Structure and Composition of Terra Firme and Seasonally Flooded V á rzea Forests in the Western Brazilian Amazon

: Research Highlights : Rare, or sparsely distributed, species drive the ﬂoristic diversity of upland, terra ﬁrme and seasonally ﬂooded forests in the central Juru á —a remote and hitherto ﬂoristically poorly known area in the Brazilian Amazon. Background and Objectives : Floristic inventories are critical for modelling and understanding the role of Amazonian forests in climate regulation, for sustainable management of forest resources and e ﬃ cient conservation planning. Yet, detailed information about the often complex spatial distributions of many Amazonian woody plants is limited. Here, we provide information about forest structure and species composition from a remote terra ﬁrme forest and an adjacent ﬂoodplain forest in the western Brazilian Amazon. More speciﬁcally, we ask (1) how ﬂoristically di ﬀ erent are the terra ﬁrme and ﬂoodplain forests? and (2) how variable is species composition within the same forest type? Materials and Methods : Between September 2016 and October 2017, we inventoried 97 plots (each 0.1 ha; 100 × 10 m) placed at least 800 m apart, with 46 plots in terra ﬁrme forest and 51 in seasonally ﬂooded forest. We included all trees, hemi-epiphytes and palms with diameter at breast height (dbh) > 10 cm and woody lianas > 5 cm dbh. We examine forest structure, family- and species-level ﬂoristic composition and species diversity within and between forest types using family and species importance values, rarefaction curves and dissimilarity matrices. Results : Terra ﬁrme forest and seasonally ﬂooded forest woody plant communities di ﬀ er both in structure and species composition, which was highly variable within forest types. Many species were shared between terra ﬁrme and seasonally ﬂooded forests, but most species were forest type-speciﬁc. Whereas species richness was greatest in the terra ﬁrme forest, ﬂoodplain species richness was among the highest regionally. Conclusions : Floodplain forests are a crucial complement to terra ﬁrme forests in terms of Amazonian woody plant diversity. and v á rzea forests. We discuss our ﬁndings in relation to wider patterns of forest structure and species distributions in the Amazon basin and conclude that v á rzea forests are an important complement to terra ﬁrme forests.


Introduction
Floristic inventories are critical for modelling and understanding the role of Amazonian forests in climate regulation, for sustainable management of forest resources and efficient conservation planning. Yet, for a number of reasons, floristic inventories in Amazonian forests are notoriously difficult and detailed information about the often complex spatial distributions of many Amazonian trees is

Study Area
This study took place in the central Juruá River basin, western Brazilian Amazonia. The region contains both seasonally flooded várzea (VZ) and lowland terra firme forests on paleo-várzea sediments (TF). The study area was located between 05 • 08 S, 67 • 01 W and 05 • 87 S, 67 • 88 W and includes the Uacari Sustainable Development Reserve (RDS Uacari, 632,949 ha), but excludes the Médio Juruá Extractive Reserve (ResEx Médio Juruá, 253,227 ha; Figure 1). The climate of the region is wet and tropical. Annual temperatures and rainfall average 27.1 • C and 3679 mm, respectively [31]. The elevation within the inventoried forests ranges from 67 to 153 m above sea level for terra firme and 68-137 m above sea level for várzea. The forests represent structurally intact vegetation.

Study Area
This study took place in the central Juruá River basin, western Brazilian Amazonia. The region contains both seasonally flooded várzea (VZ) and lowland terra firme forests on paleo-várzea sediments (TF). The study area was located between 05°08' S, 67°01' W and 05°87' S, 67°88' W and includes the Uacari Sustainable Development Reserve (RDS Uacari, 632,949 ha), but excludes the Médio Juruá Extractive Reserve (ResEx Médio Juruá, 253,227 ha; Figure 1). The climate of the region is wet and tropical. Annual temperatures and rainfall average 27.1 °C and 3679 mm, respectively [31]. The elevation within the inventoried forests ranges from 67 to 153 m above sea level for terra firme and 68-137 m above sea level for várzea. The forests represent structurally intact vegetation.  are shown in green with black borders. The map was generated in QGIS v.3.12.2, using background maps from the GADM database of Global Administrative Areas [32]. The shapefiles for the ResEx Médio Juruá and RDS Uacari were provided by Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) and the Amazonas State Environmental Agency (SEMA-DEMUC), respectively.

Floristic Inventories and Measurements
Between September 2016 and October 2017, we inventoried 97 plots (each 0.1 ha; 100 × 10 m), with 46 plots in terra firme forest (TF) and 51 in seasonally flooded várzea forest (VZ). The shortest distance between inventory plots was 800 m. To capture várzea forest at different inundation depths and periodicity as well as different soil types, topographic conditions and microhabitats, the plots were placed along transects that extended along the flooding gradient, with increasing elevation at greater distances from the main river channel.
Within each plot, all trees, hemi-epiphytes and palms ≥ 10 cm diameter at breast height (dbh), and all woody lianas ≥ 5 cm dbh, were measured and identified. Peripheral individuals straddling the boundary of the plots were included in the inventory if the mid-point of their trunks fell within the plot. We measured the dbh of buttressed trees immediately above the buttresses. When direct measurement with a dbh-tape was not possible (e.g., sulcate trunks, stranglers or where buttresses were too high), we estimated the diameters. We used a Haglöf Vertex IV and Transponder T3 to measure tree, palm and hemi-epiphyte total heights based on trigonometric calculations using the measuring angle and distance to the trunk [33], and/or estimated total heights where this was not possible. For trees and hemi-epiphytes, we also determined the height of the first branch. For palms, we determined the height of the stem. To remove observer bias, the same person (Y.K.B.) administered all height measurements.
All individuals were aluminium-tagged, numbered and identified in situ and/or in the herbarium at the National Institute of Amazonian Research (INPA), Manaus, Brazil. Skilled INPA herbarium technicians with extensive field and herbarium experience from floristic inventories in the central-western Brazilian Amazon performed all identifications. Vouchers from 1174 individuals were collected and subsequently identified at the INPA herbarium to verify the accuracy of field identifications at the level of genus and species. Individuals that could not be determined to species level were sorted to morpho-species or, where applicable, higher taxonomic levels.

Data Analyses
To test for differences in woody plant stem density, dbh, basal area (BA), total height, height of first branching, branching depth and proportion of stem with branches in relation to total height, we first ran Shapiro-Wilk's tests of normality and compared the variances of terra firme and várzea using Fisher's F-test. For normally distributed data, we ran Student's two-sample t-tests where data conformed to homoscedasticity, or Welch two-sample t-tests where they did not. Where the data did not conform to normality, we ran independent two-group Wilcoxon-Mann-Whitney tests.
We used the 'BiodiversityR' package version 2.11-1 [36] to calculate indices of species richness and diversity, and to produce species rarefaction curves estimating the expected number of additional species for every additional survey plot, in relation to the mean number of individuals per plot. Species rarefactions were based on 100 permutations.
To investigate the spatial variation in woody plant species composition, we used the 'vegan' package, version 2.5-5 [37]. We tested for spatial autocorrelation among plots using a partial Mantel test with a Bray-Curtis dissimilarity matrix for the woody plant species composition, and a Euclidian distance matrix for the geographic distances [37]. To assess variations in species composition, we used non-metric multidimensional scaling (NMDS) through the metaMDS function with the Bray-Curtis dissimilarity index. We used the envfit method to fit forest type (i.e., TF or VZ) onto the NMDS ordination as a measure of the correlation of forest type with the NMDS axes. Additionally, we performed a permutational multivariate analysis of variance test (PERMANOVA) with forest type as predictor of the woody plant composition dissimilarity matrices, with the Bray-Curtis index as the response variable.
To analyse for multivariate homogeneity of group dispersions, we used the betadisper function in vegan. Analyses for multivariate homogeneity of group dispersions inform us how the variances within groups differ among groups [38]. Defining β-diversity as the variability in species composition among sampling units within groups, tests of multivariate homogeneity of group dispersions may thus inform us about differences in β-diversity between the two forest types [39]. To minimise the influence of the most abundant species on the multivariate dispersion analyses, we first square-root transformed the species abundance matrix [40]. To adjust for potential small sample bias in the analyses, we used the bias.adjust option of the betadisper function [40,41]. All analyses were run in R, version 3.5.2 [42].

Forest Structure
In total, we recorded 4690 individual trees, 274 palms, 25 hemi-epiphytes and 450 lianas across both terra firme and várzea forests (n = 97 plots; 9.7 ha), yielding 5439 individuals or 5483 stems ( Table 1). The dbh size class distributions in the two forest types show inverse J-shaped curves for both the combined tree and hemi-epiphyte assemblages and the lianas ( Figure 2). For palms, the size class distribution was a sigmoid shape, showing a slight shift in climax towards larger diameters in várzea palms (20-25 cm dbh) compared to terra firme palms (15-20 cm dbh; Figure 2). Table 1. Number of plots (Plots) and number of ha. (Ha.) inventoried in terra firme (TF) and várzea (VZ) forests along the Juruá River, western Brazilian Amazon. Number of stems (Stems), including hollow stems, and number of individuals (Inds.), including multi-stemmed individuals, are given as count data with percentiles in parentheses (%). Mean number of stems per plot (Plot mean) is given ± standard deviations (sd). Mean diameter at breast height (dbh) ± sd is in cm, basal area (BA) in m 2 and mean height ± sd in m. All values are given per growth form, forest type and for both forest types combined. Values refer to trees, palms and hemi-epiphytes (hemi-ep.) with dbh ≥ 10 cm and woody lianas ≥ 5 cm dbh. Total height in m is also given as overall minimum (Min), maximum (Max), median and mode values. Differences in stem density, dbh, plot BA and mean total height between várzea and terra firme for normally distributed data were tested with classic Student's two-sample t-tests where group variances were homogenous or Welch two-sample t-tests where group variances were heterogenous. Where data did not conform to normality, we used independent two-group Wilcoxon-Mann-Whitney tests. Asterisks in the   Smaller trees measuring <30 cm dbh dominated both forest types. These accounted for 72.8% of all inventoried individuals in terra firme (80.9% of all terra firme trees) and 66.4% of all individuals in seasonally inundated forest (79.1% of all várzea trees). Large trees (≥70 cm dbh) represented just 2.0% of all trees (1.5% and 2.4% of the TF and VZ trees, respectively), or 1.7% of all individuals (1.4% and 2.0% of the TF and VZ individuals, respectively). Only 18 (0.4%) trees in the entire sample attained Forests 2020, 11, 1361 7 of 20 diameters greater than 100 cm dbh, two in terra firme and 16 in várzea. Six of these sixteen emergents were Hura crepitans (Euphorbiaceae) in várzea.

Floristic Diversity
In total, 931 species were recorded in the lowland terra firme and várzea forests combined ( Table 2). Of these, 625 species occurred in terra firme and 526 in várzea ( Table 2). Two hundred and twenty species (23.6%) were shared among terra firme and várzea forests, comprising 44.4% of all individuals. However, most species were unique to either terra firme (43.5%; Table 2) or várzea (32.9%; Table 2) and many species occurred only in a single plot (TF: 45.6%, n = 285; VZ: 42.2%, n = 222). Most of the shared species were rare and occurred with few observations in one forest type and single observations in the other (69.5%, n = 153), or as singletons in both forest types (13.2%, n = 29). Only 44 (20.0%) of the 220 shared species had 10 or more individuals recorded in at least one forest type. Three species (1.4%) had 10 or more individuals recorded in both forest types. Although the species rarefaction curves clearly indicate a greater species richness in terra firme compared to várzea, the curves did not reach an asymptote for either forest type ( Figure 3). Table 2. Number of species (Spp.), number of genera (Gen.) and number of families (Fam.) found within the terra firme (TF) and várzea (VZ) forests along the Juruá River, western Brazilian Amazon. Values are given as counts per growth form, forest type and for both forest types combined, with percentiles of individuals not identified to each taxonomic level in parentheses (N/A, %). In addition, the numbers of Spp., Gen. and Fam. that were unique to either forest type (Unique) or occurred as singletons in either or both forest types (Singleton) are given as counts and percentiles in parenthesis (%). All values refer to trees, palms and hemi-epiphytes (hemi-ep.) with dbh ≥ 10 cm and woody lianas ≥ 5 cm dbh.

TF VZ Total
Spp.

Family Importance Value
Leguminosae (Fabaceae) dominated the family importance value (FIV) in both forest types, mainly because of the large number of species in this super-family (TF: FIV = 40.93; 15.7%; VZ: FIV = 38.98; 13.9%; Table 3; Table 4). In both forest types, Lecythidaceae was the second most important family, followed by Sapotaceae. In terra firme, Lecythidaceae represented both the highest number of individuals (n = 383) and the greatest basal area (BA) ( Table 3). In várzea, Lecythidaceae was the second most dominant family, Sapotaceae was the second most species-rich family and Annonaceae was the second most abundant family ( Table 4). The full FIV list for all families is presented in Supplementary Table S1.   Table 3; Table 4). In both forest types, Lecythidaceae was the second most important family, followed by Sapotaceae. In terra firme, Lecythidaceae represented both the highest number of individuals (n = 383) and the greatest basal area (BA) ( Table 3). In várzea, Lecythidaceae was the second most dominant family, Sapotaceae was the second most species-rich family and Annonaceae was the second most abundant family ( Table 4). The full FIV list for all families is presented in Supplementary Table S1.

Species Importance Value Index
Three Eschweilera spp. (Lecythidaceae) top the terra firme Importance Value Index (IVI). Of these, Eschweilera coriacea was the most important, largely due to the high basal area derived from its large stems and high abundance (Table 5). In várzea, Hura crepitans (Euphorbiaceae) was the most important tree species and dominated the basal area, despite its relatively low abundance (Table 6). Palms (Arecaceae) were abundant in both forest types and both Euterpe precatoria (TF) and Astrocaryum jauari (VZ) were among the most important species. None of the 10 most important species were shared between terra firme and várzea. The IVI for all species is presented in Supplementary Table S2. Table 5. The ten most important species in lowland terra firme, listed in descending order of Importance Value Index (IVI). Values are based on all woody plants with species identifications. Growth forms (G.F.) observed for each species are given as t = tree, l = liana and p = palm. The number of individuals within each species (No. Inds.) and the number of plots in which each species occurs (Plot occ.) are given as counts. Relative density (Rel. Den.), relative dominance (Rel. Dom.) and relative frequency (Rel. Freq.) are given as percentages, where 100 equals 100% (Supplementary Table S1

Community Composition
Overall dissimilarity in species composition was high among plots. Only four between-plot Bray-Curtis dissimilarities were below 60%, all within várzea. The lowest recorded Bray-Curtis dissimilarity between forest types was 79.6% (Supplementary Table S3). No species occurred in all plots of either forest type and only two species occurred in more than half of the terra firme plots: Eschweilera wachenheimii (Lecythidaceae; n = 36) and Eschweilera coriacea (Lecythidaceae; n = 33). Despite a lower total species richness in várzea, only Virola surinamensis (Myristicaceae) occurred in at least half of the várzea plots (n = 26).
We found greater resemblance in species composition among plots within the same forest type than when comparing plots between forest types (envfit: R 2 = 0.59, p < 0.001; PERMANOVA: R 2 = 0.11, F = 11.27, p = 0.001; Figure 4), although there was spatial autocorrelation between plots (Mantel test: r = 0.19, p = 0.001). Multivariate dispersion of inventory plots indicates that neither várzea nor terra firme plots are more clustered around their respective multivariate means than the other (betadisper: F = 0.30, N.Perm = 99, p = 0.57). Thus, both forest types show a similar variation in species composition among plots.

Community Composition
Overall dissimilarity in species composition was high among plots. Only four between-plot Bray-Curtis dissimilarities were below 60%, all within várzea. The lowest recorded Bray-Curtis dissimilarity between forest types was 79.6% (Supplementary Table S3). No species occurred in all plots of either forest type and only two species occurred in more than half of the terra firme plots: Eschweilera wachenheimii (Lecythidaceae; n = 36) and Eschweilera coriacea (Lecythidaceae; n = 33). Despite a lower total species richness in várzea, only Virola surinamensis (Myristicaceae) occurred in at least half of the várzea plots (n = 26).
We found greater resemblance in species composition among plots within the same forest type than when comparing plots between forest types (envfit: R 2 = 0.59, p < 0.001; PERMANOVA: R 2 = 0.11, F = 11.27, p = 0.001; Figure 4), although there was spatial autocorrelation between plots (Mantel test: r = 0.19, p = 0.001). Multivariate dispersion of inventory plots indicates that neither várzea nor terra firme plots are more clustered around their respective multivariate means than the other (betadisper: F = 0.30, N.Perm = 99, p = 0.57). Thus, both forest types show a similar variation in species composition among plots.

Forest Structure
As seen from the high number of late-successional species characteristic of the central Juruá and the different strata that these represent, our inventory is typical of structurally intact and late-successional forests [43][44][45]. Both the terra firme and várzea forests of the central Juruá had well-stratified canopies featuring emergent trees of up to ca. 50 and 47 m, respectively.
The várzea forest had a greater diversity in growth forms than terra firme, and more multi-stemmed or hollow individuals. This may reflect the differences in disturbance regimes between the two forest types, with higher levels of disturbance in várzea forests driven by the impact of seasonal floods and their proximity to the Juruá River. For example, woody lianas typically occur in disturbed areas, such as secondary forests or forest edges [46,47]. In structurally intact terra firme forests, lianas are likely to become dominant only in treefall gaps [47], whereas with both natural clearings and the river margin, the propensity of edge habitat is considerably larger in structurally intact várzea forests. Similarly, the high number of hollow individuals that we observed in várzea forest was mostly driven by pioneer species typical of disturbed floodplain habitat, such as Cecropia spp. [48][49][50]. Moreover, since palms are associated with highly dynamic forests on weakly structured and nutrient-rich soils [51][52][53], the higher frequency and size of palms recorded in várzea may further reflect the influence of flooding on substrate properties and forest dynamics.
Both the terra firme and várzea forests had similar stem densities and high proportions of smaller trees (i.e., 10-30 cm dbh). However, trees grew taller in terra firme, whereas BA and degree of branching were significantly higher in várzea. These structural differences between terra firme and várzea woody plants may result from differences in forest dynamics and substrate fertility. High seasonality and substrate fertility in várzea might cause trees to grow quicker or better during favourable times of the year (i.e., dynamic growth in response to the changing environment) [54], thus supporting higher BA but potentially lighter wood density [55]. In contrast, less seasonal variability and lower substrate fertility in terra firme may cause woody plants to grow slower, but more evenly, throughout the year. Slow lateral growth results in more structural matter per unit volume wood, and thus greater stability, supportive of higher stems [56].
Structural and functional differences between forest types interact to determine the amount of standing, living woody biomass across the forest landscape. Therefore, the higher BA and degree of branching in várzea woody plants is potentially counterbalanced by taller terra firme stems, plus previous findings from the same region which show that terra firme trees store more carbon per unit volume than várzea conspecifics [57]. Hence, both várzea and terra firme may produce similar amounts of standing, live woody biomass in the Juruá. This would compare to a case from the southern Amazon where dry season length and storm frequency affected stem density and individual biomass of trees and palms differently across two forest types but resulted in similar forest biomass due to complementary responses in structural variables to these environmental stresses [58].

Floristic Composition and Diversity
In the Juruá, almost one quarter of all woody plant species (23.6%) occurred in both terra firme and várzea. Our findings thus support previous reports of several shared species among terra firme and seasonally flooded forests [29,59]. However, most of these occurred predominantly in one forest type or as singletons in both forest types. This could indicate that many of the shared species are generally rare within the forest matrix or represent outlier observations of individuals in one of the forest types where they would straddle the extremes of their environmental tolerance limits [60][61][62][63]. Thus, we see that differences in environmental stress, e.g., seasonal flooding versus no seasonal flooding, between várzea and terra firme forests limit species distributions and cause the woody plant communities to shift. The great dissimilarity in species composition among várzea forest plots may result from the diversity of microhabitats and successional stages they cover along the hydro-topographic gradient [29,43]. This variability in species composition contributed to a high total species richness. The species rarefactions suggest that the central Juruá várzeas are some of the most species-rich floodplains in the Amazon [24,64].
Amazonian terra firme forests are well-documented to be more species-rich than seasonally flooded forests [17]. This is supported by our study, where terra firme displayed a higher total species richness than the várzea forest. At the Amazon basin-wide scale, the greater diversity in terra firme woody plants is attributed to habitat availability (terra firme comprises ca. 87% of available forest habitat in the Amazon compared to ca. 13% forested floodplain habitat) [3,65], habitat stability [66], a diversity in climatic and edaphic conditions [67][68][69][70] and the evolutionary dynamics of land formations, e.g., through processes that undo or induce dispersal barriers and subsequent speciation [71]. At local scales, a higher diversity in terra firme woody plant communities compared to its floodplain counterpart may also be attributed to a greater stability and longer history. Terra firme habitat has been available for colonisation by woody plants for much longer than present várzea habitat. Moreover, even at this local scale, the rate of disturbance in the terra firme is much lower compared to the várzea, where forest habitat is formed and eroded on a dynamic, seasonal basis [43,72]. Given these different drivers of woody plant diversification across seasonally flooded and terra firme forests, it is perhaps not surprising that terra firme and várzea forests in the Juruá showed similar levels of variation in species composition among plots around their respective multivariate means (i.e., similar β-diversities).
A recent study found that six of the ten most common Amazonian arborescent species were palms [82]. In the Juruá, palms contributed 4%-6% of the inventoried individuals and 2%-5% of the total BA in the terra firme and várzea forests, respectively. Euterpe precatoria, potentially the most common woody species across the entire Amazon [83], was the most prominent terra firme palm species for the Juruá, where it was twice as common and more widespread than in várzea. Astrocaryum jauari was the most important várzea palm species. Overall, however, Eschweilera tree species were particularly prominent in the terra firme forest and Eschweilera coriacea was the most common tree, both for the Juruá and the Amazon at large [82]. Hura crepitans (Euphorbiaceae) was the most important floodplain species. In fact, Hura crepitans, as well as Virola surinamensis (Myristicaceae), the second most important várzea species in the Juruá, are both scarce in many floodplain areas across the Amazon basin due to logging [84,85]. Their importance in the central Juruá may therefore reflect the protected status of these floodplains [86].

Conclusions
Rare or sparsely distributed species drive most of the woody plant diversity in both low-lying terra firme forests on paleo-várzea sediments and seasonally flooded várzea forests on the floodplain of the central Juruá River basin. Both terra firme and várzea show high variation in plot-level species composition, demonstrating heterogeneity within forest types, even at small spatial scales. Although species richness was highest in terra firme, the Juruá várzea forest contain more woody species than most inventories have recorded for Amazonian floodplain forests. Given the high species turnover across terra firme and várzea, floodplain forests are clearly an important complement to terra firme woody plant diversity. The high proportion of singleton observations and forest type specialists in the central Juruá highlight the need for further floristic inventories from a wider range of geographically remote areas if we are to discover and properly describe the Amazonian flora. As a step in that direction, this study helps address the patchy botanical records of sparsely distributed Amazonian woody species.
Supplementary Materials: The following are available online at http://www.mdpi.com/1999-4907/11/12/1361/ s1, Table S1: Family Importance Value, Table S2: Species Importance Value Index, Table S3: Bray-Curtis Dissimilarity Matrix. Acknowledgments: This publication is part of the Instituto Juruá series (www.institutojurua.org.br). We thank the Secretaria do Estado do Meio Ambiente e Desenvolvimento Sustentável do Amazonas (SEMA-DEMUC) and Instituto Brasileiro do Meio Ambiente e Recursos Naturais Renováveis (IBAMA)/Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) for authorising the research. We are grateful to the members from the Projeto Médio Juruá team for logistical support and extend special thanks to Andressa Scabin. We thank Michael J.G. Hopkins, Rafael Leandro de Assis and Juliana Schietti with colleagues at the National Institute for Amazonian Research (INPA) for providing logistical assistance and a place to work in Manaus. We thank Paulo Apóstolo Costa Lima Assunção and Alexandro Elias dos Santos, assisted by Lorena M. Rincón, for identifying the species in the field and at INPA. Marcos Viejo Somoano has been of great help with the graphical layout of the figures. Finally, we thank our local field assistants, Associação dos Produtores Rurais de Carauari (ASPROC), Operação Amazônia Nativa (OPAN), and the people of the central Juruá, who, in various ways, assisted us throughout our work.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.