Introduction

Brazil ranks among the countries with the highest plant species richness in the world, and enjoys considerable diversity of forest ecosystems due to its wide variation in climatic conditions and soils (^{Gentry 1982}; Leitão-Filho 1987). In northeastern Brazil, forests occur from the transition zone of the semiarid nucleus in which the caatinga vegetation prevails, to the more humid areas at the borders of the region. However, forests are also observed in semiarid areas of greater altitude, on crystalline basement rocks, or even on sedimentary soils (^{Couto et al.2011}). The forests within the transition zone of the semiarid nucleus include the seasonal forests on the east side of Chapada Diamantina National Park, which are considered part of the Atlantic Forest (sensulato; ^{Oliveira-Filho & Fontes 2000}) and cover the largest forested section of the Chapada Diamantina. These forests are the most degraded in the region, particularly due to timber extraction and agricultural and cattle farming activities (^{Funch et al. 2005}). These anthropogenic factors have led to the destruction and loss of biodiversity of such vegetation, even though the natural richness of these regions has not been properly investigated (Funch et al. 2008; ^{Ribeiro-Filho et al. 2009}; Couto et al.2011).

Linear infrastructure work, such as the construction of roads and power lines, alter natural landscapes, and often require the deforestation of corridors inside forests, which can create linear borders (^{Murcia 1995}). These corridors magnify edge effects because their perimeter is twice as long as that of edges that delimit a conventional forest fragment, increasing the impact caused by the degraded area in the surroundings and intensifying the microclimate changes near edges (^{Goosem 1997}; ^{Pohlman et al. 2007}). Such changes affect the structure of fragments (Ribeiro et al. 2009), depending on how much time has passed since the creation of such edges. For instance, a recently opened linear edge may be structurally homogeneous, or quite similar to the inner fragment. However, with time, other changes may occurin association with edge effects and facilitate the emergence of more heterogeneous communities (^{Rodrigues & Nascimento 2006}; ^{Holanda et al. 2010}).

Edge effects play a key role in understanding the ways landscape structures influence habitat quality (^{Zanuto et al.2007}). Accordingly, these effects have been a consistent research topic for decades (Pohlman et al. 2007; ^{Prieto et al. 2014}). Studies have documented matching patterns of biodiversity deterioration in fragmented environments. These patterns are characterized by the elimination of arboreal species of specific functional groups and the impoverishment of tropical forests (^{Tabarelli et al. 2004}; ^{Lopes et al. 2009}). These factors point to the urgent need for more studies on restoration and conservation efforts.

Several studies have investigated the influence of edge effects in neotropical forests (^{Burkey 1993}; ^{Laurance et al. 1998}; ^{Olander et al. 1998}; ^{Mesquita et al. 1999}; Paciencia & Prado 2005; Laurance et al. 2009; Ribeiro et al. 2009; ^{Reznik et al. 2012}) and reported changes in composition, diversity, and dynamics of species. Nevertheless, other research suggests that edge effects may actually not have serious negative outcomes, exerting positive influences on some variables, or even no influence at all (Oliveira et al. 2013).

The present study evaluates the impact of a linear edge on the heterogeneity of the arbustive-arboreal community in a patch of submontane seasonal forest in Chapada Diamantina, Bahia, Brazil. The aim was to detect composition, abundance patterns, and successional distribution stages inside and at the edge of the forest. Since the linear vegetation suppression by periodic pruning results in two adjacent edges, we expected neutral effects or increases in the richness and heterogeneity at these two adjacent linear edges, in contrast with the expectation for a simple edge of a larger fragment (Paciencia & Prado 2005; Ribeiro et al. 2009; Rezniket al. 2012).

Material and methods

Study area

The forest fragment studied is located at 12°28'31'S, 41°23'14'W in the Chapada Diamantina National Park, Bahia, Brazil. The topography of the region is slightly wavy, with altitudes varying between 400 m and 600 m. Its soil comprised yellowish-red latosols of a clayish consistency that are poor in organic matter (Couto et al. 2011). Due to the altitudinal conditions, the climate in the region is classified as mesothermal type Cwb. The rainy season occurs in summer (December-April), peak rainfall is recorded in March-April, and the dry season occurs in winter. Rainfall varies between 35 mm (July and August) and184 mm (December). Mean monthly temperatures oscillate around 18°C in winter (April-September) and exceed 22°C in hotter months (October-February; Funch et al. 2002). These conditions have led to the classification of the vegetation in the area as seasonal submontane forest (^{IBGE 2012}).

A major highway bisects the region, and two perpendicular corridors that measure approximately 10 m in width were deforested to allow the passage of high-voltage electric power lines. Both the road and power lines were constructed in the 60's. The vegetation is characterized by emergent trees up to 26 m in height, including the species Protium heptaphyllum (Aubl.) Marchand, Maprounea guianensis Aubl., Hirtella glandulosa Spreng., and Aspidosperma discolor A. DC. Stratification is difficult to detect, but the understory is well lit and formed mainly by young individuals of species that are established in the higher strata and species of the families Rubiaceae, Melastomataceae, and Poaceae. Abundant climbing species are also seen, and epiphytic and hemiparasitic species are rare (Couto et al. 2011).

Phytosociological survey

The fieldwork was conducted between May 2006 and August 2007. Given that edge effects may stretch for approximately 100 m (Laurance et al. 1998), 30 quadrats measuring 10 m × 10 m (0.3 ha) were established in a forest fragment measuring 6.5 ha using an Etrex Summit GPS receiver for randomization of the quadrats location in a forest, compass, and measuring tape. Fifteen quadrats were defined along the forest edge (within the 0-100 m range), and the other quadrats were established in the inner fragment (more than 150 m away from the power lines). Moreover, care was taken to establish quadrats 150 m away from the highway, so as to cover the edge effect of the area deforested for the power lines alone.

In each quadrat, circumferences (C) of shrub and arboreal individuals were measured using a measuring tape. Only individuals with C at breast height (diameter at 1.3 m from ground level) equal to or larger than 15.7 cm [Diameter at breast height (DBH) = 5.0 cm] were sampled. DBH was calculated using the formula D = C.π⁻¹.Height was estimated visually, using a 2m long stick. All living individuals sampled were tagged to facilitate periodic monitoring.

Samples of botanical material were collected and identified using the collections stored in the Herbarium HUEFS and by consulting specialists and the literature; taxonomic classification was based on the ^{APG III (2009)}system. Arboreal species were classified according to successional category as pioneers (P), early secondary (Es), late secondary (Ls), and not defined (nd; ^{Gandolfi et al. 1995}). This classification was carried out using information about the biology of the species reported in the literature (Gandolfi et al. 1995; ^{Nunes 2003}; ^{Mariano Neto 2004}), supplemented with the authors'knowledge of the subject.

The structure of the forest was described according to the following phytosociological parameters: relative density (RD), relative frequency (RF), and relative dominance (RDo) as well as importance value (IV) and cover value (CV; ^{Mueller-Dombois & Ellenberg 1974}), calculated using the software FITOPAC 1.6 (^{Shepherd 2006}). The Shannon-Wienerdiversity index (H'; log_n base) and Pielou's evenness index (J') were calculated for edge and inner fragments (^{Brower & Zar 1998}; Nunes et al.2003). Values of H' were compared using Hutcheson's t-test (Diversity t-test) calculated with the software PAST (^{Hammer et al. 2001}).

Normality of data was evaluated using the Shapiro-Wilk test (p > 0.05). The variables height, diameter, area at ground level, density, floristic composition, and successional categories (in proportion of species and individuals) were compared using a one-way ANOVA. Structural variables were expressed as box plots using the software BioEstat 5.3 (^{Ayres et al. 2007}). Height and DBH were analyzed based on the distribution of categories in each quadrat.

Results

In total, 456 arboreal individuals were sampled (1520 individuals.ha^-1), belonging to 33 families, 47 genera, and 68 species (four of which were not identified; Tab.1). The inner fragment and edge differed in their most abundant species as well as highest IV and CV (Tab.1). The species with the highest IV at the edge were Tapirira guianensis, Micropholis gardneriana, Hirtella glandulosa, Maprounea guianensis, and Himatanthus bracteatus. These formed the largest and most common populations in edges, together with Protium heptaphyllum and Pogonophora schomburgkiana. Such a pattern was also evident in the sum of the density these species (43.6%). Inside the forest, Eschweilera tetrapetala, Copaifera langsdorffii, Tapirira guianensis, Ocotea nitida, and Pouteria ramiflora predominated, which, except for T. guianensis, are late successional species. Eschweilera tetrapetala was prominent not only inside the forest, but also in the whole environment studied, with 326.7 individuals.ha^-1 presenting IV higher than 40%.

When the edge and inner fragment were compared (Tab.2), even though 26 species were shared by these two environments, the floristic composition varied significantly (F = 5.221; p = 0.0239), as did diversity (Hutcheson's t = 2.40; p = 0.018). However, no variation was observed in proportions of successional categories (F_species = 0.152; p = 0.7135 and F_individuals = 0.039; p = 0.8501). A significant difference was observed between the edge and inner fragment in height but only in the class of trees between 10 m and 15 m in height (F = 6.703; p = 0.0158). However, these environments did not differ in DBH (F = 0.17; p = 0.69), D (F = 1.137; p = 0.296), and area at ground level (F = 1.45; p = 0.242), in spite of the presence of outliers (Fig. 1). In the two environments, the DBH distribution curve (Fig. 2) indicates that a higher number of individuals lie in the classes denoting larger diameters. Extreme diameter values were recorded in the edge for Hymenolobium janeirense var. stipulatum (63.4 cm), Aspidosperma discolor (48.4 cm), and Pouteria ramiflora (38.2 cm) and in the inner fragment for Copaifera langsdorffii (63.4 cm), Maytenus robusta (54.1 cm), and undetermined individual 3 (47.8 cm), which are the outliers shown in the box plot (Fig. 1). Most arboreal individuals did not exceed 10 m in height (Fig. 3). Extreme values were recorded for Hymenolobium janeirense var. stipulatum (26 m) and Hirtella glandulosa (25 m) in the edge and Copaifera langsdorffii (22 m) and Maprounea guianensis (20 m) in the inner fragment.

Figure 1. Box plot comparing height (A), diameter (B), density (C),and area at ground level (D) of trees in a linear edge and inside a forest fragment of a submontane seasonal forest in Chapada Diamantina National Park, Bahia, Brazil. Horizontal lines represent the arithmetic means (middle line) ± standard deviation (upper and lower lines); outer horizontal lines represent the minimum and maximum values; o = outliers; ha = hectares; m = meters. 

Figure 2. Distribution of the number of individuals according to diameter at breast height (DBH) in a linear edge and inside a forest fragment of a submontane seasonal forest in Chapada Diamantina National Park, Bahia, Brazil 

Figure 3. Distribution of the number of individuals according to height in a linear edge and inside a forest fragment of a submontane seasonal forest in Chapada Diamantina National Park, Bahia, Brazil. 

Discussion

The present study showed the greatest impact of linear edge on floristic composition, diversity, and abundance of species, whereas little interference was observed in the structure of the tree community. Phytosociological parameters and proportions of successional categories did not vary between the edge and the inner fragment, showing that the structure of the arboreal community in these environments is not consistently affected. Heterogeneity was also observed based on the number of pioneer and early secondary species at the edge and late secondary species inside the forest. In spite of the floristic peculiarities, the edge and inner fragment were similar in the families presenting the highest species richness, such as Lauraceae, Myrtaceae, Fabaceae, Sapotaceae, and Euphorbiaceae, which are important groups in the Brazilian as well as global flora (Gentry 1998; Oliveira Filho et al. 2004; ^{Saravanan et al. 2013}).

Higher heterogeneity in edges (H') has also been reported for other linear edges (Olander et al.1998; Laurance et al.2002; Prieto et al. 2014) as well as for other kinds of edges (Nunes et al. 2003). Disturbances and edge effects may strongly influence species diversity. Often, higher diversity in edge areas is explained by the presence of typical inner fragment species as well as plant species with different light demands (Nunes et al. 2003; Tabarelli et al. 2012; ^{Arroyo-Rodríguez et al. 2013}).

The H' value observed in the edge is in agreement with previous results reported for the Chapada Diamantina region, which vary between 3.4 and 3.6 nats/individual (^{Stradmann 2000}; Souza 2007; Ribeiro-Filho et al. 2009), and forests in northeastern and southeastern Brazil, which range from 3.0 to 4.3 nats/individual (^{Sambuichi 2002}; Mariano Neto 2004; Oliveira Filho et al. 2004; ^{Catharino et al.2006}). Despite the high species diversity, the inner fragment presented H' values below those reported for forests in other studies. Here, the most important species in this environment was Eschweilera tetrapetala (called sapucaia), which is an endemic species in submontane seasonal forests of the region.

The analysis of the successional stages of species showed that the edge had a higher proportion of early secondary species, which are less tolerant to shade. Conversely, the inner fragment had a higher number of late secondary species, which germinate and manage to grow under the low light conditions of the understory and reach maturity in this stratum, at the canopy, or in some cases, become emergent trees (Nunes et al. 2003).

According to ^{Chazdon (2012)}, fragments still undergo several changes, including changes due to edge effects, even 15-50 years after an environmental disturbance, thus, the forest studied is susceptible to edge effects. Furthermore, the linear edge observed in this study is preserved by periodic pruning, which is required for the maintenance of linear infrastructure. These procedures disrupt the regeneration process and sustain the edge effect. Ultimately, shade-tolerant species and those that produce large fruit and seeds may be substituted by shade-intolerant, small-fruited, small-seeded species typical of secondary forests (Laurance et al. 2006). Some pioneer species seem to benefit from habitat fragmentation, which have increased in abundance in small fragments (Arroyo-Rodríguez et al. 2013) and in edge zones (Oliveira et al. 2004), while abundance of the more shade-tolerant, edge-sensitive species has decreased (Chazdon 2012).

This explains the influence of linear edges in plant communities, in light of the evidence that successional processes vary with the degree of fragmentation and with the implementation of man-made edges (Rodrigues & Nascimento 2006). In a study that compared the edges and an inner fragment surrounded by sugarcane plantations in the state of Alagoas, northeastern Brazil, Oliveira et al. (2004) observed significant differences in the proportion of species in different successional stages. These authors also reported the prevalence of pioneer species at the edge, while shade-tolerant species predominated within the fragment surveyed. In the present study, the edge presented elevated mean density values and smaller mean DBH compared with the inner fragment, though with no statistically significant differences. These differences point to the existence of a relatively young community at the edge, where a higher density of trees with small DBH is observed. On the other hand, inside the fragment, thicker trees were more common, which suggests a more advanced regeneration stage (Nunes et al. 2003).

As a rule, edges exhibit higher density values than preserved inner fragments (Laurance et al. 2002). This may result from the growth of individuals (for instance, seedlings or propagules) as well as the establishment of species linked with less advanced successional stages (Rodrigues & Nascimento 2006). Assuming that this is true, edges may increase ground-level area because of greater light incidence (Murcia 1995). Moreover, here we observed that taller emergent trees were more common in the edge, similar to what was reported by Ribeiro et al. (2009), in a comparison between an inner fragment environment and natural and anthropogenic borders in an Atlantic Forest fragment in the state of Rio de Janeiro, Brazil.

The higher number of individuals with small DBH, both in the edge and in the inner fragment, and the scarcity of larger DBH trees indicate that the forest physiognomy in the environment studied develops into more advanced successional stages, in light of the great number of young individuals that may succeed over older individuals (Lopes et al. 2002). This dynamic promotes this forest's maintenance, indicating that it is undergoing a regeneration phase triggered by previous disturbances.

Thus, the present study shows that linear edge effects have little influence on the structure and successional categories in the forest. However, the linear edge effect increases species richness and heterogeneity (H' and abundances) while affecting the maintenance of endemic species, like E. tetrapetala, which was recorded only inside the forest fragment. Accordingly, the implementation of linear infrastructures causes similar impacts to those triggered by other types of edge. It should be stressed that, no matter how short the distance between forest fragments, edge effects will indeed occur, which necessitates studies addressing management and recovery from the creation of linear edges.