Advertisement
Research Article

Changes in Tree Reproductive Traits Reduce Functional Diversity in a Fragmented Atlantic Forest Landscape

  • Luciana Coe Girão,

    Affiliation: Programa de Pós-Graduação em Biologia Vegetal, Departamento de Botânica, Universidade Federal de Pernambuco, Recife, Pernambuco, Brazil

    X
  • Ariadna Valentina Lopes mail,

    To whom correspondence should be addressed. E-mail: avflopes@ufpe.br

    Affiliation: Departamento de Botânica, Universidade Federal de Pernambuco, Recife, Pernambuco, Brazil

    Current address: Department of Wildlife Ecology and Conservation, University of Florida, Gainesville, Florida, United States of America

    X
  • Marcelo Tabarelli,

    Affiliation: Departamento de Botânica, Universidade Federal de Pernambuco, Recife, Pernambuco, Brazil

    X
  • Emilio M. Bruna

    Affiliation: Department of Wildlife Ecology and Conservation, University of Florida, Gainesville, Florida, United States of America

    X
  • Published: September 19, 2007
  • DOI: 10.1371/journal.pone.0000908

Abstract

Functional diversity has been postulated to be critical for the maintenance of ecosystem functioning, but the way it can be disrupted by human-related disturbances remains poorly investigated. Here we test the hypothesis that habitat fragmentation changes the relative contribution of tree species within categories of reproductive traits (frequency of traits) and reduces the functional diversity of tree assemblages. The study was carried out in an old and severely fragmented landscape of the Brazilian Atlantic forest. We used published information and field observations to obtain the frequency of tree species and individuals within 50 categories of reproductive traits (distributed in four major classes: pollination systems, floral biology, sexual systems, and reproductive systems) in 10 fragments and 10 tracts of forest interior (control plots). As hypothesized, populations in fragments and control plots differed substantially in the representation of the four major classes of reproductive traits (more than 50% of the categories investigated). The most conspicuous differences were the lack of three pollination systems in fragments-pollination by birds, flies and non-flying mammals-and that fragments had a higher frequency of both species and individuals pollinated by generalist vectors. Hermaphroditic species predominate in both habitats, although their relative abundances were higher in fragments. On the contrary, self-incompatible species were underrepresented in fragments. Moreover, fragments showed lower functional diversity (H' scores) for pollination systems (−30.3%), floral types (−23.6%), and floral sizes (−20.8%) in comparison to control plots. In contrast to the overwhelming effect of fragmentation, patch and landscape metrics such as patch size and forest cover played a minor role on the frequency of traits. Our results suggest that habitat fragmentation promotes a marked shift in the relative abundance of tree reproductive traits and greatly reduces the functional diversity of tree assemblages in fragmented landscapes.

Introduction

Functional diversity can be defined as a variety of life-history traits presented by an assemblage of organisms [1], [2] and it has been postulated to be critical for the maintenance of ecosystem processes and properties [3]. For example, previous empirical work has suggested that ecosystems with a high diversity of functional traits have greater efficiency of water, nutrient, and light use, as well as higher productivity [3], [4]. In addition, they may also be more resilient [5] and resistant to biological invasions and to biodiversity loss [6], [7]. Nevertheless, most studies on functional diversity in plant communities have focused on the importance of traits associated with plant physiology. Consequently, we know little regarding the functional diversity of other traits that also affect both community structure and ecosystem functioning, such as those related to plant-animal interactions [1], [2].

Habitat loss and fragmentation (hereafter habitat fragmentation) have been shown to dramatically alter tree communities in tropical forests [8][12]. Fragments usually exhibit reduced species richness and diversity, particularly near edges. This reduction in species diversity is due in large part to loss of species that are “shade-tolerant” [8], [12], [13], restricted to the forest understory [10], have large-seeds [14], [15], or are dispersed by vertebrates [12], [16][19]. Furthermore, fragments tend to become dominated, both in terms of species richness and individual abundance, by pioneer trees [8], [19]. Because tropical pioneer trees usually share a similar set of life-history traits irrespective of their taxonomic affinities [20][22], this biased ratio of pioneers to shade-tolerant plants may reduce the functional diversity of tree assemblages in fragments.

More than 90% of the extant angiosperms are animal-pollinated [23], therefore pollination is considered an essential ecosystem process whose outcome can have major consequences for the maintenance of biodiversity [24], [25]. Indeed, a broad body of empirical evidence has found that the disruption of plant-pollinator interactions by habitat fragmentation can detrimentally affect plant reproductive success [26][29]. Potentially, changes in plant-pollinator interactions and pollinator abundance/composition can affect seed dispersal and seedling recruitment and consequently reduce plant population size or even promote local extinction [26], [27], [30]. Nevertheless, patterns and process regarding changes in reproductive functional diversity in fragmented tropical landscapes remain poorly investigated.

Because the long lifespan of tropical trees [31], hypotheses addressing disruptions of functional diversity driven by changes in tree composition can be properly tested in landscapes that were disturbed long enough ago to permit demographic shifts to have occurred, such as fragmented landscapes with longer histories of human occupation. The Atlantic forest of Brazil is a biodiversity hotspot that has been reduced to less than 8% of its original distribution due to forest clearing and fragmentation that dates to the 16th century [32]. In some regions (e.g. Brazil's northeast), over 90% of fragments are smaller than 50 ha and are immersed in a homogeneous and hostile matrix of sugar cane fields [33]. These archipelagos of small fragments and forest edge habitat are currently dominated by a small subset of pioneer trees, retain less than half of the tree species richness of the forest interior [19], and receive an impoverished seed rain biased towards smaller seeds [34]. This scenario offers an excellent opportunity to investigate long-term fragmentation-related changes in tree assemblages and how they influence functional diversity.

Here we test the hypothesis that the habitat fragmentation changes the frequency of tree species and individuals within categories of reproductive traits and consequently reduces the functional diversity of tree assemblages in a fragmented landscape of the Brazilian Atlantic forest. We begin by comparing the pollination systems, floral biology, sexual systems, and reproductive systems of trees in forest fragments and tracts of forest interior (control plots). We then compare the diversity of these traits in these two habitats based on the relative contribution of both species and individuals. Finally, we discuss potential mechanisms driving the patterns we observed. We conclude that habitat fragmentation promotes a marked shift in the relative abundance of tree reproductive traits, including the lack of some specialized pollination systems and a parallel increase in the frequency of generalist ones. Collectively, shifts in reproductive traits promote a conspicuous reduction in the functional diversity of tree assemblages in fragmented landscapes, which may strongly influence forest dynamics and the persistence of biodiversity.

Materials and Methods

Study site and landscape attributes

This study was conducted in the State of Alagoas in northeastern Brazil on the property of Usina Serra Grande (8°58′50″S, 36°04′30″W), a large, privately-owned sugar producer. This landholding has approximately 9,000 ha of forest included in a unique biogeographic region of the Atlantic forest known as the Pernambuco Center of Endemism [sensu 35] or the Atlantic forest of Northeast Brazil, the most threatened sector of the South American Atlantic forest [17]. We selected a large (666.7 km2), severely fragmented landscape within this property containing 109 forest fragments (total forest cover = 9.2%), including the 3,500-ha Coimbra forest–the largest and best preserved remnant in this region [19]. All fragments are entirely surrounded by a uniform matrix of sugar-cane monoculture (Figure 1). In addition to the Coimbra forest, the patches ranged in size from 1.67–295.7 ha.

thumbnail

Figure 1. Study landscape at the Atlantic forest of northeast Brazil.

(A) Northeastern Brazil, where this study was conducted. (B) Distribution of the Atlantic forest of northeast Brazil ( = Pernambuco Center of Endemism), note original (grey) and current (black) distribution of this forest in the region; white rectangle represents the study landscape (amplified in C). (C) Study landscape with the fragments used in this study (dark grey polygons), including the 3,500 ha Coimbra forest (lower right). Light grey and white areas represent remaining forest fragments (not sampled) and sugar-cane cultivation, respectively.

doi:10.1371/journal.pone.0000908.g001

Our study landscape consists of a low-altitude plateau (300–400 m above sea level) containing two similar classes of dystrophic soils with high clay fractions: yellow-red latosols and yellow-red podzols (according to the Brazilian soil classification system [36]). Annual rainfall is ~2000 mm, with a 3-month dry season (<60 mm/month) from November to January. Forests in this landscape consist of lowland terra firme forest (<400 m a.s.l.) [37], with the Fabaceae, Lauraceae, Sapotaceae, Chrysobalanaceae and Lecythidaceae accounting for most tree species (≥10 cm DBH) [38], [39]. Sugar cane cultivation in this landscape, which dates to the early 19th century, and possibly as early as the 18th century [see 40], provided the strongest incentive for clearing large tracts of pristine old-growth forests. Remaining forest fragments have been protected against fire and logging to ensure watershed protection and water supply for sugar cane irrigation (C. Bakker, pers. communication). This protection has guaranteed the stability of forest fragment borders and the occurrence of both pioneer and shade-tolerant adult trees along forest edges as evidenced by local patterns of seed rain [15]. The Serra Grande landscape therefore provides a rare and interesting opportunity for Atlantic forest fragmentation studies.

Tree species surveys and habitat classification

We compared the frequencies and the functional diversity of tree reproductive traits in 10 of the forest fragments (range = 3.4–295.7 ha) and 10 ‘forest interior plots’ [sensu 41] located in the region's largest remnant (Coimbra forest; here adopted as the control site) using floristic data from previously conducted botanical surveys. Although we are aware that the Coimbra forest does not represent a true ‘continuous forest’, it is the largest remaining Atlantic forest patch in Northeast Brazil [see 19] and is more than twice as large as the largest fragment analyzed by Ranta et al. [33] in this same center of endemism. In addition, the Coimbra forest still retains the full complement of ecological groups occurring in more continuous tracts of Atlantic forest, such as large-seeded trees and frugivorous vertebrates [16], [17], [42], [43]. It is therefore representative of the largest tracts of forest remaining in the hotspot, making its core area [sensu 41] the best possible control site for assessing persistent and long-term effects of habitat loss and fragmentation.

The tree surveys, upon which we randomly selected our fragments and control plots, were carried out from 2003–2005 by Oliveira et al. [19] and Grillo [39] as part of a regional plant survey. Briefly, all trees ≥10 cm DBH were measured, marked, and identified in one 0.1-ha plot per fragment. Plots were located in the geographic center of fragments to standardize procedures and minimize edge effects [44]. Depending on the size of the fragment, plots were 60.5–502.77 m from nearest edge. The ten control plots, also measuring 0.1-ha, were haphazardly located in the interior of Coimbra forest at distances 200–1012.73 m from nearest edge, in locations consisting of old-growth forest with no detectable edge effects (i.e. forest interior [sensu 41]). Vouchers collected by Oliveira et al. [19] and Grillo [39] are deposited in the Herbarium UFP (No. 34.445 to 36.120), and the checklist of the flora of Usina Serra Grande (ca. 650 plant species) is available at www.cepan.org.br and in Pôrto et al. [45]. Since 2001, the number of botanical investigations carried out in our study landscape has increased [e.g. 15], [19], [34], [38], [39], [45], [46], providing detailed knowledge about the taxonomy and life-history traits of the woody flora.

Reproductive traits of tree species

Floristic surveys revealed a total of 629 individuals from 77 tree species in the forest fragments (32 families, 58 genera), whereas 878 individuals from 119 species (37 families, 87 genera) were recorded in the control plots. Pooling the data from all sites resulted in 1507 individuals from 156 species (41 families, 105 genera) (see Table S1 in the Supporting Information). For each species we identified the following “reproductive traits”: pollination system, floral biology, sexual system, and reproductive system (Table 1). Classification of species into each category was based on (1) floras and botanical monographs [e.g. 47], [48][50], including several issues of Flora Neotropica; (2) web searches including only published and referenced data; (3) field observations and a survey of specimens from the UFP and IPA Herbaria; and (4) personal knowledge and previously published observations [see 51 for a review]. For each fragment and control plot we then calculated the proportion of tree species and individuals within the 50 categories that comprise the four major classes of reproductive traits (Table 1). Although not all categories could be identified for a few of the species (see results), it is unlikely that this biases the qualitative outcome of our analyses because habitats were compared in terms of frequency of species and individuals within categories.

thumbnail

Table 1. Tree reproductive traits with their respective categories adopted in this study.

doi:10.1371/journal.pone.0000908.t001

Explanatory variables

Because a number of patch and landscape-scale environmental variables may affect the structure of tree assemblages in tropical forests [8], [52], we also considered the effects of soil type, distance to the nearest forest edge, forest fragment size, the spatial distribution of plots (i.e. plot location in the landscape), and the amount of forest cover retained in the surrounding landscape (hereafter forest cover) as independent variables for the frequency of reproductive traits in the tree assemblages. Forest cover is positively correlated with overall connectivity between patches [53] and was quantified as the percentage of forest within a 1-km width buffer set from the border of each fragment. Patch and landscape metrics were quantified using a combination of three Landsat and Spot images acquired in 1989, 1998, and 2003, a set of 160 aerial photos (1:8,000) taken from commissioned helicopter overflights on April 2003, a soil map by IBGE [36], and a soil map provided by the Usina Serra Grande Agriculture Office. Analyses were conducted using ArcView 3.2 and Erdas Imagine 8.4.

Functional diversity of reproductive traits

Here we operationally define a functional group as a set of tree species within the same category of reproductive trait, i.e. a set of species sharing a life-history trait as previously adopted elsewhere [1]. To calculate the functional diversity of reproductive traits in forest fragment and control plots, we used Shannon's (log base 2) and Simpson's indices [54]. We used both indices to elucidate the contribution of both the richness of categories and the evenness to diversity scores (note that the use of evenness-based indices for estimating functional diversity has been recommended by some authors [55][57]). We calculated these indices twice for each of the 20 plots: first, using categories as the equivalent of species, and the number of tree species within each category as the equivalent of individuals; and second using categories as the equivalent of species and the number of individual within each category.

Statistical analysis

Differences in (1) the average percentage of species and individuals within each category of reproductive trait, and in (2) the average functional diversity of reproductive traits between the control area and fragments were compared with t or Mann-Whitney tests [58]. General linear models (GLM) were used to detect any effect of explanatory variables on the frequency of traits in tree assemblages by first examining the effects of habitat type (fragments vs. control plots), soil type and distance to the nearest edge considering all 20 plots in the two habitats, and then the effects of forest fragment size and surrounding forest cover considering the 10 fragments (since these patches and landscape metrics had no variance in Coimbra forest). Normality of all response variables were checked using Lilliefors tests; for GLMs the percentage-expressed dependent variables were arcsine transformed as suggested by Sokal & Rohlf [58].

Additionally, to examine the effect of habitat and soil type on species similarity between plots these variables were considered as factors in Analysis of Similarities (ANOSIM) tests [59]. Plots were ordered according to their Bray-Curtis dissimilarities of species composition [54]. Species abundance were square-root transformed and standardized [59] to avoid any bias resulting from very abundant species and differences in the sample size of individuals recorded within each plot. We also performed a Mantel test with Weighted Spearman rank correlations to address the effect of plot geographic location on levels of taxonomic similarity. Straight-line distances between plots were ln-transformed, as suggested by Condit et al. [60] and Jones et al. [61]. The Mantel test was carried out considering a group of 20 fragments and 75 0.1-ha plots from which information on tree species composition is available [19], [39]. Here we assume that the lack of significant relationships between soil type, plot location and plot floristic similarity discard soil and plot location as variables driving the frequency of tree reproductive traits in the landscape. All analyses were carried out using SYSTAT 6.0 [62], PRIMER v. 5 [63], and PC-ORD 4.36 [64].

Results

Reproductive traits of tree species

Fragments and control plots differed significantly in more than 50% of the categories of reproductive traits investigated, but differences were much more notable when evaluating individuals within categories (over 60% of the categories differed) than species (ca. 40%).

For pollination systems, fragments and control plots markedly differed in 50% of all categories (6 out of 12 categories) (Table 2; Figure 2A). The most conspicuous differences concerning species richness within categories of pollination systems can be summarized in four aspects. First, fragments lacked three categories of pollination systems–pollination by birds, flies and non-flying mammals. Second, scores for hawkmoth- and bat-mediated pollination in fragments were about half of the scores recorded in the control plots. Third, when comparing pollination by vertebrates as whole (birds, bats, and non-flying mammals) fragments had a ca. threefold decreased frequency than control plots. Finally, fragments had a 33% increase in the proportion of tree species pollinated by diverse small insects (DSI) in comparison to control plots (Table 2). The proportion of tree individuals within categories of pollination systems showed similar trends (Figure 2A), although for some categories the differences between fragments and control plots were even more dramatic than for species richness (e.g. hawkmoth and vertebrate pollination). Fragments and control plots also differed dramatically when pollination systems were pooled into two categories of pollen vectors-generalists and specialists [sensu 65]. In summary, fragments had proportionately more tree species pollinated by generalist vectors (66.43±14.08%) than control plots (58.18±7.87%; t = 1.616; d.f. = 18; P = 0.06); the relative abundance individuals pollinated by generalists was also higher in fragments than control plots (71.71±16.5% vs. 46.10±15.53, U = 13.0; P = 0.0052).

thumbnail

Figure 2. Effect of habitat fragmentation on pollination systems, floral sizes and floral rewards.

Percentage of tree individuals within categories of pollination systems (A; N = 137 spp.), floral sizes (B; N = 105 spp.), and floral rewards (C; N = 124 spp.) at 10 fragments and 10 control plots of an Atlantic forest landscape, northeastern Brazil. Frequencies represented by boxes that are significantly different are indicated with asterisks: *P<0.05; **P<0.01; ***P<0.001.

doi:10.1371/journal.pone.0000908.g002
thumbnail

Table 2. Percentages (mean±SD) of tree species within categories of reproductive traits in forest fragments (N = 10) and control plots (N = 10) in a fragmented landscape of Atlantic forest, northeastern Brazil (data on the reproductive traits for the species are available upon request).

doi:10.1371/journal.pone.0000908.t002

The proportion of species within categories of floral size was similar in fragments and control plots (Table 2). However, fragments had twice as many individuals with inconspicuous flowers than control plots (50.75±25.44% vs. 29.99±15.86%; Figure 2B). An opposite trend was observed for large and very large flowers, fragments with more than a 10-fold lower proportion of individuals with large flowers (0.5±0.84%) and almost a three-fold decrease of the very large ones (13.74±11.77%) in comparison to control plots (7.54±8.58% and 35.4±13.54%, respectively). By grouping the five categories of flower size into two [i.e. inconspicuous/small (≤10 mm) and medium/very large (>10 mm)] results were similar. Fragments showing a prevalence of individuals with inconspicuous/small flowers (68.85±21.43%) in contrast with control plots (45.98±16.04%), and a significant lower proportion of individuals with medium to very large flowers ones (31.15±21.43%) than control sites (54.02±16.04%) (Figure 2B).

Nectar was the most frequent floral reward observed in tree species of fragment and control sites, however, these habitats differed in two of the other five categories of floral rewards adopted in this study (Table 2). Nectar/pollen-flower species were twice as higher in fragments than in control plots, and fragments had also higher frequency of species with oil-flowers in comparison with control plots (Table 2). Similar patterns were observed with respect to the proportions of individuals within categories of floral rewards in each habitat, but, additionally, fragments faced a slight and statistically significant reduction on the proportion of individuals with BMFT flowers (0.19±0.41%) in contrast with control plots (1.25±1.56%) (Figure 2C).

As expected, fragments and control plots largely differed in terms of floral types considering the proportion of both species and individuals. Noticeable differences refer to significantly lower scores of species with flag and brush flowers, and higher scores of inconspicuous flowers in fragments in comparison with control plots (Table 2). Similar patterns were detected by analyzing the eight categories of floral types based on reward accessibility: (1) inconspicuous+open/dish flowers (easily accessible resource [sensu 66]), and (2) non-inconspicuous/open (concealed resource, at least some degree of hiddenness [sensu 66]). Under this approach, fragments showed a prevalence of species with inconspicuous/open type, which was significantly higher than in control sites. In terms of relative abundance of tree species within floral types categories, figures described fragments facing the same patterns observed to species regarding flag, inconspicuous (with even stronger differences), and brush flowers. Additionally, fragments showed lower proportions of individuals bearing camera and tube flowers in contrast with control areas (Figure 3A). Similarly, when observing proportions of individuals within categories of floral types according to reward accessibility, fragments had significant higher frequency of individuals with flowers of the inconspicuous/open type than control plots (Figure 3A), differences being yet more expressive than for species richness. Moreover, fragments revealed to be particularly impoverished in terms of tree species with nocturnal anthesis, showing a frequency more than two times lower (8.17±9.21%) than control plots (19.58±6.44%) (t = −3.211; d.f. = 18; P = 0.002). Difference was even more marked when the relative abundance of tree species with nocturnal anthesis is analyzed (4.93±6.67% in fragments vs. 21.18±11.41% in control plots) (t = −3.889; d.f. = 18; P = 0.001).

thumbnail

Figure 3. Effect of habitat fragmentation on floral types, sexual systems and reproductive systems.

Percentage of tree individuals within categories of floral types (A; N = 111 spp.), sexual system (B; N = 129 spp.), and reproductive system (C; N = 79 spp.) at 10 fragments and 10 control plots of an Atlantic forest landscape, northeastern Brazil. Frequencies represented by boxes that are significantly different are indicated with asterisks: *P<0.05; **P<0.01; ***P<0.001.

doi:10.1371/journal.pone.0000908.g003

Both habitats, fragments and control, were dominated by hermaphrodite species and showed similar scores for species within the five categories of sexual systems (Table 2). However, habitats were absolutely contrasting with respect to the frequency of individuals, as fragments were dominated by hermaphrodite individuals (61.05±15.33%), whereas non-hermaphrodite systems were prevalent (50.12±10.66%) among individuals of the control plots. Particularly expressive, as well, was the lower representation of monoecious individuals in the fragments–more than 12-times lower (1.72±1.84%) than control plots (24.51±13.92%) (Figure 3B). Fragments also had a slight but statistically significant decrease in the proportion of self-incompatible and overall obligatory outcrossing species (self-incompatible+dioecious; Table 2). In terms of the relative abundance of tree species within categories of reproductive system, fragments had significant lower scores of outcrossing individuals (87.82±6.84% vs. 95.40±8.54 in control plots) and highest frequency of self-compatible ones (11.59±6.56% vs. 3.90±8.03 in control sites) (Figure 3C).

Explanatory variables

GLMs did not reveal any significant influence of soil type on the proportion of traits in tree assemblages. Habitat was consistently the strongest explanatory variable for the proportion of tree species and individuals within categories of reproductive traits, explaining between 19.4% and 69.4% of their variation, influencing 38 categories (Table 3). GLMs also detected 10 categories of reproductive traits that were influenced by log-distance to edge (considering forest fragments and control plots), two categories influenced by log-fragment area, and eight affected by forest cover (considering forest fragments only) (Table 3). These three fragmentation-related variables explained between 20.7% and 68.6% of the variation on reproductive traits in forest fragments and control plots (Table 3). Additionally, ANOSIM revealed no significant correlation between soil type and level of taxonomic similarity between plots (R = 0.024; P = 0.54), but detected a stronger effect of habitat type (R = 0.95; P = 0.001). A Mantel test failed to uncover any spatial effects on the taxonomic similarity among plots (Rho = 0.155; P = 0.9).

thumbnail

Table 3. Scores from General Linear Models applied to the proportion of tree species and individuals within categories of reproductive traits (48 categories for species, 48 categories for individuals) in forest fragments (N = 10) and control plots (N = 10) in a fragmented landscape of Atlantic forest, northeastern Brazil.

doi:10.1371/journal.pone.0000908.t003

Functional diversity of reproductive traits

When using the number of reproductive categories (see Table 1) and the species richness per category, fragments were significantly less diversified (H′) with respect to pollination systems (−18.4%) and floral types (−12.65%) in comparison with control plots (Table 4). Simpson's values also evidenced fragments with significant lower functional diversity of pollination systems (Table 4). Differences were much more expressive, both biologically and statistically, when using number of categories (as equivalent of species) and number of individuals within categories for calculating diversity indices. In this case, fragments were significantly less diversified (H′ scores) not only in terms of pollination systems (−30.3%) and floral types (−23.6%), but they also presented significant lower functional diversity of floral sizes (−20.8%) in contrast with control plots (Table 4). Simpson's values also evidenced fragments with significant reduced functional diversity of pollination systems (−20.7%) and floral types (−19.62%) (Table 4). Based on Simpson's index, fragments were slightly more diversified than control plots in terms of floral rewards, however, when applying Bonferroni correction, values for floral rewards were not significantly different any more (Table 4).

thumbnail

Table 4. Functional diversity (mean±SD) of pollination systems, floral size, floral type and floral reward categories in tree assemblages of forest fragments (N = 10) and control plots (N = 10) in a fragmented landscape of Atlantic forest, northeastern Brazil.

doi:10.1371/journal.pone.0000908.t004

Discussion

Patterns and underlying mechanisms

Our findings suggest that habitat fragmentation promotes marked changes in both the presence and relative abundance of the reproductive traits of tree species, resulting in a reduced functional diversity of tree assemblages in forest fragments. Moreover, small forest patches in severely-fragmented landscapes may be strongly impoverished in terms of the number of species and individuals with particular pollination systems (e.g. pollination by bats, birds, non-flying mammals, Sphingids) and may be dominated by tree species pollinated by generalists. Finally, strategies that are more dependent on long-distance pollen movement and animal-mediated services, such as self-incompatibility, may be negatively affected. These statements are supported by the fact that the differences we found between fragments and control plots could not be explained by soil type or the relative spatial position of the plots in the landscape. Although the distribution of tropical trees has been found to be influenced by variation in soil types [52], [67], there is no evidence that this also influences the spatial distribution of ecological groups (based on reproductive traits, regeneration strategy, etc.) in terra firme forests [8], [68].

An increasing body of evidence has shown that as fragments become older, tree assemblages become drastically altered [12], [69][71]. Plant assemblages in small fragments (<10 ha) and forest edges are impoverished (scores of alpha diversity reduced by a half) and biased in taxonomic and ecological terms towards pioneer species. These patch-level findings suggest that fragmented landscapes tend to retain just a small subset of species from the original biota. Despite the recent findings on this topic, our study is one of the first to document a marked shift on the signature of tree assemblages inhabiting a fragmented landscape with respect to the frequency of reproductive-related traits and its functional diversity. Similar results were reported by Chazdon et al. [72] for tree assemblages in second-growth, logged, and old-growth forests in Costa Rica. They found lower relative abundance of mammal-pollinated trees in second-growth forests in comparison to old-growth ones, as well as a higher relative abundance of hermaphroditic trees in second-growth forests. In addition, Murcia [27] suggested fragmented forests tended to have an increased frequency of self-compatible hermaphrodites at the expense of other sexual systems. Our findings are consistent with these results, as well as recent ones indicating self-incompatible systems are more negatively affected than self-compatible ones following habitat loss and fragmentation [12], [25], [29].

Two fragmentation-related processes may be the principal mechanisms driving the changes in reproductive traits and functional diversity we observed: 1) the proliferation of pioneer species with a concomitant decline in the abundance of shade-tolerant trees and 2) depressed population sizes of animal pollinators, which over time led to changes in tree abundance in forest fragments. In tropical forests, myriad processes triggered by the creation of forest edges promote a proliferation of short-lived pioneers [8] and the local extirpation of shade-tolerant trees, including canopy and understory species [10], [19], emergent trees [73] and large-seeded trees [15], [34]. In our study site, pioneer species represent over 80% of all tree species and individuals recorded in the fragments, whereas they represent less 50% in core areas [19], [39]. Furthermore, recent surveys in this site have documented an outstanding predominance of pioneer species in seed rain [15] and seedling assemblages [34], [74] which suggests that pioneer dominance may represent a more pervasive, long-term feature of old and severally fragmented landscapes.

Assuming that pioneer plants are r-strategists and shade-tolerant (climax) ones are K- strategists [21], it is reasonable to expect that these two species groups differ in terms of reproductive traits, sexual systems, and reproductive systems. Some of our findings, such as higher scores of pollination by DSI and flowers with easily accessible resources (inconspicuous+open/dish flowers) in fragments, may simply reflect the dominance of pioneer trees in this habitat as these traits appear to be more frequent among pioneers (65% of the DSI-pollinated species and over 68% of the species with inconspicuous/open/dish flowers are pioneers). On the other hand, a trait such as pollination by bats that was significantly more frequent in control plots (richness and abundance) is also positively associated with a subset of shade-tolerant species–75% of the bat-pollinated species are shade-tolerant (e.g. Bauhinia, Hymenaea-Fabaceae; Manilkara-Sapotaceae; Quararibea-Malvaceae sensu APG II [75]). Because the pioneer species recorded in the fragments - including both short- and long-lived pioneers - belong to 16 orders and eight superordinal clades (sensu APG II [75]), the patterns documented here cannot be explained by phylogenetic clustering among pioneers. Even pioneer species that were recorded exclusively in forest fragments belong to four families in four orders and three superordinal clades. Unfortunately, because of the large number of categories for each reproductive trait and the low number of tree species within each category, it was not possible to properly test trait-associated differences between pioneer and shade-tolerant tree species.

In tropical forests, 98–99% of the flowering plant species (and 97.5% of the trees) rely on biotic vectors such as insects and vertebrates for successful pollination [76], [77], and it has been broadly assumed that plant-pollinator interactions are largely detrimentally affected by habitat loss and fragmentation [26][29], [78][81]. Some of the changes we documented in our fragments are therefore expected, particularly the lack or reduced occurrence of some pollination systems [27], [28], [82]. For instance, fragmented habitats may support less pollinators than continuous habitats due to limited resource availability for pollinators (area-related effects on animal populations). In turn, plants can have a depressed reproductive output as consequence of changes in pollinator diversity, composition, or behavior [25], [28], i.e., reproductive impairment driven by pollination limitation [sensu 29]. Studies on pollinator diversity carried out in our landscape have documented a decreased diversity of nectarivorous bats [83] and hawkmoths [84] in small fragments. However, empirical evidence to determine which pollination-related traits and plant-pollinator mutualisms are particularly susceptible to habitat disturbance is still scarce [28]. Our results suggest that the reduced number of tree species and individuals pollinated by bats and Sphingids in fragments and the absence of fly-, bird-, and non-flying-mammal-pollinated trees, together with the changes in floral traits and sexual systems, may be a higher order effect promoted by habitat fragmentation.

Implications of reduced functional diversity

The reduced reproductive functional diversity documented in our study landscape's fragments resulted primarily from the lack or skewed representation of some pollination systems, floral types, and floral size categories in terms of both species and individual (see Table 3). In other words, tree assemblages in this habitat appear to carry a narrower range of floral traits and pollination systems in comparison to patches of forest interior, particularly for pollinators such as mammals and hawkmoths (reduced support capacity). Regardless the underlying mechanism, this narrow range may (1) promote the collapse of pollinator populations; (2) restrict the ecological range of plant and animal groups able to colonize remaining patches of forest or even turn fragments into sink habitats for both plants and their pollinators; and (3) alter the course of natural regeneration or the dynamics of forest fragments toward the establishment of impoverished assemblages in terns of species richness, ecological composition and trophic structure.

Unfortunately, few studies have addressed shifts on the diversity of plant reproductive traits in human-disturbed habitats, especially those traits associated with plant-pollinator interactions [1], [72]. Studies linking these shifts to functional diversity are even more scarce [1], [2], despite the fact that pollination processes influence biodiversity maintenance and ecosystem functioning. Fontaine et al. [2], for example, argued that even simple structured plant-pollinator communities may have their persistence threatened due to reduced functional diversity, thereby suggesting that functional diversity of pollination networks is critical to avoid biodiversity loss.

In summary, it is reasonable to propose as a working hypothesis that the persistence of biodiversity and consequently the long-term conservation value of isolated tropical forest fragments may be negatively affected by reduced functional diversity to such an extent yet not anticipated by conservation biologists. Collectively, the proliferation of pioneer species, extirpation of shade-tolerant trees, and reduced functional diversity have the potential to disrupt some trophic interactions [e.g. 85]; even landscapes such as ours that were fragmented long ago and are dominated by pioneers may face future biodiversity loss. We believe it would be beneficial for future research to 1) validate and assess the generality of both the patterns and the underlying mechanisms observed here and 2) address more ecosystem level effects driven by reduced functional diversity in fragmented landscapes, such as changes in biodiversity persistence, primary productivity, nutrient cycling, succession, and ecosystem resilience.

Supporting Information

Table S1.

Species studied and their abundance in forest fragments (N = 10) and control plots (N = 10) in a fragmented landscape of Atlantic forest, northeastern Brazil.

doi:10.1371/journal.pone.0000908.s001

(0.55 MB DOC)

Acknowledgments

We are very grateful to Alexandre Grillo and Marcondes Oliveira (UFPE-Brazil) for their immeasurable cooperation and for permitting us to use data from their botanical surveys; Luís Antônio Bezerra and Clodoaldo Bakker for permitting this research on the private property of the Usina Serra Grande; Carlos Peres (University of East Anglia-England) for discussions and help with analyses; André Santos (UFPE-Brazil) and Antônio Aguiar-Neto (UF-USA) for discussions and help in several phases of this study; Jérôme Chave (Academic Editor) and an anonymous reviewer for valuable suggestions on the manuscript.

Author Contributions

Conceived and designed the experiments: AL MT. Performed the experiments: AL LG. Analyzed the data: AL LG MT EB. Wrote the paper: AL LG MT EB.

References

  1. 1. Mayfield MM, Boni ME, Daily GC, Ackerly D (2005) Species and functional diversity of native and human-dominated plant communities. Ecology 86: 2365–2372.
  2. 2. Fontaine C, Dajoz I, Meriguet J, Loreau M (2006) Functional diversity of plant-pollinator interaction webs enhances the persistence of plant communities. PLoS Biol 4: 129–135.
  3. 3. Tilman D, Knops J, Wedin D, Reich P, Ritchie M, et al. (1997) The influence of functional diversity and composition on ecosystem processes. Science 277: 1300–1302.
  4. 4. Hooper DU, Vitousek PM (1997) The effects of plant composition and diversity on ecosystem processes. Science 277: 1302–1305.
  5. 5. Tilman D, Downing JA (1994) Biodiversity and stability in grasslands. Nature 367: 363–365.
  6. 6. Prieur-Richard AH, Lavorel S (2000) Invasions: the perspective of diverse plant communities. Austral Ecol 25: 1–7.
  7. 7. Mason NWH, MacGillivray K, Steel JB, Wilson JB (2003) An index of functional diversity. J Veg Sci 14: 571–578.
  8. 8. Laurance WF, Nascimento HEM, Laurance SG, Andrade AC, Fearnside PM, et al. (2006) Rain forest fragmentation and the proliferation of successional trees. Ecology 87: 469–482.
  9. 9. Tabarelli M, Silva MJC, Gascon C (2004) Forest fragmentation, synergisms and the impoverishment of neotropical forests. Biodivers Conserv 13: 1419–1425.
  10. 10. Tabarelli M, Mantovani W, Peres CA (1999) Effects of habitat fragmentation on plant guild structure in the montane Atlantic forest of southeastern Brazil. Biol Conserv 91: 119–127.
  11. 11. Metzger JP (2000) Tree functional group richness and landscape structure in a Brazilian tropical fragmented landscape. Ecol Appl 10: 1147–1161.
  12. 12. Laurance WF, Nascimento HEM, Laurance SG, Andrade A, Ribeiro J, et al. (2006) Rapid decay of tree-community composition in Amazonian forest fragments. Proc Natl Acad Sci U S A 103: 19010–19014.
  13. 13. Laurance WF, Ferreira LV, Rankin-De Merona JM, Laurance SG, Hutchings RW, et al. (1998) Effects of forest fragmentation on recruitment patterns in Amazonian tree communities. Conserv Biol 12: 460–464.
  14. 14. Terborgh J, Lopez L, Nunez P, Rao M, Shahabuddin G, et al. (2001) Ecological meltdown in predator-free forest fragments. Science 294: 1923–1926.
  15. 15. Melo FPL, Dirzo R, Tabarelli M (2006) Biased seed rain in forest edges: Evidence from the Brazilian Atlantic forest. Biol Conserv 132: 50–60.
  16. 16. Galetti M, Donatti CI, Pires AS, Guimaraes PR, Jordano P (2006) Seed survival and dispersal of an endemic Atlantic forest palm: the combined effects of defaunation and forest fragmentation. Bot J Linnean Soc 151: 141–149.
  17. 17. Silva JMC, Tabarelli M (2000) Tree species impoverishment and the future flora of the Atlantic forest of northeast Brazil. Nature 404: 72–74.
  18. 18. Cordeiro NJ, Howe HF (2001) Low recruitment of trees dispersed by animals in African forest fragments. Conserv Biol 15: 1733–1741.
  19. 19. Oliveira MA, Grillo AS, Tabarelli M (2004) Forest edge in the Brazilian Atlantic forest: drastic changes in tree species assemblages. Oryx 38: 389–394.
  20. 20. Richards PW (1996) The tropical rain forest - an ecological study. Cambridge: Cambridge University Press.
  21. 21. Turner IM (2001) The ecology of trees in the tropical rain forest. Cambridge: Cambridge University Press.
  22. 22. Whitmore TC (1990) An introduction to tropical rain forests. Oxford: Clarendon Press.
  23. 23. Buchmann SL, Nabhan GP (1996) The forgotten pollinators. Whashington D.C.: Island Press.
  24. 24. Kevan PG (1999) Pollinators as bioindicators of the state of the environment: species, activity and diversity. Agric Ecosyst Environ 74: 373–393.
  25. 25. Vamosi JC, Knight TM, Steets JA, Mazer SJ, Burd M, et al. (2006) Pollination decays in biodiversity hotspots. Proc Natl Acad Sci U S A 103: 956–961.
  26. 26. Rathcke BJ, Jules ES (1993) Habitat fragmentation and plant pollinator interactions. Curr Sci 65: 273–277.
  27. 27. Murcia C (1996) Forest fragmentation and the pollination of neotropical plants. In: Schelhas J, Greenberg R, editors. Forest Patches in tropical landscapes. Washington, DC: Island Press. pp. 19–36.
  28. 28. Harris LF, Johnson SD (2004) The consequences of habitat fragmentation for plant-pollinator mutualism. Int J Trop Insect Sci 24: 29–43.
  29. 29. Aguilar R, Ashworth L, Galetto L, Aizen MA (2006) Plant reproductive susceptibility to habitat fragmentation: review and synthesis through a meta-analysis. Ecol Lett 9: 968–980.
  30. 30. Wilcock C, Neiland R (2002) Pollination failure in plants: why it happens and when it matters. Trends Plant Sci 7: 270–277.
  31. 31. Chambers JQ, Higuchi N, Schimel JP (1998) Ancient trees in Amazonia. Nature 391: 135–136.
  32. 32. Galindo-Leal C, Câmara IG (2005) Status do hotspot Mata Atlântica: uma síntese. In: Galindo-Leal C, Câmara IG, editors. Mata Atlântica: Biodiversidade, Ameaças e Perspectivas. Belo Horizonte: Fundação SOS Mata Atlântica & Conservação Internacional. pp. 3–11.
  33. 33. Ranta P, Blom T, Niemela J, Joensuu E, Siitonen M (1998) The fragmented Atlantic rain forest of Brazil: size, shape and distribution of forest fragments. Biodivers Conserv 7: 385–403.
  34. 34. Melo FPL, Lemire D, Tabarelli M (2007) Extirpation of large-seeded seedlings from the edge of a large Brazilian Atlantic forest fragment. Écoscience 14: 124–129.
  35. 35. Prance GT (1982) Forest refuges: evidence from woody angiosperms. In: Prance GT, editor. Biological diversification in the tropics. New York: Columbia University Press. pp. 137–158.
  36. 36. IBGE (1985) Atlas nacional do Brasil: Região Nordeste. Rio de Janeiro: Instituto Brasileiro de Geografia e Estatística.
  37. 37. Veloso HP, Rangel-Filho ALR, Lima JCA (1991) Classificação da vegetação brasileira adaptada a um sistema universal. IBGE.
  38. 38. Grillo A, Oliveira MA, Tabarelli M (2006) Árvores. In: Pôrto KC, Almeida-Cortez JS, Tabarelli M, editors. Diversidade biológica e conservação da Floresta Atlântica ao norte do Rio São Francisco. 1 ed. Brasília: Ministério do Meio Ambiente. pp. 191–216.
  39. 39. Grillo A (2005) As implicações da fragmentação e da perda de habitat sobre a assembléia de árvores na Floresta Atlântica ao norte do rio São Francisco [PhD Thesis]. Recife: Universidade Federal de Pernambuco.
  40. 40. Coimbra-Filho F, Câmara IG (1996) Os limites originais do Bioma Mata Atlântica na Região Nordeste do Brasil. Rio de Janeiro: Fundação Brasileira para Conservação da Natureza - FBCN.
  41. 41. Harper KA, Macdonald SE, Burton PJ, Chen JQ, Brosofske KD, et al. (2005) Edge influence on forest structure and composition in fragmented landscapes. Conserv Biol 19: 768–782.
  42. 42. Chiarello AG (1999) Effects of fragmentation of the Atlantic forest on mammal communities in south-eastern Brazil. Biol Conserv 89: 71–82.
  43. 43. Tabarelli M, Peres CA (2002) Abiotic and vertebrate seed dispersal in the Brazilian Atlantic forest: implications for forest regeneration. Biol Conserv 106: 165–176.
  44. 44. Bruna EM, Kress WJ (2002) Habitat fragmentation and the demographic structure of an Amazonian understory herb (Heliconia acuminata). Conserv Biol 16: 1256–1266.
  45. 45. Pôrto KC, Tabarelli M, Almeida-Cortez JS (2006) Diversidade biológica e conservação da Floresta Altântica ao norte do Rio São Francisco. Brasília: Ministério do Meio Ambiente.
  46. 46. Pimentel DS, Tabarelli M (2004) Seed dispersal of the palm Attalea oleifera in a remnant of the Brazilian Atlantic Forest. Biotropica 36: 74–84.
  47. 47. Ribeiro JELS, Hopkins MJG, Vicentini A, Sothers CA, Costa MAS, et al. (1999) Flora da Reserva Ducke: guia de identificação de uma floresta de terra-firme na Amazônia Central. Manaus: INPA-DFID.
  48. 48. Backes P, Irgang B (2004) Mata Atlântica: As árvores e a paisagem. Porto Alegre: Editora Paisagem do Sul.
  49. 49. Lorenzi H (2002) Árvores Brasileiras: manual de identificação e cultivo de plantas arbóreas nativas do Brasil (vol.1). Nova Odessa: Instituto Plantarum.
  50. 50. Lorenzi H (2002) Árvores Brasileiras: manual de identificação e cultivo de plantas arbóreas nativas do Brasil (vol.2). Nova Odessa: Instituto Plantarum.
  51. 51. Machado IC, Lopes AV (2002) A Polinização em ecossistemas de Pernambuco: uma revisão do estado atual do conhecimento. In: Tabarelli M, Silva JMC, editors. Diagnóstico da Biodiversidade de Pernambuco. Recife: Secretaria de Ciência Tecnologia e Meio-Ambiente, Fundação Joaquim Nabuco, Editora Massangana. pp. 583–596.
  52. 52. ter Steege H, Pitman NCA, Phillips OL, Chave J, Sabatier D, et al. (2006) Continental-scale patterns of canopy tree composition and function across Amazonia. Nature 443: 444–447.
  53. 53. Gorresen PM, Willig MR (2004) Landscape responses of bats to habitat fragmentation in Atlantic forest of Paraguay. J Mammal 85: 688–697.
  54. 54. Krebs C (1989) Ecological metodology. New York: Harper & Hall.
  55. 55. Petchey OL, Gaston KJ (2002) Extinction and the loss of functional diversity. Proc R Soc Lond Ser B-Biol Sci 269: 1721–1727.
  56. 56. Petchey OL, Hector A, Gaston KJ (2004) How do different measures of functional diversity perform? Ecology 85: 847–857.
  57. 57. Ricotta C (2005) A note on functional diversity measures. Basic Appl Ecol 6: 479–486.
  58. 58. Sokal RR, Rohlf FG (1995) Biometry. New York: W.H. Freeman and Company.
  59. 59. Clarke KR, Gorley RN (2001) PRIMER v5: User manual/tutorial. Playmouth: PRIMER-E Ltd.
  60. 60. Condit R, Pitman N, Leigh EG, Chave J, Terborgh J, et al. (2002) Beta-diversity in tropical forest trees. Science 295: 666–669.
  61. 61. Jones MM, Tuomisto H, Clark DB, Olivas P (2006) Effects of mesoscale environmental heterogeneity and dispersal limitation on floristic variation in rain forest ferns. J Ecol 94: 181–195.
  62. 62. Wilkinson L (1996) SYSTAT. Version 6.0. Chicago: SPSS.
  63. 63. Anon (2001) Primer 5 for Windows. Plymouth: PRIMER-E.
  64. 64. McCune B, Mefford MJ (1999) PC-ORD. Multivariate Analysis of Ecological Data. Version 4.36. Gleneden Beach: MjM Software.
  65. 65. Kang H, Bawa KS (2003) Effects of successional status, habit, sexual systems, and pollinators on flowering patterns in tropical rain forest trees. Am J Bot 90: 865–876.
  66. 66. Faegri K, Pijl L van der (1979) The principles of pollination ecology. Oxford: Pergamon Press.
  67. 67. Clark DB, Clark DA, Read JM (1998) Edaphic variation and the mesoscale distribution of tree species in a neotropical rain forest. J Ecol 86: 101–112.
  68. 68. Michalski F, Nishi I, Peres CA (2007) Disturbance-mediated drift in tree functional groups in Amazonian forest fragments. Biotropica. doi:10.1111/j.1744-7429.2007.00318.x.
  69. 69. Laurance WF, Lovejoy TE, Vasconcelos HL, Bruna EM, Didham RK, et al. (2002) Ecosystem decay of Amazonian forest fragments: A 22-year investigation. Conserv Biol 16: 605–618.
  70. 70. Laurance WF (2001) Fragmentation and plant communities: synthesis and implications for landscape management. In: Bierregaard RO Jr, Gascon C, Lovejoy TE, Mesquita RCG, editors. Lessons from Amazonia: the Ecology and Conservation of a Fragmented Forest. New Haven: New Haven: Yale University Press. pp. 158–168.
  71. 71. Laurance WF, Peres CA (2006) Emerging threats to tropical forests. Chicago: The University of Chicago Press.
  72. 72. Chazdon RL, Careaga S, Webb C, Vargas O (2003) Community and phylogenetic structure of reproductive traits of woody species in wet tropical forests. Ecol Monogr 73: 331–348.
  73. 73. Laurance WF, Delamonica P, Laurance SG, Vasconcelos HL, Lovejoy TE (2000) Conservation - Rainforest fragmentation kills big trees. Nature 404: 836–836.
  74. 74. Costa JBP (2007) Efeitos da fragmentação sobre a assembléia de plântulas em um trecho da floresta Atlântica Nordestina [Master thesis]. Recife: Universidade Federal de Pernambuco.
  75. 75. APG ( = Angiosperm Phylogeny Group) II (2003) An update of the Angiosperm Phylogeny Group classification for the orders and families of flowering plants: APG II. Bot J Linnean Soc 141: 399–436.
  76. 76. Bawa KS (1990) Plant-pollinator interactions in tropical rain-forests. Annu Rev Ecol Syst 21: 399–422.
  77. 77. Bawa KS, Bullock SH, Perry DR, Coville RE, Grayum MH (1985) Reproductive biology of tropical lowland rain forest trees. II. Pollination systems. Am J Bot 72: 346–356.
  78. 78. Ghazoul J (2005) Pollen and seed dispersal among dispersed plants. Biol Rev 80: 413–443.
  79. 79. Kearns CA, Inouye DW, Waser NM (1998) Endangered mutualisms: The conservation of plant-pollinator interactions. Ann Rev Ecol Syst 29: 83–112.
  80. 80. Aizen MA, Feinsinger P (1994) Forest fragmentation, pollination, and plant reproduction in a chaco dry forest, Argentina. Ecology 75: 330–351.
  81. 81. Renner SS (1998) Effects of habitat fragmentation on plant pollinator interactions in the tropics. In: Newbery DM, Prins HHT, Brown ND, editors. Dynamics of Tropical Communities. London: Blackwell Scientific Publishers. pp. 339–360.
  82. 82. Hobbs RJ, Yates CJ (2003) Impacts of ecosystem fragmentation on plant populations: generalising the idiosyncratic. Aust J Bot 51: 471–488.
  83. 83. Sá-Neto RJ (2003) Efeito da fragmentação na comunidade de morcegos (Mammalia: Chiroptera) em remanescentes de floresta Atlântica, Usina Serra Grande - Alagoas [Master thesis]. Recife: Universidade Federal de Pernambuco.
  84. 84. Lopes AV, Medeiros PC, Aguiar AV, Machado IC (2006) Esfingídeos. In: Pôrto KC, Almeida-Cortez JS, Tabarelli M, editors. Diversidade biológica e conservação da floresta Atlântica ao norte do Rio São Francisco. Brasília: Ministério do Meio Ambiente. pp. 229–235.
  85. 85. Wirth R, Meyer ST, Leal IR, Tabarelli M (2007) Plant–herbivore interactions at the forest edge. Progr Bot 69: 423–448.
  86. 86. Endress PK (1994) Diversity and evolutionary biology of tropical flowers. Cambridge: Cambridge University Press.
  87. 87. Proctor M, Yeo P, Lack A (1996) The natural history of pollination. London: Harper Collins Publishers.
  88. 88. Machado IC, Lopes AV (2004) Floral traits and pollination systems in the Caatinga, a Brazilian tropical dry forest. Ann Bot 94: 365–376.
  89. 89. Richards AJ (1997) Plant breeding systems. London: Chapman & Hall.
  90. 90. Oliveira PE, Gibbs PE (2000) Reproductive biology of woody plants in a cerrado community of Central Brazil. Flora 195: 311–329.