Diversity and Structure of Natural Regeneration in Swamp Forests in Southeastern Brazil

Many factors can influence the structures and distributions of plant communities. Plant diversity of swamp forests reflect responses to water stress conditions. We evaluated the floristic diversity, structure, and composition of the regenerating shrub-tree species of six swamp forests in veredas ecosystems (northern Minas Gerais, Brazil), and the floristic similarities between them. All individuals ≥ 1 cm in diameter at ground level and < 3 cm in diameter at breast height were sampled in one hundred 25 m2 plots in each area. A total of 5442 individuals were recorded, distributed among 134 species, 85 genera, and 47 botanical families, with a diversity (Shannon) of 3.38 and equability (Pielou) of 0.69. Beta diversity was high, while similarities between the areas were low, with only two species common to all six sites. The forests showed different compositions, diversities, and natural regeneration structures, reflecting their ecotone nature and past anthropic impacts.


INTRODUCTION
The Brazilian Cerrado (neotropical savanna) constitutes a mosaic of several phytophysiognomies . Veredas (palm swamps) ecosystems are complex systems associated with Cerrado watercourses (Bijos et al., 2017). They are characterized by the presence of the palm tree Mauritia flexuosa L.f. (Arecaceae), "buriti", which grows in water-saturated sites and are of great importance to biodiversity (Endress et al., 2017). Depending on the ages of those systems, swamp forests will develop in their lower sections and demonstrate successional stage changes (Boaventura, 2006), although few studies have yet been undertaken in areas of swamp forests to examine that phenomenon (Nunes et al., 2015) while focusing on their regenerative strata (Ávila et al., 2016).
Natural regeneration refers to forest growth from the initial stages of establishment and development (Pastório et al., 2018), and describes gradual processes that can restore the structure, function, and composition of an ecosystem to its pre-disturbed state (Chazdon & Guarigata, 2016). That regeneration is shaped by natural socio-ecological processes, but can be sustained, improved, or impaired over time through human interventions (Chazdon et al., 2014). Studies of the regenerative strata allow insights into their potential for regeneration and for the success of environmental restoration efforts (Menino et al., 2012); such studies are also fundamental to establishing economic strategies for forest management and conservation (Santos et al., 2017).
The present study sought to examine the floristic compositions of six swamp forests in northern Minas Gerais State, analyze their natural regeneration and vegetation similarities, as well as address the following questions: (i) how does the diversity and composition of those swamp forests vary?; (ii) is there high floristic similarity between the study sites?; and, (ii) is local diversity (α diversity) lower than the diversity between areas (β diversity)?

Study area
The present study was carried out in the Pandeiros River Environmental Protection Area (EPA) in northern Minas Gerais State, in southeastern Brazil. The EPA covers an area of 393,060 ha and includes the entire Pandeiros River basin, within the municipalities of Januária, Bonito de Minas, and Cônego Marinho (Nunes et al., 2015). The regional climate is tropical humid and dry (type Aw according to the Köppen classification system), with well-defined dry winters and rainy summers (Alvares et al., 2014). Our sampling focused on regenerating arboreal and shrub vegetation in six swamp forest: Água Doce (AD), Almescla(AL), Buriti Grosso (BG), Capivara (CA), Pindaibal (PI), and São Francisco (SF) ( Fig. 1; Tab.1).
The vegetation in the study area is represented by Cerrado and Caatinga transition phytophysiognomies, culminating in a mosaic of riparian and dry forests, savanna, and palm swamps (Menino et al., 2012). The swamp forests associated with the palm swamps, the target of this study, occur mainly in the upper-mid Pandeiros River basin (Nunes et al., 2009;.

Structural survey of the regenerating community
We systematically established one hundred 5×5 m plots (totaling 0.25 ha, distance 15 m between plots, along the water body) where the regenerating arboreal/shrub individuals having ≥ 1 cm DGL (diameter at ground level) and < 3 cm DBH (diameter at breast height -measured 1.30 m above ground level) were sampled. We measured the DGL and total heights of regenerating individuals using a digital caliper and graduated meter respectively, except for regenerating M. flexuosa palm seedlings, which did not produce visible stipes capable of measurement. The data were collected during dry season.
We collected a sample of each inventoried plant for identification and deposited prepared specimens in the Montes Claros Herbarium (MCMG), at the Universidade Estadual de Montes Claros. Identifications of the collected botanical material were performed by consulting specialists and the specialized literature. The classification system used for botanical families was the Angiosperm Phylogeny Group IV (APG IV, 2016).

Data analysis
We calculated the phytosociological parameters: absolute and relative density, dominance and frequency, and importance value (Mueller-Dombois & Ellenberg, 1974). The Shannon diversity index (H') and Pielou equability (J') were calculated on a natural logarithmic basis (Brower et al., 1997;Magurran, 2011). The Hutcheson t-test was performed to compare diversities between areas using the H' diversity index and PAST program software (Hammer et al., 2001).
We determined the floristic similarities between the sampled areas by comparing their species in a presence/ absence matrix using the Single Linkage method and Euclidean metric distances with Statistica 10 software (Statsoft, 2011). Additive diversity partitioning was examined in 100 permanent plots established in the six areas at different scales. The total species diversity (γ) was divided into additive components that represented diversity within the community (α, richness of each plot) and diversity between communities (richness between plots, β1, and between areas, β2). We used the Vegan package for the partitioning of α and β diversities in R (Oksanen et al., 2015;R Development Core Team, 2018). We used a null model with 1000 randomizations to evaluate the statistical significance of the observed values in relation to expected values.

RESULTS
A total of 5442 regenerating individuals belonging to 134 species, 85 genera, and 47 families were encountered in the six sampled areas (supplementary material). The most important families were Fabaceae, with 21 species, followed by Myrtaceae (11 species), Melastomataceae (10), Clusiaceae (7), and Rubiaceae (5). The genera with the highest number of individuals (57.49% of the total) were: Xylopia (827 individuals), Miconia (665), Calyptranthes (520), Vernonanthura (454), Zygia (347), and Trembleya (318). The total diversity (H ') was 3.38 and the equability (J') 0.69. The Hutcheson t-test demonstrated significant differences in diversity between areas, except for AL and BG (t = 0.842; df = 1924.5; p > 0.05) and BG and CP (t = 23.786; df = 1479.2; p > 0.01) (Tab. 2).   (54) Only two species were common to all six areas: Tapirira guianensis Aubl. and Miconia sp., while 10 species were exclusive to AD, 11 to AL, 22 to SF, 11 to BG, seven to CP, and 11 to PI (a total of 72 exclusive species). Floristic similarity showed the formation of a group composed of the BG, CP and PI forests, and another group formed by areas AD and AL (Fig. 2). Area SF was separated from the others. Additive diversity partitioning showed that local species (α) are responsible for 4.29% of the total richness. The exchange of species between plots in each path (β1) represented 36.43%, and the exchange of species between paths (β2) represented 49.28% of the total species richness (γ diversity). The diversity of those components was different from that expected by the null model (p = 0.001; Fig. 3).

DISCUSSION
The swamp forests studied here showed varied structures and compositions, with low similarity between them. Diversity is determined by the numbers of species and their abundances within a community (alpha diversity) and by differences in species compositions and their abundances between areas within the same community (beta diversity) (Magurran, 2011). Beta diversity was thus the component that most explained the total diversity of the swamp forests studied and indicated that each area had a particular flora. The low numbers of species common to those different areas also reinforced that result, because more than half of the species sampled were unique to a single site. Different environmental characteristics, conservation conditions, and land use in each forest can determine those variations, as the areas studied had experienced different anthropic impacts (Neves, 2011) and, consequently, were responding to different degrees of disturbance. The forests in this study experienced external impacts from local inhabitants who use the palm swamps as sources of water, food, and wood, as well as for agriculture purposes (personal observations).
The reduced numbers of species common to all of the forests studied, especially in the same geographic area and watershed, suggest great floristic heterogeneity (Menino et al., 2012). Differences in topography, flooding intensity, and soil conditions at any site will affect the spatial distributions of species and promote phytosociological variations (Teixeira & Assis, 2011).
Plant diversity in each of the six study areas (H' varied from 1.82 to 3.18) was relatively low in relation to the high diversity commonly observed in tropical forests, as swamp forest formations are flooded for most of the year and show low diversity because of those anoxic conditions (Kurtz et al., 2013). Additionally, flooding durations and depths will influence swamp forest compositions (Parolin et al., 2010), as water stress selects for tolerant species that can form large populations, with low resulting diversity (Kurtz et al, 2015).
The Brazilian Cerrado includes forest, savanna, and grassland formations, so that both forest and savanna communities experience the same climate regime (Ribeiro & Walter, 2008). Soil moisture and fertility levels, as well as geographic proximity, are also important factors contributing to species distributions in those environments (Silva Júnior et al., 2001). Swamp forests, as well as palm swamps, are humid formations within the Cerrado biome, with different surrounding plant communities that will influence their species compositions (Teixeira & Assis, 2011). The existence of differing Cerrado vegetation types results in numerous species co-occurring throughout the biome, increasing the diversity of any given region. Menino et al. (2012), for example, reported high species diversity between naturally regenerating riparian vegetations in the Pandeiros River study area that was influenced by the occurrence of diverse neighboring vegetation types. In addition to ecological conditions, geographical proximity can also be determinant for floristic composition (Kurtz et al., 2013). The conditions necessary for the establishment of different species in humid forests support the hypothesis that the surrounding vegetation contributes to their varied compositions (Teixeira & Assis, 2011), resulting in a vegetation mosaic. Although some studies have indicated a group of species characteristic of swamp forests, the contributions of the surrounding vegetation were determinant to the diverse of each study area (Kurtz et al., 2014).
The species sampled were organized into two groups: those found more frequently in poorly drained and flooded sites, such as X. emarginata, Z. latifolia and R. grandis (Maas et al., 2015;Garcia et al., 2015;Secco et al., 2015); and those typical of cerrado and forest edges, such as Trembleya sp., T. sellowiana, and B. pachyphylla (Mendonça et al., 2008). Flood regime durations and water depths will influence plant ecophysiology, and will require special adaptations to the periodicity of local anoxic conditions (Kurtz et al., 2013). Thus, species tolerant of waterlogging will define the plant typology, and will be both abundant and frequent. In this study, species such as X. emarginata, C. brasiliensis, Z. latifolia, Miconia sp., and Trembleya sp. were apparently tolerant of soil water saturation, as they were generally very abundant in the regenerative strata of the forests studied. There are groups of species in swamp forests that demonstrate dominance and high densities in saturated soils, composing a group of flooding specialists (Kurtz et al., 2014). Those species significantly contributed to increasing richness in well-drained soils, but less so in terms of their phytosociological parameters. Teixeira & Assis (2011) indicated that swamp forests generally show 6 -7 Nunes,YRF,Ávila,MA,Mota,NM,Souza,SR,Santos,R 6 low plant diversity and are populated by species with high local densities, and identified C. brasiliense and T. guianensis as indicator species.
The high densities of species typical of humid areas in the swamp forests studied here contributed more to the similarities between the six areas that did their floristics (the presence of species represented by just a few or even a single individual) (Kurtz et al, 2015). Our study areas showed expressive numbers of exclusive species, with SF having the largest number (22 species). The SF area is flooded in the center (where the water table surfaces), but dryer along the edges, with the absence of any grassy zone. It therefore shows no typical separation of the swamp forest and cerrado vegetations, with no middle zone (composed mainly of grasses) of palm swamp (Araújo et al., 2002) -which probably allowed the establishment of numerous unique species with varied ecological requirements. The proximity of surrounding vegetation formations, characterized here as the forest-savanna ecotone, explains the co-occurrence of species on soils with different degrees of drainage (Kurtz et al., 2013;. The varied environmental conditions in swamp forests, particularly in relation to their flooding regimes and topographic variations, allow the establishment of species with different ecological requirements, including generalist species from neighboring areas (Teixeira & Assis, 2011;Kurtz et al., 2013).
The species X. emarginata (in AD and AL), B. salicifolius and C. brasiliense (in AD); Z. latifolia and Miconia sp. (in AL), and P. gaudichaudianum, M. umbellata, and L. elegans (in SF), were the most important taxa in terms of vegetation structure regeneration in the present study. Those species occur in humid environments, especially in riparian, gallery, and flooded forests, as well as palm swamp forests or wetlands (Ribeiro & Walter, 2008;Teixeira & Assis, 2011). The importance of the species B. pachyphylla, V. phosphorica, T. sellowiana, and Landerbergia cujabensis in areas BG, CP and PI, demonstrated the influence of the adjacent vegetations, different soil conditions, and anthropic disturbances in those areas.
In general, the connections between swamp forest in palm swamps and adjacent Cerrado formations favored floristic diversity because they allowed colonization by species with different ecological requirements, thus increasing diversity. Anthropic interferences in the areas (being most intense in areas CP, BG, and PI), however, also appeared to be a determining factor of diversity.
The present work indicates that the regeneration potentials of the six areas studied are different (as seen in their respective diversities and compositions), so that conservation, management, or restoration strategies for those environments will be complex, and must be associated with a detailed knowledge of the important species in each area.

CONCLUSION
Floristic compositions varied among the forests studied here, with many species being unique to each site, and their presence was determined mainly by environmental characteristics, the influences of the adjacent vegetation, and anthropic impacts -all leading to low floristic similarity between them. Diversity, however, was generally low, due to ecological conditions linked to water stress. The different areas showed high variations in terms of their species contents (β2 diversity) and represent a major contribution to the total diversity of swamp forests within palm swamps in northern Minas Gerais State, Brazil.