INTRODUCTION
Natural and anthropogenic disturbances may cause some ruderal species to dominate large forested areas and change forest structure (Farah et al. Reference FARAH, RODRIGUES, SANTOS, TAMASHIRO, SHEPHERD, SIQUEIRA, BATISTA and MANLY2014, Schnitzer et al. Reference SCHNITZER, DALLING and CARSON2000) and functioning (Muler et al. Reference MULER, ROTHER, BRANCALION, NAVES, RODRIGUES and PIZO2014). Native and exotic bamboo species are disturbance-adapted and profit from the opening of gaps in tropical forests (Griscom & Ashton Reference GRISCOM and ASHTON2003). Bamboo negatively affects the recruitment of tree seedlings into older age categories, mainly through seedling mortality due to a falling branch, trunk or bamboo culm, leading to a self-perpetuating bamboo-dominated area with fewer trees in the canopy, lower density and basal area of regenerating trees, and a thicker litter depth compared with forest patches without bamboo (Griscom & Ashton Reference GRISCOM and ASHTON2003, Guilherme et al. Reference GUILHERME, OLIVEIRA-FILHO, APPOLINÁRIO and BEARZOTI2004, Lima et al. Reference LIMA, ROTHER, MULER, LEPSCH and RODRIGUES2012, Rother et al. Reference ROTHER, RODRIGUES and PIZO2016a).
Under natural conditions, this proliferation of ruderal species occurs in isolated disturbed patches that gradually return to pre-disturbance conditions (Muler et al. Reference MULER, ROTHER, BRANCALION, NAVES, RODRIGUES and PIZO2014). However, the contemporary increase in human-induced disturbances in tropical forests may favour the proliferation of ruderal species and amplify their impacts on community assembly during regeneration (Arroyo-Rodriguez et al. Reference ARROYO-RODRIGUEZ, MELO, MARTINEZ-RAMOS, BONGERS, CHAZDON, MEAVE, NORDEN, SANTOS, LEAL and TABARELLI2015, Asner et al. Reference ASNER, RUDEL, AIDE, DEFRIES and EMERSON2009, Laurance et al. Reference LAURANCE, ANDRADE, MAGRACH, CAMARGO, VALSKO, CAMPBELL, FEARNSIDE, EDWARDS, LOVEJOY and LAURANCE2014). The illegal harvesting of the endemic palm species, Euterpe edulis, in the Brazilian Atlantic forest has led to the creation of an increasing number of forest gaps that have been further colonized and dominated by the bamboo Guadua tagoara in some regions (Rother et al. Reference ROTHER, RODRIGUES and PIZO2009). Euterpe edulis plays an important role in forest ecosystem dynamics as a food source for birds and mammals that are important seed dispersers (Galetti et al. Reference GALETTI, ZIPPARRO and MORELATTO1999, Rother et al. Reference ROTHER, PIZO and JORDANO2016b), thus negatively impacting seed rain taxonomic diversity (Muler et al. Reference MULER, ROTHER, BRANCALION, NAVES, RODRIGUES and PIZO2014). Despite previous studies on the impact of bamboo dominance and palm-heart harvesting on taxonomic diversity, a phylogenetic analysis has not been conducted. In this context, phylogenetic ecology frameworks complement taxonomic studies by linking species in a community with their evolutionary history (Gerhold et al. Reference GERHOLD, CAHILL, WINTER, BARTISH and PRINZING2015, Swenson Reference SWENSON2011, Tan et al. Reference TAN, PU, RYBERG and JIANG2012). The loss of phylogenetic diversity and consequently of functional diversity would reduce forest stability and resilience to environmental changes such as those triggered by anthropogenic disturbances and a changing climate (Cadotte et al. Reference CADOTTE, DINNAGE and TILMAN2012, Craven et al. Reference CRAVEN, FILOTAS, ANGERS and MESSIER2016, Díaz & Cabido Reference DÍAZ and CABIDO2001).
Our study was initially set up to compare the diversity and structure of old-growth forest patches of Atlantic forest dominated or not by bamboos, however, between the first and second census an unexpected harvest of palm-heart took place. As such our study evaluated changes in seed rain and seedling recruitment in a spatial (forest patches dominated or not by bamboos) and temporal (before and after palm-heart harvesting) context. We hypothesized that the seed rain and seedling communities on bamboo patches would show phylogenetic clustering (more close relatives than expected by chance). In addition, we also hypothesized that palm-heart harvesting would accentuate phylogenetic clustering in the seed rain in bamboo patches and lead to clustering in non-bamboo patches. For the seedling community, we assumed that palm-heart harvesting would lead to phylogenetic overdispersion (more distant relatives than expected by chance) in both bamboo-dominated and non-bamboo patches.
METHODS
Study site
This work was carried out in the Carlos Botelho State Park (CBSP) (37793 ha, 24°00′–24°15′S, 47°45′–48°10′W), São Paulo state, south-eastern Brazil. Carlos Botelho State Park is covered by dense tropical forest (Veloso & Góes-Filho Reference VELOSO and GÓES FILHO1982). The local climate is classified as humid subtropical (CFA – Köppen system), an average annual temperature of 22°C and rainfall of 1584 mm, concentrated from December to March (Muler et al. Reference MULER, ROTHER, BRANCALION, NAVES, RODRIGUES and PIZO2014). This protected area of the Atlantic forest has the highest levels of plant species richness of São Paulo state with about 1143 vascular species belonging to 140 families. The vegetation is characterized by the occurrence of the palm Euterpe edulis and almost one third of the area is covered by the native woody bamboo Guadua tagoara, forming a mosaic throughout the area.
Conservation efforts in the last decades have resulted in the establishment of a large network of protected areas in the coastal zone of south-eastern Brazil. This zone contains well-conserved areas of Atlantic forest, such as the Serra do Mar biogeographic region with 32.2% remaining forest cover (Ribeiro et al. Reference RIBEIRO, METZGER, MARTENSEN, PONZONI and HIROTA2009). Although deforestation has declined in this region, forest remnants are still submitted to chronic human-mediated disturbances caused by illegal palm-heart harvesting, which have led to the extirpation of the now threatened palm: Euterpe edulis.
Data collection
We collected data in a permanent plot of 10.24 ha composed by a grid of 256 subplots (20 × 20 m), in which all individuals with dbh > 4.8 cm were sampled. In the permanent plot we selected patches dominated and not dominated by the bamboo G. tagoara.
Seed rain – We collected seed rain data in two periods. In the first period (June 2004–May 2005), we randomly established 80 1-m2 seed traps (40 in each bamboo and non-bamboo patches), made of wood and nylon screen and positioned 10 cm above ground (Rother et al. Reference ROTHER, RODRIGUES and PIZO2009). Traps were established with a minimum separation of 20 m. After the first period, all individuals of the palm Euterpe edulis present in the study area were illegally harvested to collect palm-heart. Therefore, in the second period (January–December 2008) we evaluated the seed rain in bamboo and non-bamboo patches after the harvest of the palm-heart, using 61 0.25-m2 seed traps (30 in bamboo and 31 in non-bamboo patches). These traps were also made of wood and nylon screen and installed 10 cm above ground.
Because of the spatial dynamics of the bamboo patches as a result of the death of culms and growth of new culms, the sampling points of the second period differed in number and partially differed in position from those sampled in the first period. We avoided resampling areas where bamboo had died. Death of G. tagoara occurs after its reproductive phase (Alves Reference ALVES2007). In both periods, seeds were collected monthly and identified at the species level whenever possible.
Seedlings – Following the same framework of the seed rain evaluation, we established 80 1-m2 plots (40 in each forest patch type) adjacent to the seed traps to monitor seedling recruitment in the first period. In the second period, we established 61 0.25-m2 plots in the patches (30 in bamboo and 31 in non-bamboo patches). In both periods, each seedling ≤ 0.30 m height was tagged, identified to species level whenever possible, and monitored monthly according to its survivorship.
Data analysis
We estimated the phylogenetic structure of the observed species in bamboo- and non-bamboo-dominated patches, before and after palm-heart harvest. We employed a local species pool surveyed by Lima et al. (Reference LIMA, DITTRICH, SALINO, BREIER and AGUIAR2011) to generate the phylogenetic tree. We included in the species pool all species surveyed at our study sites that were not listed in Lima et al. (Reference LIMA, DITTRICH, SALINO, BREIER and AGUIAR2011). We removed individuals that were not identified to the species level for the phylogenetic analysis. We compared 1027 seeds of 93 species in bamboo patches versus 1488 seeds of 131 in non-bamboo patches and 787 seedlings of 106 species versus 512 seedlings of 107 species in bamboo and non-bamboo patches, respectively.
We estimated the phylogenetic relationships among the species using the most recent angiosperms supertree, R20120829. To estimate the ages of the interior nodes and evenly space the nodes between them, we employed the Bladj algorithm from Phylocom and evolutionary ages published by Wikstrom et al. (Reference WIKSTROM, SAVOLAINEN and CHASE2001). Before aging the file we checked for internal node inconsistencies as recommended by Gastauer & Meira-Neto (Reference GASTAUER and MEIRA-NETO2013). We estimated the evolutionary distance matrix between all species pairs observed in our patches using Phylomatic, an algorithm implemented in the software Phylocom (Webb et al. Reference WEBB, ACKERLY and KEMBEL2008).
We compared the phylogenetic structure of bamboo-dominated versus non-bamboo patches, before and after palm-heart harvest, using the net relatedness index (NRI) and the nearest taxon index (NTI) (Webb Reference WEBB2000). The first index estimates the mean pair-wise evolutionary distance among all species found in a given sample and compares this distance to the distance observed between the same number of taxa sampled at random from the species pool:
$$\begin{eqnarray}
{\rm{NRI}} &=& - 1 \times (( {\rm{MP}}{{\rm{D}}_{{\rm{observedPD}}}} \nonumber\\
&&-{\textrm{Mean MP}}{{\rm{D}}_{{\rm{nullPD}}}}){\rm{/sd MP}}{{\rm{D}}_{{\rm{nullPD}}}})
\end{eqnarray}$$
The second index estimates the distance between one species and its closest relative in the community:
$$\begin{eqnarray}
{\rm{NTI}} &=& - 1 \times (({\rm{MNT}}{{\rm{D}}_{{\rm{observedPD}}}} \nonumber\\
&&-{\textrm{Mean MNT}}{{\rm{D}}_{{\rm{nullPD}}}}){\rm{/sd MNT}}{{\rm{D}}_{{\rm{nullPD}}}})
\end{eqnarray}$$
These measures assess if the phylogenetic dispersion in a community is greater or less than that expected at random given its species richness, meaning there are more or less closely related species than expected by chance. The values MPD and MNTD were compared to a null estimate generated by randomly sampling 9999 times from the local species pool, keeping observed plot abundance. Abundance-weighted phylogenetic indices are better than occurrence-based indices at capturing assembly processes such as environmental filtering or competition acting on a given community (Freilich & Connolly Reference FREILICH and CONNOLLY2015). We employed the randomization routine called ‘taxa labels’, which shuffles the labels of a phylogenetic distance matrix across all taxa, as a null model to estimate the random communities. This metric is considered as a conservative method for avoiding Type-I errors (Hardy Reference HARDY2008). In both cases, a positive index value indicates a higher presence of close relatives (phylogenetic clustering), whereas a negative value indicates a higher than expected presence of far relatives (phylogenetic overdispersion) in the plant community. To estimate which taxa in the communities were responsible for the observed phylogenetic pattern indicated by the indices we employed the function nodesig, implemented in the Picante package (R version 3.1.1). We conducted all analyses separating seeds and seedlings.
RESULTS
Bamboo patches showed significant seed rain overdispersion (higher than expected presence of distant relatives), whereas non-bamboo patches showed a random phylogenetic structure (Figure 1a). Palm-heart harvesting changed the phylogenetic structure of the seed rain from significantly overdispersed in bamboo patches pre-harvest to random post-harvest as measured by the Net Relatedness Index (NRI, Figure 1a). Palm-heart harvesting did not affect the phylogenetic structure of the seedling community, which was significantly overdispersed pre- and post-harvesting in both bamboo-dominated and non-dominated patches (Figure 1b). The phylogenetic structure at the tips of the phylogeny (NTI index) was random in bamboo- and non-bamboo-dominated patches, for both seeds and seedlings, and that structure did not change post-harvest (Figure 2).
Figure 1. Net relatedness index (NRI) for seeds (a, b) and seedlings (c, d) at bamboo and non-bamboo patches pre-harvest (2004) and post-harvest (2008) of Euterpe edulis palm-heart, Carlos Botelho State Park, Sete Barras, São Paulo state, Brazil. Asterisks denote phylogenetic patterns that significantly differ from random. Data come from 80 seed and seedling collection areas in 2004 and 61 in 2008. The phylogenetic tree was estimated based on the most recent angiosperm tree found online, R20120829. We compared the seeds of 93 species in bamboo patches versus 131 in non-bamboo patches and the seedlings of 106 versus 107 species in bamboo and non-bamboo patches, respectively.
Figure 2. Nearest taxon index (NTI) for seeds (a, b) and seedlings (c, d) at bamboo and non-bamboo patches pre-harvest (2004) and post-harvest (2008) Euterpe edulis palm-heart harvesting, Carlos Botelho State Park, Sete Barras, São Paulo state, Brazil. The phylogenetic patterns observed did not differ from random. Data come from 80 seed and seedling collection areas in 2004 and 61 in 2008. The phylogenetic tree was estimated based on the most recent angiosperm tree found online, R20120829. We compared the seeds of 93 species in bamboo patches versus 131 in non-bamboo patches and the seedlings of 106 versus 107 species in bamboo and non-bamboo patches, respectively.
Seeds of the pioneer species Cecropia glaziovii were more abundant than expected by chance in bamboo patches before palm-heart harvesting but no other taxa were more or less abundant than expected in either area. After the harvest, Cecropia and other pioneer taxa were more abundant than expected in the seed rain and seedlings of the palm, E. edulis, were also more abundant in bamboo-dominated patches (Table 1).
Table 1. Analysis of taxonomic nodes that significantly affected the phylogenetic structure of the seed rain and seedling community at bamboo and non-bamboo patches pre- (2004) and post- (2008) Euterpe edulis palm-heart harvesting, Carlos Botelho State Park, Sete Barras, São Paulo state, Brazil. Data come from 80 seed and seedling collection areas in 2004 (pre-harvest) and 61 in 2008 (post-harvest).
DISCUSSION
The use of a phylogenetic ecology framework for assessing the impacts of disturbances in the regenerating community in an Atlantic forest old-growth area allowed us to distinguish the differential effects of bamboo dominance and palm-heart harvesting in the phylogenetic structure of seeds and seedlings. Contrary to what we expected, neither the presence of bamboo nor palm-heart harvesting significantly affected the phylogenetic patterns of the seed and seedling communities, which remained mostly random for the seed rain, except in the bamboo patches pre-harvest, and overdispersed in the seedling recruits for the Net Relatedness Index (NRI), and random for the seedlings in the Nearest Taxon Index (NTI). We did not observe the expected phylogenetic clustering that is exhibited in some plant communities with strong environmental filtering or after a disturbance (Dinnage Reference DINNAGE2009).
Greater bird species richness visiting bamboo patches (Rother et al. Reference ROTHER, ALVES and PIZO2013) may explain the greater presence of far relatives in the seed rain and thus the significant overdispersed phylogenetic structure in bamboo-dominated patches. Frugivorous bird species visiting bamboo patches are not the same as the ones in non-bamboo patches (Rother et al. Reference ROTHER, ALVES and PIZO2013), a difference that may be responsible for changes in the number of seeds deposited, the taxonomic and the phylogenetic composition of the seed rain. Contrary to our hypothesis, bamboo does not appear to be posing a negative environmental filter to the recruitment of a diverse seedlings assemblage. A strong environmental filter acting on conserved seed traits related to germination has been shown to lead to phylogenetically clustered communities (Vanderlook et al. Reference VANDERLOOK, VERDÚ and HONNAY2012). Although bamboo litter may be harmful to woody species regeneration by establishing a physical barrier for seedling establishment and releasing allopathic substances shown to inhibit growth of native Atlantic forest species (Grombone-Guaratini et al. Reference GROMBONE-GUARATINI, JESSEN, CARDOSO-LOPES and TORRES2009), this negative effect may be counterbalanced by higher habitat heterogeneity created by differential light availability in the understorey (Stevens et al. Reference STEVENS, GAVILANEZ, TELLO and RAY2012). We may observe a stronger negative effect of bamboo on later recruitment stages (Griscom & Ashton Reference GRISCOM and ASHTON2003, Rother et al. Reference ROTHER, RODRIGUES and PIZO2016a). Guadua tagoara has been shown to function as a demographic bottleneck in later stages of Euterpe edulis recruitment, arresting its regeneration cycle (Rother et al. Reference ROTHER, RODRIGUES and PIZO2009). Similar effects may be expected for saplings and young trees. Other research has shown a lower adult tree density in bamboo-dominated patches in tropical forests (Fantini & Guries Reference FANTINI and GURIES2007, Guilherme et al. Reference GUILHERME, OLIVEIRA-FILHO, APPOLINÁRIO and BEARZOTI2004, Oliveira-Filho et al. Reference OLIVEIRA-FILHO, VILELA, GAVILANES and CARVALHO1994).
In addition, the overdispersion observed in the seedling communities, which did not differ between bamboo-dominated and non-dominated patches, could be due to the study sites being immersed in a matrix of old-growth, species-rich forest, thus with a great probability of arrival of species from distant clades. Strong environmental filters, such as the conditions present in abandoned agricultural sites, foster the colonization of close relatives due to the conservatism of traits related to seedling recruitment in disturbed patches (Baeten et al. Reference BAETEN, DAVIES, VERHEYEN, VAN CALSTER and VELLEND2015). On the other hand, overdispersed patterns as observed in the study sites resemble the patterns observed in old-successional forests and in continuous forest stands (Li et al. Reference LI, CADOTTE, MEINERS, HUA, JIANG and SHU2015, Munguía-Rosas et al. Reference MUNGUÍA-ROSAS, JURADO-DZIB, MEZETA-COB, MONTIEL, ROJAS and PECH-CANCHÉ2014). In sites with weak environmental filtering, it is expected that once species arrive at a site, distant relatives can more readily coexist due to greater niche and fitness differences than close relatives (Godoy et al. Reference GODOY, KRAFT and LEVINE2014).
Abundance of the pioneer taxa Cecropia glaziovii after palm-heart harvesting indicates that harvesting leads old-growth forest patches to early successional stages (Muler et al. Reference MULER, ROTHER, BRANCALION, NAVES, RODRIGUES and PIZO2014, Tabanez & Viana Reference TABANEZ and VIANA2000). Cecropia glaziovii was the most abundant species sampled in the seed rain, in both patches, before palms were harvested, but seed number increased after harvesting (Muler et al. Reference MULER, ROTHER, BRANCALION, NAVES, RODRIGUES and PIZO2014). Cecropia species produce a large number of fruits and its seeds are dispersed by a wide variety of dispersers (birds, bats, primates, marsupials), therefore, it does not suffer from disturbances as much as other species do (Fleming & Williams Reference FLEMING and WILLIAMS1990). More seedlings of the palm E. edulis recruited after palm-heart than expected by chance, however, the observed decrease of around 90% of palm seeds after harvesting (Muler et al. Reference MULER, ROTHER, BRANCALION, NAVES, RODRIGUES and PIZO2014) indicates that this increase in seedling recruitment may come from the seed bank but that continued palm harvest will have strong effects on the future recruitment of the palm and the attraction of dispersers of other large-seeded species.
The phylogenetic analysis conducted in the present study complemented previous analyses regarding the effects of bamboo dominance and E. edulis harvesting on species taxonomic richness (Muler et al. Reference MULER, ROTHER, BRANCALION, NAVES, RODRIGUES and PIZO2014), showing that bamboo dominance and palm-heart harvesting not only affect the taxonomic diversity of the forest but also the phylogenetic structure, mainly of its seed rain, with implications for the stability and resilience of the forest in the future. Closely related taxa generally have a higher likelihood of sharing functional traits than further related ones (Manel et al. Reference MANEL, COUVREUR, MUNOZ, COUTERON, HARDY and SONKÉ2014), therefore the maintenance of overdispersed patterns in the seedling community indicate that despite bamboo dominance and palm-heart harvesting, functional diversity is being preserved at least at the early regenerating stages and in the time frame of the study. Despite challenges in phylogenetic trees resolution, this approach can be included in analyses of community composition as it contributes to a more comprehensive picture that can lead to further studies of processes of species assembly.
ACKNOWLEDGEMENTS
We gratefully acknowledge the staff of the Carlos Botelho State Park for permission to conduct our study in the park. We are especially grateful to Eric Teixeira, Paulo Cesar Teixeira, Henrique Teixeira and Natália Guerin for their help in the fieldwork. We really appreciate the help of Marcelo Pinho and Renato Augusto Ferreira de Lima in the seedlings identification. This study received financial support from the São Paulo Research Foundation (FAPESP) (#1999/09635-0, #2013/50718-5). Individual research grants were provided by the Brazilian National Research Council (CNPq) to DCR (#143107/2006-2), MAP (#303559/2008-0) and PHSB (#304817/2015-5), and by FAPESP to DCR (#2003/11073-7) and AEM (#2007/06745-7).
