Structure and fl oristic pattern of a coastal dunes in southeastern Brazil

Th e Brazilian shrublands area (restinga) is composed of marine coast vegetation on Quaternary sandy plains, where the species composition can vary depending on the surrounding ecosystems. Th e aims of this study were to describe the structure and fl oristic composition of a restinga near the community of Itaúnas, in the northern part of the state of Espírito Santo, to identify any relationships between this restinga and surrounding plant formations, and to determine which are the species that occur preferentially in the coastal forests of Espírito Santo. We sampled woody plants with a diameter at breast height ≥ 2.5 cm, excluding lianas, in 50 plots of 100 m2. We selected studies of coastal forests in the states of Espírito Santo and Bahia to prepare a database that would reveal patterns of fl oristic variation among these areas. We used two-way indicator species analysis for the identifi cation of the species that occur preferentially in the coastal forests of Espírito Santo. We identifi ed 114 species belonging to 38 families. Species richness was greatest for Myrtaceae (26 species), followed by Fabaceae (10). Th e Shannon index for the study area was 3.96. Th e estimated total density was 3,330 individuals/ha and basal area was 32.02 m2/ha. Th e highest importance value (IV) was for Protium heptaphyllum (IV, 23.4), indicating that it is characteristic of the Espírito Santo restinga. Th e results of our fl oristic analysis indicate that the species composition of the Itaúnas restinga is infl uenced by the so-called tabuleiro forests (coastal lowland forests on Tertiary deposits), which are most common in northern Espírito Santo. Th is seems to be the main factor responsible for the gradual reduction in fl oristic similarity between the restingas in the north of Espírito Santo and those in the south of the state, each constituting a distinct fl oristic block. In addition, we generated lists of species that occur preferentially in the restinga and tabuleiro forests. Th ose lists could inform decisions regarding environmental restoration programs.


Introduction
One of the major factors responsible for the biological diversity in the Atlantic Forest is its environmental heterogeneity, resulting from signifi cant variations in latitude, longitude and altitude, which function as key determinants of such diversity (Scarano 2002;Marques et al. 2011).Floristic studies conducted along the southeastern coast of Brazil have demonstrated the heterogeneity of the biome as a whole, revealing considerable similarities between forest formations, even those that are geographically distant from each other (Oliveira-Filho & Fontes 2000).Th ese conclusions were made possible by fl oristic analyses that considered the Atlantic Forest stricto sensu, despite having been recognized, within a broader approach, as a complex of vegetation patterns, including those distributed on the periphery and classifi ed as associated patterns, such as the shrublands, or restinga (IBGE 1992).
Distributed along the entire coastline of Brazil and therefore the most extensive vegetation type associated with the Atlantic Forest, the restinga is composed of a combination of established vegetation on sandy plains of marine origin formed either during the Quaternary (Menezes & Araujo 2005) and the Pleistocene, constituting the oldest, innermost portions of the plains, or during the Holocene, constituting the most recent, outermost portions (Martin et al. 1981. Martin et al. 1997).In these environments, the vegetation communities face marked variations in temperature, fl oods, constant wind, high salinity and nutrient-poor soil, adverse conditions that are refl ected in less diversity and structural complexity (Pereira 1990;Scarano 2002;Pimentel et al. 2007).Th e fl oristic composition of the restinga can also vary in function of the adjacent ecosystems (Freire 1990;Scherer et al. 2005), acquiring unique characteristics in each region.Th ese characteristics can result in marked fl oristic and structural diff erences between geographically proximate areas, creating distinct distribution patterns (Araujo 2000).
Phytogeographically, the restinga in the state of Espírito Santo is characterized by the fact that there is greater fl oristic similarity between areas of restinga located in the south of the state and those located along the coast of the neighboring state of Rio de Janeiro than between the former and those located in the north of Espírito Santo (Araujo 2000).Floristic studies of the restinga in the southern part of Espírito Santo have indicated that the fl ora of the adjacent hillside forests have a strong infl uence (Fabris & Cesar 1996;Assis et al. 2004b).In fact, the crystalline uplands are closer to the coast in the south of Espírito Santo than in the north of the state.In the north, the most conspicuous geomorphologic feature is constituted by the Tertiary deposits of the Barreiras formation, which extends from the crystalline uplands to near the coastline (Lani et al. 2008).Th e so-called tabuleiro forests (semi-deciduous, coastal lowland forests) develop on the Barreiras formation, between the restinga and the hillside forests (IBGE 1992;Jesus & Rolim 2005).Phytophysiognomically, this forest is primarily composed of mata alta (tall forest) vegetation and the so-called muçununga vegetation (similar to that observed in some parts of the restinga).Th e fi rst, which dominates the landscape, is distributed on clay soils with trees up to 35 feet high, and the second, forming enclaves within the mata alta vegetation, is on sandy soils of marine origin with some trees of small stature and taller trees reaching approximately 10 m in height (Peixoto et al. 2008;Simonelli et al. 2008).
Considering that the restinga is a geologically recent ecosystem and the colonizing species are from adjacent ecosystems (Freire 1990;Araujo 2000), we can hypothesize that the fl oristic composition of the restingas in the north of Espírito Santo are infl uenced by tabuleiro forest and that similarity diminishes toward the south of the state.In this context, the objectives of this study are to describe the structure and fl oristic composition of an area of Pleistocene restinga in Itaúnas, in northern Espírito Santo; to determine whether relationships exist between the restinga and adjacent fl oristic formations; and to identify the species that occur preferentially in the coastal forests of the state.

Study area
As shown in Fig. 1, the fragment of restinga studied is located on private land in Itaúnas, a district of the municipality of Conceição da Barra, located in the northern part of the state of Espírito Santo (18°25'16"S; 39°43'8"W), in an area of 51.3 ha and a border along the edge of the Itaúnas State Park, lying approximately 2 km from the coastline.Th e forest studied is associated with fi elds of Pleistocene sand ridges (Kenitiro Suguio, personal communication) that were formed aft er the peak of the penultimate marine transgression, initiated at approximately 123,000 B.C., when the relative sea level was 8 ± 2 m higher than the present--day level (Suguio & Martin 1978;Suguio et al. 1982).Th e dune ridge extends approximately 10 m above sea level, without any waterlogging of the soil due to fl uctuations in the groundwater level.Th e soil is composed of a surface layer of leaf litter ≤ 15 cm deep; 5 cm further down, there is a network of thin, intertwined roots, below which there is sandy sediment, the initial layer of which is white.

Vegetation sampling
Data were collected on 0.5 ha of restinga in Itaúnas, divided into 50 plots of 10 × 10 m (Müeller-Dombois & Ellenberg 1974).Th e plots were delineated systematically with use of a compass in 11 rows in a northwest-southeast orientation, providing the basis for implementation of alternately positioned plots at least 50 m apart.
We sampled all woody plants, except lianas, with a diameter at breast height (DBH) ≥ 2.5 cm, including dead and standing trees.For each sampled individual, we also estimated height with the aid of a pole pruner of known length (6 m).Individuals profi led above ground level and with a DBH < 2.5 cm were marked and measured when one of the branches fi t the inclusion criteria, with measurements from all branches then recorded to calculate the basal area.Th e fl oristic list was supplemented with individuals collected outside the plots, following the inclusion criteria established for the structural analysis.
All fertile material collected was categorized according to Fidalgo & Bononi (1984) and incorporated into the collection of the Central Herbarium of the Federal University of Espírito Santo (VIES).Th e botanical identifi cations were performed by consulting classical works and taxonomic reviews, as well as through expert consultation and by comparisons with material on fi le at the Central Herbarium of the Federal University of Espírito Santo, the Natural Vale Reserve Herbarium and the Herbarium of the Rio de Janeiro Botanical Gardens.Th e taxonomic classifi cation system used was that proposed by the Angiosperm Phylogeny Group (APG III 2009).For the abbreviations of the authorities for specifi c binomials, we followed the example of Brummitt & Powell (1992).

Vegetation structure
To evaluate the vertical structure, we used the absolute and relative phytosociological parameters described in Brower & Zar (1984).We chose to use the Shannon diversity index (H') and Pielou's evenness index (J').
To analyze the vertical stratifi cation of the forest, we developed an abundance matrix, organizing the sampled individuals into 1-m classes by height.To determine whether our sample was fl oristically representative of the vegetation formation, we performed a sampling suffi ciency analysis using Mao Tau rarefaction curves, with a confi dence interval of 95% (Colwell et al. 2004), and an abundance matrix by plots, calculated with the program PAST, version 2.0 (Hammer et al. 2001).Th is method consists in sorting the order of entry of the plots or individuals and calculation of the cumulative number of species, thus avoiding the infl uence that the randomness of the order of the sampling units might have on the construction of the curve (Kersten & Galvão, 2011).

Similarity analysis
For the similarity analysis, we included data from 14 fl oristic and phytosociological studies conducted areas in that were in a good state of conservation, including the present study, totaling 669 shrub or tree species (Fig. 1).For restingas and tabuleiro forests (Tab.1), we included mostly studies conducted in Espírito Santo, the exceptions being two studies conducted in the southern part of the state of Bahia.Th e forest referred to here as the muçununga of Caravelas, in Bahia, is, phytophysiognomically, a poorly drained system (Meira-Neto & 2005).Nevertheless, we treated it in a broader context, in order to elucidate fl oristic diff erences in comparison with the muçununga forest in Espírito Santo.Th e 14 studies selected were used in order to compile a binary matrix (presence vs. absence) of tree and shrub species occurring in those two areas, considering only the specifi c level determinations and verifi ed synonyms in the database from the list of species   1979).In order to rank the areas according to the fl oristic similarity and thus reveal patterns of variation in species composition along an environmental gradient (ter Braak 1995), we performed reciprocity averaging, equivalent to a correspondence analysis, the latter name being chosen due to its wider usage.Th e most numerous genera were Eugenia (15 species), Marlierea, Myrcia, Ocotea and Pouteria (5 species each).Th e mean height of the forest was 8 ± 4 m, the lower stratum comprising individuals 2-3 m in height (Fig. 2) and accounting for 44.7% of all sampled species.Th e most representative species, in terms of the numbers of individuals, were Unonopsis aurantiaca, Ocotea lobbii, Myrcia vittoriana, Eugenia astringens and Ouratea cuspidata.Th ese species collectively accounted for 7.5% of the individuals and 0.8% of the total basal area.Th e middle stratum comprised individuals 4-9 m in height, accounting for 76.2% of all sampled individuals, 92% of the species and *Species outside the complementary sampling of the fl oristic list.17.6% of the total basal area, O. lobbii, U. aurantiaca, M. vittoriana, E. excelsa and E. astringens showing the greatest density in this stratum.Individuals in the upper stratum (10-22 m in height) accounted for 16.7% of the species, the tallest individuals being those of Protium heptaphyllum, Tapirira guianensis, Kielmeyera albopunctata, Macrolobium latifolium and Aspidosperma pyricollum.Upper-stratum species accounted for 25.8% of the individuals sampled and covered 81.6% of the total basal area.
The behavior of the species-area rarefaction curve indicated a tendency towards stabilization, given that, by the time half of the allocated plots had been sampled, approximately 85% of the species had been included (Fig. 3).Of the 114 species sampled, 20 (17.5%) were represented by a single individual.
In Itaúnas, the H' was 3.96 and the J' was 0.836.Among trees and bushes, we sampled 1665 living individuals, collectively covering a basal area of 32.02 m²/ha.We recorded 107 dead individuals totaling 1.10 m²/ha basal area, with 65.5% of individuals with DBH ≤ 5 cm and 24% ≤ 10 cm DBH.As can be seen in Tab. 3, Burseraceae and Anacardiaceae stood out in that they accounted for a considerable portion of the basal area mainly because of a single species.P. heptaphyllum occupied 10.4% of the total basal area, accounting for 64.5% of the basal area covered by Burseraceae, and T. guianensis occupied 6.1% of the total basal area, accounting for 98.2% of the basal area covered by Anacardiaceae.Myrtaceae was the most important family, accounting for 16% of the total importance value (IV), E. astringens and E. excelsa showing IVs of 10.8 and 9.9, respectively (Tab.4).
Taking all of the phytosociological information, as well as the fl oristic composition and the soil water saturation, into account, we can classify the forest formation analyzed as a dune-ridge forest, according to the classifi cation system devised by Menezes & Araujo (2005) and Silva & Britez (2005).

Similarity analysis
Th e similarity coeffi cient between the coastal forests of Espírito Santo and those of southern Bahia ranged from 0.03 to 0.86, which, by TWINSPAN, could indicate that some species were preferential to these forests (Chart 1).According to the similarity dendrogram (Fig. 4), restingas constitute a cohesive cluster associated by a similarity coeffi cient ranging from 0.27 to 0.57, with only Capparis fl exuosa, Garcinia brasiliensis, P. heptaphyllum and Guapira opposita occurring in all of the restingas considered.Within the cluster consisting of restingas, one subcluster comprised restingas in northern Espírito Santo (Conceição da Barra and Itaúnas) and another comprised restingas in the southern and northern parts of the state (Interlagos, as well as two locations with the Paulo César Vinha State Park, previously known as Setiba Park, in the south; and Pontal do Ipiranga in the north).
Th e fl oristic similarity between the restinga of Itaúnas and the other restingas analyzed ranged from 0.29 to 0.43, varying even when compared with the survey conducted in Conceição da Barra (Gomes Pereira & 1994), which is quite near our study area.Nevertheless, Itaúnas and Conceição da Barra proved to be cohesive in the similarity analysis and consistent in the TWINSPAN divisions (eigenvalue, 0.499), Eriotheca macrophylla, Ficus gomelleira, Joannesia princeps, M. latifolium and Pterocarpus rohrii being identifi ed as responsible for this clustering.Given the fl oristic similarities between the restinga of Itaúnas and the other formations evaluated (0.08-0.43), the Itaúnas restinga, as would be expected, showed greater affi nity with the other restingas.However, despite the low degree of fl oristic affi nity between the Itaúnas forest and the tabuleiro forests (0.15-0.22),Itaúnas was more similar to the tabuleiro forest than to the other restingas analyzed (0.04-0.16).
According to the similarity dendrogram, the muçununga forest in Linhares (Rio Doce Valley Nature Reserve) had greater fl oristic affi nity with the restinga.Th e muçununga forest in Caravelas constituted an outlier to all of the other forests analyzed, with a similarity coeffi cient of 0.06.
Floristically, the areas of mata altaformed a very cohesive cluster, generating the highest similarity indices (0.46-0.86).Although the mata alta in Caravelas (southern Bahia) is geographically the most remote of the mata alta surveyed, the mata alta in the Sooretama Biological Reserve showed the least similarity.
Th e correspondence analysis eigenvalues for the fi rst two axes were 0.739 and 0.604 for axis 1 and axis 2, respectively, the fi rst explaining 17.88% of the total variance, compared with 14.63% for the second, collectively explaining 32.51%.Th erefore, the ordination diagram distinguished a cluster that was highly associated by the proximity among the mata alta areas (Fig. 5).Th e restingas formed another cluster, which, unlike the mata alta areas, comprised formations that were more or less aligned in order.Along axis 1, the muçununga forests were in an intermediate position between restinga and mata alta.According to axis 2, the muçununga forest in Caravelas was fl oristically distant from the other forests.

Floristics, structure and diversity
In fl oristic surveys of shrub or tree formations in well--drained restingas in northern Espírito Santo, 82 species have been observed in Linhares (Pereira et al. 1998), comparable to the 83 observed in Conceição da Barra (Pereira & Gomes 1994).In the cities of Vila Velha (Pereira & Zambom 1998) and Guarapari (Assis et al. 2004a), coastal communities in the south of the state, 74 and 103 species, respectively, were observed.In the state of Rio de Janeiro, which has the longest tradition of studying restinga communities, species richness has been shown to range from 26 to 108 species (Assumpção & Nascimento 2000;Lobão & Kurtz 2000;Sá & Araujo 2009).Th e Itaúnas forest presented even higher species richness, indicating that it is an important forest remnant that should be preserved.
Th e great species richness of Myrtaceae is commonly known and has also been detected in other restingas along the Brazilian coast (Araujo 2000;Assumpção & Nascimento 2000;Scherer et al. 2005;Sacramento et al. 2007;Amaral et al. 2008).Th is might be related to the adaptability of Myrtaceae species to low fertility soils (Ashton 1988), such as those of the restinga (Scarano 2002).
Th e high species richness of the Itaúnas restinga is attributable to the presence of families that are also plentiful in other neotropical formations (Leitão-Filho 1987).Along the southern coast of Brazil, where Waechter et al. (2000) and Guedes et al. (2006) each found only one Sapotaceae species (Sideroxylon obtusifolium (Roem.& Schult.)T.D. Penn.and Manilkara Subsericea, respectively), Sapotaceae is the family that has had the highest VI.Along the northern and northeastern coasts, Sapotaceae is also not among the most species-rich families, only four species or fewer having been observed (Freire 1990;Sacramento et al. 2007;Santos-Filho et al. 2011).Studies addressing the structural contribution of these species in the communities of the northern and northeastern Brazil are scarce, which precludes comparisons across studies.On the coast of the state of São Paulo, in southeastern Brazil, Guedes et al. (2006) found that a single Sapotaceae species was the most important species in a fl ooded restinga and the fi ft h most important in a terra fi rme forest.In the restinga within Paulo César Vinha State Park, on the southern coast of Espírito Santo, Assis et al. (2004b) found that the Sapotaceae family was also important, Pouteria coelomatica having the second highest IV.Unlike the authors of the aforementioned studies, those authors also found Sapotaceae to be one of the most species-rich families, being represented by nine species.Similarly, in that study, Sapotaceae stood out for its contribution to the structure and species richness of the community.Sapotaceae has also been identifi ed as being among the most species-rich families in the restinga in the state of Rio de Janeiro (Pereira & Araujo 2000).Given its wide distribution along the Brazilian coast, Sapotaceae can also be considered characteristic of nutrient--poor soils such as that of the restinga.In the northern and northeastern regions, with the data currently available, Sapotaceae can be identifi ed as being of key importance for the fl ora of the restinga forests.For the southern and southeastern regions, Sapotaceae is of relevant structural importance.In the southeast, Sapotaceae plays an important role in the fl ora and structure of restinga forests in the    region.Th e pattern observed allows us to conclude that the environmental conditions of the areas evaluated, including climatic conditions generated by latitude, proximity to other vegetation types and the infl uence of human activity, might explain the similarities and diff erences found for this family along the coast (Guedes et al. 2006).
Although there was considerable density of Annonaceae species, mainly U. aurantiaca, in the lower and middle strata, the great numbers of Myrtaceae species in these strata made the latter family predominant.However, in the upper strata, Myrtaceae species become less representative for all parameters, confi rming that it is a family typical of the lower strata, as stated by Tabarelli & Mantovani (1997).Th e middle stratum is representative of species in development, which are also present in the upper strata, thus contributing to the maintenance of species diversity in this forest.However, individuals in the upper stratum contribute little to the richness and density, although they are primarily responsible for biomass stocks, given their high representation in terms of basal area.Because the low nutrient content in the soil of the restinga makes biomass the principal nutritional reserve maintained through nutrient cycling (Guedes et al. 2006), the upper and lower strata are both quite important for maintaining the structural complexity of this type of forest.
For the species-area curve, our sampling can be considered suffi cient despite the high proportion of species represented by only one individual.In restingas, this does not necessarily indicate that the species are rare.For example, in a restinga in southern Espírito Santo, Schinus terebinthifolius was found to be represented by a single individual (Assis et al. 2004b).In shrub formations along the backshore, as well as in disturbed environments, S. terebinthifolius has a higher density (Assumpção & Nascimento 2000;Lobão & Kurtz 2000;Menezes & Araujo 2005).In previous studies, the number of shrub and tree species observed in restinga vegetation formations has varied considerably, species richness typically being highest in terra fi rme forest, although rarely exceeding 150 species (Pereira 1990, Pereira & Araujo 2000;Silva & Britz 2005;Menezes & Araujo 2005;Amaral et al. 2008).In the present study, we observed a high number of species with low density, in comparison with other restingas of Brazil, which might be an intrinsic characteristic of the community evaluated and could be responsible for the increase in species diversity.
In the Itaúnas restinga, the diversity index, density and basal area were higher than in other Brazilian restingas (Assumpção & Nascimento 2000;Waechter et al. 2000;Scherer et al. 2005;Guedes et al. 2006;Sá & Araujo 2009), which could be indicative of a good state of preservation.In addition to the higher diversity index, Myrtaceae richness and overall species richness were higher in our sample than in those of other studies of the vegetation of restingas in Espírito Santo (Fabris & César 1996;Assis et al. 2004b).However, this result might have been infl uenced by the fact that our inclusion criteria were broader than were those of some of the aforementioned studies, resulting in a more inclusive sample.By adopting a DBH ≥ 5 cm as an inclusion criteria, we eliminated 16 species (three Myrtaceae) and 850 individuals (51% of the total) from our sample, the fi nal sample therefore totaling 98 species.Nevertheless, the values of H' and J' (3.93 and 0.856, respectively) were still higher than those obtained for other restingas in Espírito Santo and for those in other Brazilian states.Th erefore, the exclusion of individuals with a DBH < 5 cm implied partial exclusion of the undergrowth, interfering little in the total number of species, given that the middle stratum comprises the largest portion of the wealth of this forest.
Tree death is a natural process triggered by biotic or abiotic conditions that contribute to the forest dynamics of tropical ecosystems (Franklin et al. 1987).In the Itaúnas restinga, dead trees, if they were to be considered in a specifi c category, would rank 4th in IV, with a density of 214 ind./ ha.Th is value is higher than that found for restingas with well-drained soils along the coast-such as the 58.9 ind./ha reported for the Tavares restinga, in the state of Rio Grande do Sul (Dorneles & Waechter 2004); the 67.0 ind./ha and 104.0 ind./ha reported for the restingas in Paulo César Vinha State Park, in Espírito Santo (Assis et al. 2004b;Fabris 1995); and the 120.0 ind./ha reported for the Armação de Buzios restinga, in the state of Rio de Janeiro (Lobão & Kurtz 2000)-as well as that found for poorly drained restingas, such as the 152.0 ind./ha reported for the Marambaia restinga, in the state of Rio de Janeiro (Menezes et al. 2010).However, the basal area of dead trees in Itaúnas was similar to that found for those well-drained restingas and lower than that found for the poorly drained Marambaia restinga (1.28 m²/ha).Th is can be attributed to the fact that fl ooding creates soil conditions that restrict tree development (Menezes et al. 2010).In addition, the signifi cant diff erence between our fi ndings and those of other studies of restingas, in terms of the density of dead trees, can be attributed to the inclusion criterion adopted in our study (DBH ≥ 2.5, a diameter smaller than that used in other studies).Th is allowed for the inclusion of a high density of thin individuals, as evidenced by the concentration of individuals in the fi rst two diameter classes, which are more susceptible to disease, climate change and senescence due to competition for light (Franklin et al. 1987).
Th e concentration of the IV in a few species is common in Brazilian restingas.In the present study, this became apparent when we summed the IVs of the 13 most important species, which were found to account for approximately 50% of the total density of IV and approximately 60% of the total dominance, as reported for other restingas along the coast (Assumpção & Nascimento 2000;Assis et al. 2004b).Th is indicates that environmental factors regulate the composition of species, limiting their establishment and development, resulting in a highly oligarchic structure that can refl ect severe environmental conditions, such as the highly porous and nutrient-poor nature of the restinga soil, which is unfavorable for the establishment of many species (Scarano 2002).However, some species seem to fl ourish in these environmental conditions, particularly those belonging to the family Myrtaceae, which stood out in this forest due to their high density, which is common in restingas.Trindade (1991), studying a shrub/tree restinga near the city of Natal, noted the predominance of Myrtaceae in the sample, attributing this to some power over other species or to a better ability to take advantage of available resources.
Th e low similarity coeffi cients found for the coastal forests of Espírito Santo showed high fl oristic diversity, given that similarity coeffi cients greater than or equal to 0.5 indicate high fl oristic similarity (Kent & Coker 1992).Th ese conclusions must be interpreted with caution, because this fl oristic heterogeneity might, in part, refl ect diff erences among the sampling methods employed for the various areas evaluated.Nevertheless, this does not invalidate the results showing that, despite their proximity to each other, these forests show fl oristic characteristics that are particular to each region, a pattern commonly found in Brazilian restingas (Pereira 1990;Assis et al. 2004a;Scherer et al. 2005;Amaral et al. 2008).Th is high fl oristic heterogeneity complicates the implementation of environmental restoration programs, because of uncertainties over which species to be used, given that the introduction of an exotic species can harm the ecological system (Lani et al. 2008).Th erefore, the identifi cation of preferentially occurring species in this region is aimed at facilitating this process.
According to the TWINSPAN, C. fl exuosa, G. brasiliensis and P. heptaphyllum were identifi ed as preferential to restinga and as characteristic of the restinga of Espírito Santo, in relation to its tabuleiro forests, the latter making an important contribution to the structure of the restinga tree community (Assis et al. 2004b).In the TWINSPAN division, we found G. opposita to be non-preferential, which is in agreement with studies indicating that it is a generalist species (Oliveira-Filho & Fontes 2000).
Th e similarity dendrogram shows that the restingas of Espírito Santo do not form clusters clearly distinguishing those in the north of the state from those in the south.For example, the Pontal do Ipiranga restinga, along the northern coast, showed greatest fl oristic similarity to the Interlagos restinga, located on the southern coast.However, this fi nding should also be interpreted with caution, because, despite sharing characteristics with the southern restinga cluster, the Pontal do Ipiranga restinga has few species in common with those forests, which precludes any hierarchical divisions in the TWINSPAN.Th erefore, the inclusion of the Pontal do Ipiranga restinga weakens the structure of that cluster, because it is an artifact of the analysis, probably attributable to diff erences in methodology and sampling power among the studies.
Th e low fl oristic similarity between the Itaúnas and Conceição da Barra restingas denotes an intrinsic character of the restingas, refl ecting the high heterogeneity among the sites surveyed, in terms of species composition.Th is uniqueness provides evidence that the tabuleiro forest, given its proximity, had an infl uence on the Itaúnas restinga, providing seedlings of uncommon species that fi nd conditions favorable to their establishment in the restingas.Other species identifi ed by the TWINSPAN as preferential to the tabuleiro forests of northern Espírito Santo can also be found in the Itaúnas restinga, including Eugenia pisiformis, Jacaratia heptaphylla, Marlierea sucrei, Swartzia simplex and Tabebuia roseoalba (Chart 1).
Although the muçununga forest in Linhares (Natural Vale Reserve) was established on Tertiary land, forming enclaves in the mata alta, it features soil conditions similar to those observed in the restingas analyzed, the soil being essentially sandy and well-drained (Simonelli et al. 2008), which justifi es its position in the dendrogram.Th e muçununga forest in Caravelas is fl ooded periodically (Meira- -Neto et al. 2005), which explains the low degree of fl oristic similarity between this and the other formations.Chief among the species identifi ed by the TWINSPAN as preferential is Tabebuia cassinoides (Lam.)DC., which is known to tolerate poorly drained environments (Scarano 2002).
Th e northern Espírito Santo/southern Bahia region is a recognized focus of plant species diversity (Peixoto & Silva 1997), presenting a unique biogeographic pattern in that it constitutes an area within which semideciduous lowland forests and northern rainforests overlap, which could generate fl oristic affi nities even between geographically distant areas, given the strong fl oristic links with tabuleiro forests in that region (Oliveira-Filho & Fontes 2000; Rolim et al. 2006).Some of the species identifi ed by the TWINSPAN as responsible for the clustering have also been cited for Atlantic Forest, in the northern lowlands of Espírito Santo (Oliveira-Filho & Fontes 2000), including Brosimum guianense, Carpotroche brasiliensis, Caryocar edule, Cedrela odorata, J. princeps, Melanoxylon brauna, T. roseoalba and Virola gardneri, as well as in the lowlands to the south, including Allophylus petiolulatus, Ecclinusa ramifl ora, Micropholis crassipedicellata, Pouteria venosa, Psychotria carthagenensis and Sweetia fruticosa.
In the correspondence analysis, we obtained eigenvalues that are considered high and therefore indicative of good separation of areas, as well as indicating complete turnover of species along its axes (ter Braak 1995).According to the ordination diagram, the restingas formed an ordered cluster that was essentially aligned toward a gradient of fl oristic similarities, graphically illustrating the possibility that the distance between the restingas of Espírito Santo and the tabuleiro forests (in the north of Espírito Santo and the south of Bahia) creates gradual variations in species composition, in which the Itaúnas restinga was notable for its degree of similarity with the tabuleiro forests.In addition, the diff erence between the Itaúnas restinga and the southern restingas, in terms of fl oristic composition, might refl ect the latitudinal position of the former, which is infl uenced by the dense tropical rain forest on Precambrian terranes that is more common in this region (Fabris & Cesar 1996;Assis et al. 2004a).However, this result should be interpreted with caution given the sensitivity of the correspondence analysis to species of low occurrence, and it is suggested that those be given less weight, because they can introduce noise and tend to contribute little to the patterns generated by the analysis (ter Braak 1995).Th erefore, when species occurring in only one area were excluded from the analysis, the ordering pattern became conservative, even when species occurring in two, three or four areas were excluded (Fig. 6).Furthermore, the division of the restingas of southern and northern Espírito Santo into two fl oristic blocks became more evident, providing evidence of a historical separation that was probably caused by the infl uence of the surrounding vegetation on the fl oristic composition of the restinga.
In the ordination diagram, the muçununga forest in Linhares can be seen to be in close fl oristic proximity to the restingas and the infl uence of the mata alta on species composition in the muçununga forest is evident.Th e TWINSPAN confi rmed this pattern of diff erentiation, allowing the identifi cation of species responsible for the division between the muçununga forest and the mata alta.Th ose species include Andira nitida, Cupania emarginata, Eugenia astringens, G. brasiliensis, Myrcia splendens and Rhodostemonodaphne capixabensis, all of which are commonly found in the restingas of Espírito Santo (Pereira & Araujo 2000).As in the similarity dendrogram, the muçununga forest in Caravelas, in southern Bahia, was found to be distant from all other forests in the ordination diagram, confi rming that the species composition of the community is quite distinct from that of the others analyzed here.
Th e Itaúnas forest retains unique fl oristic characteristics because of many factors, among which is its high diversity.Th e fl oristic affi nity between the Itaúnas restinga and tabuleiro forests revealed important fl oristic patterns, suggesting that the species composition of the restingas of Espírito Santo is a refl ection of distinct environmental and geomorphological conditions, the surrounding vegetation having a decisive infl uence.Th is seems to be the main factor responsible for the gradual reduction in fl oristic similarities between the restingas of the north and south.In addition, the indication of preferential species generated by the TWINSPAN can inform the preparation of species lists used in environmental restoration programs, mainly in the restinga, given its naturally limiting conditions and the intense process of fragmentation that it is undergoing.

Figure 1 .
Figure 1.Location of the study area and distribution of sites used in the fl oristic analyses of the coastal forests in the state of Espírito Santo and in the southern part of the state of Bahia, Brazil.Restinga -shrublands; muçununga -similar to restinga but established on Tertiary land, forming enclaves in the mata alta (tall forest).

Figure 2 .
Figure 2. Distribution of frequencies of classes of total height (m) of the individuals in the restinga (shrublands) in Itaúnas, near the municipality of Conceição da Barra, in the state of Espírito Santo, Brazil.

Figure 3 .
Figure 3. Species-area rarefaction curve, generated with the Mao Tau method, for the restinga (shrublands) in Itaúnas, near the municipality of Conceição da Barra, in the state of Espírito Santo, Brazil.CI -confi dence interval.
Th e closest weather station is located in the municipality of São Mateus, approximately 35 km from the study area.Climatic data recorded over the last 34 years by the Espírito Santo State Institute for Research, Technical Support and Rural Extension Services/National Meteorology Institute show that average annual temperature ranges from 21.7°C to 26.7°C, with a relative humidity of 83% and an average annual rainfall of 1308 mm.According to the Köppen climate classifi cation system, the study area climate is type Aw (tropical savanna), a humid tropical climate in which rainfall rates are highest during the summer months and lowest during winter months.

Table 1 .
Information about the sites used in the fl oristic analyses of 14 coastal forests in the state of Espírito Santo and in the southern part of the state of Bahia, Brazil.

Table 2 .
Floristic list of shrub/tree species occurring in the restinga (shrublands) in Itaúnas, near the municipality of Conceição da Barra, in the state of Espírito Santo, Brazil.

Table 3 .
Phytosociological parameters of families sampled in the restinga (shrublands) in Itaúnas, near the municipality of Conceição da Barra, in the state of Espírito Santo, Brazil.

Table 4 .
Parameters of the species sampled in the restinga (shrublands) in Itaúnas, near the municipality of Conceição da Barra, in the state of Espírito Santo, Brazil.