODUCTION
Ecosystems with extensive plantations of non-native trees, such as the eucalyptus monocultures, can extinguish local species by replacing native vegetation and simplifying the forest heterogeneity, alter the regional water resources, and diminish soil fertility (Marsden et al., 2001;Borsboom et al., 2002;Kanowski et al., 2005;Carnus et al., 2006;Valduga et al., 2016;Bayle, 2019). On the other hand, eucalyptus plantations can be used as fuelwood, replacing the use of indigenous species, and conserving native forests and wildlife, protecting soil from erosion, and help in habitat restoration by promoting the recruitment, establishment and succession of native woody species (Hobbs et al., 2003;Barlow et al., 2007;Brockerhoff et al., 2008;Bremer & Farley, 2010;Brockerhoff et al., 2013;Bayle, 2019). All these negative and positive effects have been much debated on whether they prevent forest regeneration (Altieri, 2009;Edwards et al., 2014;Benton et al., 2021). However, little is known about the consequences of eucalyptus plantations for seed dispersal.
Seed dispersal is considered the most important ecological process for the natural regeneration of vegetation (Bascompte & Jordano, 2007), occurring in most cases through frugivory (fruit consumption). Frugivory allow mutualistic animal-plant interactions in which more than two species are usually involved, that may form a complex network of interactions, capable of maintaining the resilience, and even the regeneration, of the ecosystem through the spread of propagules (Bascompte & Jordano, 2007). Mutualistic animal-plant interactions are influenced by environmental disturbances, such as habitat fragmentation and introduction of exotic species, by the exclusion or introduction of key species in the system, preventing pre-existing relationships in the interaction network (Markl et al., 2012). To understand how a species becomes a key player in a network of mutualistic interactions, we need to assess the importance of the species in a complex context and through quantitative operational metrics (Mello et al., 2015).
In South America, specifically in Brazil, eucalyptus is the most widespread exotic monoculture, which places the country as the second in the world with the largest area of native forest replaced by plantations of this tree (Ceccon, 2001). This fact led to changes in many ecosystems, such as the Atlantic Forest, considered one of the most fragmented domains in the Neotropics (Ceccon, 2001;FAO, 2016). This is the domain with the largest number of endemic species in Brazil (it is considered a biodiversity hotspot; Myers et al., 2000;Rezende et al., 2018). Because of this problem, some areas of the Atlantic Forest were converted into protected areas, such as the Itacolomi State Park (PEIT), located in southeastern Brazil, which include both native vegetation and a 30-year abandoned eucalyptus plantation. The PEIT is considered an important historical and geological natural park in the region (Tafuri, 2008;Messias et al., 2017). Studies, such as those of Andrade (1998), Governo do Estado de Minas Gerais et al. (2007) and Pereira et al. (2015), highlighted the high bird diversity of the PEIT, indicating the probability of existence of a complex birdplant interaction networks.
Birds play a fundamental role in the establishment and evolution of native vegetation, with many species that consume and potentially disperse seeds, hence contributing to the forest restoration in degraded areas (Fleming & Kress, 2011). In addition, birds have the largest number of frugivorous species in the Neotropics, with some families being dependent or highly dependent on eating fruits (Fadini & De Marco, 2004;Hernández-Ladrón de Guevara et al., 2012). For these reasons, it would be important to evaluate the contribution of birds to seed dispersal inside the PEIT, evaluating if the replacement of native vegetation by eucalyptus plantations interfere with birds-plant interaction network. This study aimed to compare the relationships between frugivorous birds and zoochoric plants in the forests of the PEIT, by comparing bird-plant interaction networks in an area of native forest with an area of eucalyptus plantation. We hypothesized that the richness of seed dispersing birds would be lower in the eucalyptus plantation than in native forest areas due to the replacement of native vegetation by the monoculture. Also, we hypothesized that the interaction network would be more connected in the native forest than in the eucalyptus plantation, due to the exclusion of frugivorous birds and their mutualistic native plant species by the eucalyptus trees (Valduga et al., 2016).

Study Area
The study was carried out at the Itacolomi State Park (PEIT), a protected area located in the southeast of the state of Minas Gerais, between the municipalities of Ouro Preto and Mariana (43°32′30″W,20°22′30″S). The PEIT has an area of 7,543 ha, with altitudes varying between 700 to 1,772 m.a.s.l. The region's climate, according to the Koeppen classification, is dry-winter subtropical highland climate (CWb), as temperatures can vary from 4 to 33℃, with an annual average of 20℃. Precipitation is concentrated between the months of October and March, with a notorious wet season during summer and another dry season in winter (Tafuri, 2008). In the PEIT, rupestrian grasslands occupy 51% of its territory (3,846.93 ha), 40% is covered by montane Atlantic Forest (3,017.2 ha), and 9% is occupied by a 30-years abandoned eucalyptus plantation (two blocks totaling 678.87 ha), placed in an area previously occupied by the native Atlantic Forest, as long as abandoned tea plantations and human constructions (Messias et al., 2017). In this protected area, have being recorded 1,623 plant species, 251 birds, 43 mammal species, 37 amphibians and 22 reptiles, showing its high biodiversity (Governo do Estado de Minas Gerais et al., 2007;Messias et al., 2017).
Eucalyptus grandis W. Hill ex Maiden formed the abandoned eucalyptus plantations, with trees reaching heights up to 45 m. A developed understory, formed by native plant species with up to 15 m height, could be observed in the abandoned eucalyptus plantation. Common native plant species such as Schinus terebinthifolius Raddi and Drimys brasiliensis Miers are found in native forest. A dense understory, with plants measuring from 15 to 18 m is present in the native vegetation area.

Fecal Samples
Sets of 10 mist nets (35 mm mesh, 12 m length x and 2.5 m width) were installed in the three sampling units of both the native forest and the eucalyptus plantation. The nets remained open for sampling between 06:00 h until 12:00 h, totaling 180 hours in each habitat, for 10-month samplings. Each net was inspected at 30 min intervals during the sampling period. Once a bird was captured, the species was identified using a field guide (Sigrist, 2014). Then, each captured bird was placed in a cloth bag with filter paper for 15 or 30 minutes until it defecated, and the feces were preserved in 70% alcohol for later laboratory analysis (León, 2010;Hernández-Ladrón de Guevara et al., 2012).
All fecal samples were placed in petri dishes and diluted with 70% alcohol to separate intact seeds using a stereoscopic microscope in the laboratory. The seed identification in feces was carried out using seed identification guides (Kuhlmann, 2012;Frigieri et al., 2016) and by comparison with collections of fruits and seeds of exsiccates present at the Herbário Professor Jose Badini of the Federal University of Ouro Preto (OUPR). Interaction data were collected from each fecal sample, and an interaction record was considered valid whenever a fecal sample of a bird species contained intact seeds (seeds not broken during gut passage) from a given plant species. The number of seeds within fecal samples was not recorded due to the difficulties that would lead to an underestimation of some species with very tiny and numerous seeds, such as those of the Melastomataceae family.

Visual Recordings
Focal observations of the plants whose fruits were consumed by frugivorous birds were conducted in the six transects of each sampling areas, from 06:00 h to 11:00 h and from 15:00 h to 18:00 h (Pizo & Galetti, 2010), for 10 months, totaling 240 hours of observation in each type of habitat. The number of bird-plant interactions per species was recorded, considering an interaction when a bird species consumed fruits of a plant species, in a way that every plant species may have numerous/ many interactions when fruits were consumed by several bird species, and a bird species could have multiple interactions when consuming fruits of several plant species (Silva et al., 2002;Pizo & Galetti, 2010;Cullen et al., 2004). In each transect, the plant species with fleshy fruits were sampled, and the fruits were placed in alcohol 70% and taken to the laboratory for later identification (Caziani, 1996;Amico & Aizen, 2005).

Data Analysis
An interaction network is the graphic representation of a complex system of multiple potentially connected elements, in which the connection may be a mutualistic relationship, and the elements are the plants and birds (Bascompte & Jordano, 2014). To identify the importance of species and groups of species in the interaction network in the sampled habitats, the following network metrics were calculated following Mello et al. (2015), Costa et al. (2016) and Mello et al. (2016). The degree of connection or connectivity, which is used to compare the pair of species that interact the most within the network, thus showing the strongest bird-plant relationships in the community. The degree of centrality, which points out the species with greater number of interactions. The centrality by proximity or closeness centrality, that estimates which species are more likely to interact with others within the network, in other words, showing the species with the greatest indirect relationship with other species in the network. The centrality by intermediate or betweenness centrality, which indicates the species that are mediating the interaction between two other species within the community. Finally, the Louvain modularity (Q) was used to assess if the relationships are equally distributed between birds and plants or if they are separated into subgroups (modules) within the community.
All metrics were calculated using the software Pajek 4.09 (Batagelj & Mrvar, 1998) and non-parametrical Mann-Whitney tests were used to compare the resulting metrics between the two types of sampled habitats using the software R 3.5.0. (R Development Team, 2018). Complementarily, the Sorensen Index for similarity between communities, the Index of dispersion importance, and the Index of importance for plants were calculated for each habitat. The Sorensen Index for similarity between communities (IS) indicates the similarity in the species composition of the compared communities was calculated using the formula IS = 2c / a + b -c, where "c" represents the number of species recorded in both environments, "a" the number of species in the first environment (native forest), and "b" the number of species in the second environment (eucalyptus plantation) (Rabinovich, 1981). The Dispersion Importance Index (DII) indicates the importance of each bird species as a seed-disperser in the studied habitats and it was calculated using the formula DII = (S * B) / 1000), where S represents the relative abundance of each captured species, and B represents the percentage of fecal samples with seeds of each species; DII values ranges from 0 to 10, where 0 represents no seeds found on the feces and 10 represents a unique bird species dispersing all seeds (Galindo et al., 2000). Finally, the Index of importance for plants (IVIP), which represents the importance of a bird species for seed dispersal of a given plant, was calculated using the formula IVIP = Σ (r / R), where r represents the number of times a bird species presented seeds of a plant species in its feces and R represents the total number of plant species consumed by a bird species (Galindo et al., 2000;León, 2010). Then, after testing data normality using the Lilliefors test, we used paired t-tests to evaluate if the total richness of seed dispersing birds were higher in the native forests than in the eucalyptus forests.

RESULTS
A total of 419 individuals from 21 bird species were captured with mist nets, from which 318 samples of feces were obtained. The intact seeds found in the feces belong to 17 species of plants (one seed type was identified to genus, two to family, and three remained as morphospecies). In the native forest area, 42 species of birds were recorded, with 17 captured by the mist nets and 35 regis-tered through focal observations. On the other hand, 36 species of birds were recorded in the eucalyptus plantation, with 13 captured by the mist nets and 30 registered by the focal method (Table 1). In both native forest and eucalyptus plantation, the most common plant genus found in bird droppings was Miconia (Melastomataceae), followed by Leandra (Melastomataceae) and Myrsine (Primulaceae), the first two are undergrowth shrubs in the studied area, and the third being arboreal. No differ-ences in seed dispersing by birds were found between native forests and eucalyptus plantations (t = 1.20, p = 0.25).
Within the native forest fragments, 50 interactions between 17 bird species and 20 plant species were recorded through fecal analysis (Fig. 1A); while, in the eucalyptus plantation fragments 41 interactions, between 13 species of birds and 19 species of plants were recorded (Fig. 1B). On the other hand, with focal observation sam- pling, 92 interactions were recorded in native fragments among 36 bird species and 14 plants ( Fig. 2A); while in eucalyptus fragments, 70 interactions among 32 birds and 11 plants were registered (Fig. 2B). Statistically, there was no differences in the number of interactions between the two types of habitats for the fecal analysis (W = 4.0, P-value = 0.99) or for the focal observations (W = 3.5, P-value = 0.83) (Fig. 3). Almost all bird species were recorded in both types of habitats, with some exceptions such as Phibalura flavirostris Vieillot, 1816 (Cotingidae) and Dacnis cayana Sundevall, 1836 (Thraupidae), and the plant species Vismia parviflora Cham. & Schltdl (Hypericaceae) which were found only in the native forest; and Phaeomyias murina (Spix, 1825, Tyrannidae) and Miconia valtheri Naudin (Melastomataceae) that were recorded only in the eucalyptus plantation.
Using the data obtained via fecal analysis, the highest degree of connection values in the native forest were  In relation to the closeness centrality in both habitats, the bird species C. caudata and the plant M. theaezans were the ones that showed the highest proximity. The betweenness centrality metric showed that bird species N. chrysolophum and C. caudata presented the highest values in both habitats. The Dispersion Importance Index (DII) indicated N. chrysolophum and Trichothraupis melanops (Vieillot, 1818; Thraupidae) as the most important seed disperser bird species in the studied native forest; while in the eucalyptus plantation, C. caudata and Tachyphonus coronatus (Vieilott, 1822; Thraupidae) were the most important seed dispersers. Conversely The network connectivity in the native forest obtained through the focal observations indicated the interaction between the tree Schinus terebinthifolius Raddi (Anacardiaceae) with the bird Elaenia obscura (d'Orbigny & Lafresnaye, 1837; Tyrannidae) as the one with the largest number of records. In the eucalyptus plantation, the greatest connectivity was observed for M. valtheri and Tangara cyanoventris (Vieillot, 1819; Thraupidae). The metrics of degree, closeness and betweenness centrality, in both habitats indicates that plants Myrsine umbellate Mart. (Primulaceae), S. terebinthifolius, and the bird E. ob-scura as the species with the highest number of links and importance for the network (Supplementary material 2). None of the network metrics showed significant statistical differences between native forest and eucalyptus plantations (Table 2). Also, according to the Sorensen Similarity Index, the frugivorous bird communities of both habitats presented a similarity of 0.71, while the zoochoric plant communities of both habitats presented a similarity of 0.72, indicating a high similarity (up to 70%) in the composition of species which play a role in seed dispersion in both habitats.
Finally, results for the modularity generated six groups of birds and plants more related to each other in both habitats, and these results were highly identical, with the values found from the mist nets (native forest: Q = 0.49, eucalyptus plantation: Q = 0.52) and focal observations (native forest: Q = 0.42, eucalyptus plantation: Q = 0.39), with an average of Q = 0.455 in the two habitats.

DISCUSSION
The richness of seed dispersing birds in the eucalyptus plantation and in native forest were similar. The interaction networks were remarkably similar in both habitats, partially corroborating the initial hypothesis. These similarities between the native forest and the eucalyptus plantation found in our study could be caused by environmental characteristics of the eucalyptus plan-  tation; the area have not been disturbed by human extractive actions for 30 years, it remains close to the native forest fragments, and exhibit a well-developed understory. Similar results were found by Stallings (1991) and Machado & Lamas (1996), who observed that almost half of the species of mammals and birds found in the fragments of native forest were also found in eucalyptus plantations with well-developed understory; in contrast to the species found in eucalyptus plantations without understory. The understory can provide important habitat resources for wildlife species, thus improving the conservation value of eucalyptus plantations: Additionally, decreasing the use of herbicides during eucalyptus growth would allow the development of the understory and the maintenance of a greater biodiversity (Brockerhoff et al., 2008;Pereira et al., 2015). However, monocultures of introduced species, including those with a well-developed understory, certainly do not support all the diversity that would be found in a native forest (Brockerhoff et al., 2013). Therefore, they must be avoided, always giving preference to the maintenance of native forests.
The results of the network metrics in the present study were moderately higher, at least for the degree of centrality and centrality by intermediate, when compared to those found by Mello et al. (2015), who evaluated the interaction networks of bird frugivory across the Neotropics, indicating that the patterns vary enormously between the interaction networks depending on the location. On the other hand, according to Vidal et al. (2014), the number of interactions between species in mutualistic networks is strongly associated with each species contribution to modularity, which would explain why the result of the Louvain modularity (Q) in this study was notoriously similar between the two types of habitats. Silva et al. (2002) also found a similarity of 70% of frugivorous birds between a fragment of non-protected Atlantic Forest and a conserved one, inside the Intervales state park in Brazil, which could indicate the efficiency of frugivory in impacted forests within reserves.
The study of Vidal et al. (2014)  These studies showed the influences of anthropogenic impacts on the interaction networks of frugivore birds and plants, with the human actions (mainly deforestation) decreasing the richness of birds, plants, and their interactions, with important changes in the composition of the community and its succession processes.
According to Lopes et al. (2005), in the Atlantic Forest domain, it is common to find intact seeds, mainly from the Melatomataceae family, in the stomach contents of the frugivorous birds of the families Thraupidae, Turdidae, and Pipridae. León (2010) also found that several species of the Thraupidae, Pipridae, and Tyrannidae families presented higher IVID values in the forests of Mexico, being the main seed dispersers. Among these families, Thraupidae is considered by Schleuning et al. (2014) as the most important group of seed dispersers in frugivory networks. In the present study, birds of the families Thraupidae, Pipridae, and Tyrannidae were the main seed dispersers in both native forest and eucalyptus plantation. Fadini & De Marco (2004), Lopes et al. (2005) and Athie & Dias (2012) pointed out some of the same species found on the present study, such as C. caudata, Tangara sayaca (Linnaeus, 1766), T. coronatus and Turdus rufiventris Vieillot, 1818, as being fundamental to the interaction network of potential seed dispersers in the Atlantic Forest of the southeastern Brazil. These species are considered tolerant to the presence of exotic monocultural species, for having a generalist diet and a wide distribution, being equally registered in preserved native vegetation and in eucalyptus plantations along the Atlantic Forest domain (Machado & Lamas, 1996;Manhaes et al., 2010). In other domains, such as in the Cerrado (savannah-like vegetation), bird T. sayaca also showed the highest interaction values (Purificação et al., 2014), while in Restinga (sandbank vegetation) in southern Brazil T. rufiventris was one of the most important bird species (Scherer et al., 2007), showing that some of these species are important seed dispersers also in other vegetation types.
According to Athie & Dias (2012), S. terebinthifolius is one of the most important plant species in a frugivory network in the Atlantic Forest and have also been recommended for restoration processes due to its rapid growth and its importance as a resource for the avifauna (Avila et al., 2010) and entomofauna (Somavilla et al., 2010;Sühs et al., 2009). Manhaes et al. (2010) and Vidal et al. (2014) found many species of the genus Miconia, including M. theaezans, as important for the diet of birds and possibly for mutualistic networks, corroborating Snow (1981), who indicates the Melastomataceae family as the one with the largest number of species that offer food for seed dispersing frugivores in Neotropical forests, results in accordance with the findings of the present study. León (2010) found that several species of the genus Ficus (Moraceae) have high index of importance of plants values, being important for seed dispersion. However, plants of this genus were uncommon in the PEIT areas (Messias et al., 2017), not being recorded in the interactions observed in the present study.
Most of the species of birds and plants were recorded in both habitats, and they are common in the Atlantic Forest of southeastern Brazil, being not considered sensitive to anthropogenic impacts and do not presenting habitat restrictions in the region (Snow, 2004;Messias et al., 2017). The exceptions were the Swallow-Tailed Cotinga (P. flavirostris), registered only in the native forest and considered near threatened by Birdlife International CONCLUSION Species richness and the complexity of interactions between frugivorous birds and zoochoric plants were similar in both, the abandoned eucalyptus plantation and native forests in the present study. This indicates a possibly same potential for seed dispersal in both studied forests, also suggesting the same potential for regeneration. Therefore, in the PEIT, the presence of the abandoned eucalyptus plantation is not being considered an obstacle to the regeneration of the Atlantic Forest, thanks to the fact that the natural succession process is being allowed in the understory. It is important to mention that adding more sampling points in different areas of the two habitats in future studies would increase the representativeness of the mutualistic interaction networks and the power of the analysis, resulting in more robust and refined information.

ACKNOWLEDGMENTS
Authors would like to thank UFOP and CAPES for the logistic support during the development of this study the licenses granted by IEF (093-2018), SISBIO (63431-2) and CEUA (4425170918), and to the Herbário José Badini (OUPR) for the identification of the plant samples. CDVB thanks the scholarship received from OEA/GCUB. The authors would also like to thank all the PEIT staff who helped directly and indirectly with this study.