The ArboreAl ComponenT of A Dry foresT in norTheAsTern brAzil

The dry forests of northeastern brazil are found near the coastal zone and on low, isolated mountains inland amid semi-arid vegetation. The floristic composition of these dry montane forests, as well as their relationship to humid forests (Atlantic forest sensu stricto) and to the deciduous thorn woodlands (Caatinga sensu stricto) of the brazilian northeast are not yet well known. This paper sought to determine if the arboreal plants in a dry forest growing on a low mountain in the semi-arid inland region (serra negra, 8° 35’ 8° 38’ s and 38° 02’ 38° 04’ W) between the municipalities of floresta and inajá, state of pernambuco have the same floristic composition and structure as that seen in other regional forests. in fifty 10 x 20 m plots all live and standing dead trees with trunk measuring ≥ 5 cm diameter at breast height were measured. floristic similarities between the forest studied and other regional forests were assessed using multivariate analysis. The results demonstrate that the dry forest studied can be classified into two groups that represent two major vegetational transitions: (1) a humid forest/dry forest transition; and (2) a deciduous thorn-woodland/ dry forest transition.

research supported by the project: "floristic composition and diversity of the upland forests of pernambuco" (Darwin initiative/ Cnpq/Apne/Ufrpe).

INTRODUCTION
Tropical dry forests are found in areas having high year-round temperatures and a well-defined dry season that limits both plant growth and animal activity (lüettge, 1997).Although climate is seasonally dry, enough moisture is available to support forest vegetation.
The concept of humid and dry tropical forests varies from author to author throughout the phytogeographical regions in which they occur (richards, 1996).Gentry (1995), for example, using a floristic approach, classified neotropical dry forests as those occurring in areas receiving 700-1600 mm of rain per year, and having at least five months of less than 100 mm.
Although dry forests cover extensive areas throughout the world, most research efforts have focused on tropical humid forests (Gillespie et al., 2000).Dry forest degradation has been similar to (and often greater than) that seen in humid tropical forests (Gentry, 1995;mooney et al., 1995;Janzen, 1997). in fact, Trejo & Dirzo (2000) argue that, for lack of non-disturbed study sites, few quantitative studies have been carried out on dry forests, which have a long history of human use, resulting in either their outright destruction or their conversion into sites with very different vegetation forms.
As the length of the dry season shows extreme variation throughout the tropics, no clear distinction exists between dry and humid tropical forests (Gentry, 1995), with one biome possibly changing into another over a distance of hundreds of miles (sarmiento, 1972). in addition, although tropical forest researchers have determined that annual precipitation rates and seasonality are closely correlated with plant diversity, this relationship may not always be clear in dry neotropical forests (Gillespie et al., 2000).furthermore, Gentry (1995) attributed differences between dry forests mainly to biogeographical factors, since these forests show no significant increases in either species richness or diversity as rainfall increases. pennington et al. (2000) noted that in seasonally dry regions of brazil the diverse forest and savanna types have been traditionally defined as belonging to the Amazonian forest, Atlantic forest, or Cerrado biomes.These same authors observed that the term "seasonally dry forest" includes several forest types (such as forests in more humid areas, and shrub vegetation in drier regions), and that the largest areas of seasonally dry forests in south America are found in northeastern brazil (the Caatinga biome, which extends southward almost to the state of minas Gerais).for these authors, the dry forests on the low hills in the semi-arid region of northeastern brazil represent the most favorable end of a humidity gradient that trails off into the deciduous thorn woodland (DTW) (Caatinga sensu stricto).
however, other authors regard these same forests as part of the Atlantic coastal forest biome that extends all along the brazilian seacoast, between the states of rio Grande do norte and rio Grande do sul.This biome is composed of a number of different vegetation types, ranging from humid forest (Atlantic forest sensu stricto) to seasonally dry forest (Atlantic forest sensu lato), within an area of high environmental heterogeneity (Coimbra-filho & Câmara, 1996).While the bulk of the Atlantic coastal forest (ACf) hugs the coast, it can extend inland along rivers into the savanna region (Cerrado) (oliveira-filho & ratter, 2000), and can cover the tops of low mountains in the borborema highlands or those scattered throughout the semi-arid region of northeastern brazil (Joly et al., 1999).The ACf has been intensely studied in southeastern brazil (scudeller et al., 2001), while its floristic composition and physiognomy are much less studied in the northeastern part of the country, with the possible exception of the forests in southern bahia (Thomas et al., 1998).
in the absence of sufficient data on this area, the present study sought to characterize the arboreal component of a dry forest community in the low mountains of brazilian semi-arid northeastern region (serra negra) and to answer the following questions: (1) Are there floristic and structural differences among dry forests in northeastern brazil?and (2) do the dry and humid forests in northeastern brazil represent a single vegetation unit?

Study area
The serra negra biological reserve (8° 35' -8° 38' s and 38° 02'-38° 04' W), between the municipalities of floresta and inajá, pernambuco state, is a low mountain located on a residual plain in the semi-arid region of northeastern brazil.According to Andrade (1954) and Dantas (1980), the serra negra region, which originated in the Cretaceous, has a cuesta landform.The landscape is a relict plain derived from a larger sedimentary formation that was reworked throughout a series of paleo-environments under varying humidity conditions.The reserve itself contains 1,100 ha, and is located approximately 400 km from the Atlantic coast, between the municipalities of floresta and inajá, in pernambuco state (ibAmA, 1989).The forest-covered summit measures 800-900 m in width, and is approximately 3,000 m long in an e/ne-W/sW direction.According to Andrade (1954), the maximum altitude is 963 m asl (fig.1).
Although no systematic rainfall records exist for the study area, estimates of precipitation based on regression (rodal, 1983) show that the average annual rainfall on the summit is approximately 900 mm per year.Average annual temperature is 25.3 °C, with an absolute maximum of 38 °C, and an absolute minimum of 8 °C (ibAmA, 1989).soils are deep and sandy (Jacomine et al., 1973).
The altitude of serra negra, combined with its exposure to humid southeast winds from the são francisco river valley, results in an amenable local climate in the midst of an otherwise semiarid landscape.A dense forest, with trees such as Manilkara salzmanii (ADC.)lam.(sapotaceae) and Trichilia emarginata (Turcz.)C. DC (meliaceae) reaching up to 35 m in height, covers the serra negra peak.
The main forested area in the eastern part of the brazilian northeast (Alagoas, pernambuco, and paraíba states) is located principally between the coastal plain and the borborema highlands.The latter consists of a series of highly weathered massifs, composed mainly of metamorphosed rocks, such as gneisses and eruptive granites (moreira, 1977).These massifs are the most outstanding regional geomorphological feature, the great diversity of which is paralleled by that of its vegetation, which includes deciduous thorn woodland (Alcoforado-filho et al., 2003), as well as both dry (nascimento, 2001) and humid forests on the eastern slopes (Tavares et al., 2000).both of these forest types are known as brejos de altitude.
Down from these highlands and eastward to the sea, precambrian rocks of the borborema highlands give way to the hills and coastal plains of the barreiras formation (Tertiary) whose vegetation ranges from humid (Atlantic forest sensu stricto) (siqueira et al., 2001) to dry forest (Atlantic forest sensu lato) (Andrade & rodal, 2004).The western slopes, which are dominated by DTW vegetation (rodal, 2001) of the borborema, descend steeply towards the semi-arid inland region known as the Depressão Sertaneja.scattered throughout this inland depression are a number of low mountains covered by dry forests, such as the serra negra.According to sampaio (1995) and rodal ( 2001), the climate in pernambuco state (where most of the quantitative surveys analyzed in this study were undertaken) is a complex combination of various climatic elements (precipitation, temperature, wind, etc.), which in turn are influenced by physiographic factors such as orography (borborema highlands) and continent effect.precipitation diminishes inland (westward), decreasing from 2000 mm year -1 in the coastal plains to 500 mm year -1 in the semi-arid depression.Temperatures are generally much higher in these inland areas, although the borborema highlands are a local exception, having slightly lower temperatures than the surrounding areas, especially during the night.

Field methods
fieldwork was carried out for 13 consecutive months, between January 1997 and January 1998.fifty 10 x 20 m plots on the top of the mountain (930-940 m asl) were set along three parallel transects separated from one another by 100 m.The plots themselves were spaced 20 m apart from one another.All trunks (living and dead), except for lianas and palms, with a ≥ 5 cm diameter at breast height (Dbh) were marked and identified to species; their diameters were recorded, and heights measured using an Abney level.

Data analysis
Voucher specimens were processed, identified, and stored at the professor Vasconcelos sobrinho herbarium (peUfr) of the Universidade federal rural de pernambuco (Ufrpe).The classification system used follows Cronquist (1981).
To identify different height classes, a 4 m interval was used to evaluate the species, the number of trunks, and their basal areas (1.3-4 m, 4.1-8 m, ..., 32.1 + m).Density, frequency, and basal area (Kent & Coker, 1992) for all families and species were calculated using fiTopAC 2 software (shepherd, 1995).The Cover importance index (matteucci & Colma, 1982) and shannon Diversity index (magurran, 1988) for families and species were also calculated.
in order to identify the similarities between the serra negra and the other forests in pernambuco, multivariate analyses were performed using presence/absence of species in 12 surveys, including the study area.The surveys covered humid forests (Tavares et al., 2000;siqueira et al., 2001;lins-esilva & rodal, in press); dry forests (the present survey; Correia, 1996;moura, 1997;nascimento, 2001;Andrade & rodal, 2004), specifically at 900 m asl (ferraz et al., 2003) and 700 m asl (rodal et al., accepted);and DTW (ferraz et al., 2003), specifically at 500 m asl (rodal et al., accepted).The Kendall correlation coefficient was used to test the relationship between altitude and basal area in these humid and dry forests (Ayres et al., 2000).
The terminology used to define montane (altitude > 600 m asl) and lowland forests (altitude < 100 m) followed Veloso et al. (1991).The authors that have worked in the northeastern forests differ with respect to nomenclature used in classifying ombrophylous (humid) and seasonal (dry) forests.for example, Andrade & rodal (2004) employed definitions suggested by Veloso et al. (1991) while Tavares et al. (2000) used floristic composition to distinguish forest types.The quantitative surveys used in our comparative analyses classified forests as either humid or dry in accordance with the descriptions supplied by the various authors cited above.Table 1 presents the analyzed surveys listed by municipality and includes bibliographic references, geomorphology (silva et al., 1993), altitude, and precipitation.Total annual rainfall at each survey site was obtained from either the studies themselves or sUDene (1990).
All species present in more than one survey were taken into consideration, regardless of the number of individuals in each.scientific names listed were synonymized according to the literature.Cluster analysis and principal component analysis (pCA) were performed using the same matrix.for the cluster analysis, the sørensen index, and the weighted pair-group method using arithmetic averages (WUpGmA) were employed.for the pCA analysis, a variance-covariance matrix was used.Analyses were undertaken using fiTopAC 2 software (shepherd, 1995).

Floristic composition and structure of the Serra Negra Forest
in 50 sample plots, a total of 1,525 stems (including dead ones) ha -1 were found, with a total basal area of 50.32 m 2 ha -1 .Average tree height was 8.9 m (± 5.55 m), while the average maximum height per plot was 22.8 m.Average diameter was 14 cm (± 14.74 cm), with an average maximum diameter per plot of 64.2 cm.fig.2a shows that the majority of trunks were less than 10 m tall.Approximately 50% of the total basal area was accounted for by only 7% of the total number of stems ( 103), all of which were over 20 m tall.While taller trees accounted for most of the basal area (fig.2b), the largest number of species was found among trees less than 12 m tall (fig.2c).basal-area distribution and trunk number per height class indicate a double-layered stratification, with emergent trees between 28 and 35 m tall.
The understory occupies the 4 to 12 m layer, with canopy beginning at 16 m.
fifty-one species belonging to 30 families were found in the survey (Table 2).shannon diversity index (h') for families was 2.06 nats ind. - The most species-rich family was myrtaceae (10 species), followed by fabaceae and sapotaceae (3 each).These three families accounted for 34% of all species recorded.
The families most heavily represented in the importance cover index (IC) were myrtaceae, mimosaceae, meliaceae, lauraceae and Urticaceae.All showed a reduced number of species, except myrtaceae, which accounted for 49.38% of the total density and 18.19% of the total basal area.if sapotaceae and Combretaceae (families having less than 50 individuals per hectare, Table 2), are added to this group, the sum of the total basal area increases to 71%.Although fabaceae contained more species in this survey than mimosaceae (both families of the order fabales), this is not important within the forest structure.
The shannon diversity index (h') for species was 2.72 nats ind -1 .seven species accounted for approximately 72% of the total iC, whereas other 30 species represented less than 1% of the iC.Myrcia fallax (myrtaceae) alone represented 35.87% of all trunks found in this survey.Myrcia multiflora (myrtaceae), Urera baccifera (Urticaceae), and Maprounea guianensis (euphorbiaceae) all showed high iC values, due in great part to their high relative densities and frequencies.Manilkara salzmanii (sapotaceae), Trichilia emarginata (meliaceae), Terminalia brasiliensis (Combretaceae), and Gallesia integrifolia (phytolaccaceae) stood out for their high relative dominance (Table 2): these four species accounted for 46.8% of total dominance and 10.2% of total density.moving vertically from understory to canopy, sapotaceae (mainly Manilkara salzmanii), meliaceae (Trichilia emarginata), and Combretaceae (Terminalia brasiliensis) became increasingly important.myrtaceae, with several species, mostly occupied the intermediate height classes, and Myrcia fallax was the most important species found in this study.it accounted for 39.4% of all individuals less than 12 m tall, and almost completely dominated the understory layer.Urticaceae (Urera baccifera) stood out in the < 4 m height class.species such as Manilkara salzmanii and Terminalia brasiliensis were also represented in different height classes, while Gallesia integrifolia (phytolaccaceae) was only present in the taller classes.

Floristic Similarities between the Serra Negra Forest and other Regional Forests
Cluster analysis of 155 species in 12 surveys allowed identification of three distinct groups, based primarily on their floristic composition (fig.3).The first group comprised all humid forests (1, 2, 3) and some dry forests (4,5,6). in the second group were two deciduous thorn woodlands (11,12), and -in the third group -only dry forests (7,8,9,10).

Are there floristic and structural differences between different dry forests in northeastern
Brazil? in the first place, the basis for distinguishing montane dry forests that are more related with humid forest near the Atlantic coast, from others that are more closely associated with seasonal plant formations characteristic of the Caatinga biome, is the flora and structure of each florest area.Thus, rodal (2001) argues that the dry montane forests studied by ferraz (1994( ), Correia (1996( ), and moura (1997) ) are more similar to the Caatinga biome, because of the absence of species characteristic of wet environments, and the presence of many species common to the northeastern Caatinga and other deciduous plant formations in south America.our multivariate analysis results suggest that dry forests are quite heterogeneous.These dry forests are classifiable by either of two transitional states: one, related to the Atlantic forest biome, that represents a transition from humid (1, 2, 3) to dry forest (4, 5, 6), and a second representing a transition from DTW (11,12) to dry forests (7,8,9,10).
Although in the Caatinga biome a strong relationship due to rising precipitation rates and decreasing temperatures is evident between altitude and plant height on a regional scale, this correlation may not be as noticeable on a local scale, at which plant height appears to be more strongly linked to soil water supply and local topography (sampaio et al., 1981).At 900 m asl, for example, dry forest can be found that is dominated by tall trees and associated with only a few low, branched, thorny, deciduous species (typical of DTW -Caatinga sensu stricto), e.g., the serra negra, whereas other dry forest with tall trees exists at the same altitude but associated with a large number of those same deciduous species, as in an area studied by ferraz et al. (2003).
Although murphy & lugo (1986) observed that seasonally dry forests have a smaller stature and a lower basal area than tropical rain (humid) forests, that pattern differs from the results of the present work on the serra negra forest and those of other dry and humid forests in studies by other authors using similar inclusion criteria (Table 1).
half of the surveys, such as those of dry forests (5, 7) and humid forests (2, 3), showed a total density (TD) of approximately 1,500 trunks ha -1 .The occurrence of significantly lower TD values in other studies (e.g., 780 stems ha -1 in lins-e-silva & rodal, in press) may be related to successional stages of regrowth, or to the degree of disturbance suffered in these areas.Considering all surveys using a 5 cm Dbh, variation in basal area was positively associated only with increasing altitude (0.32 with p ≥ 0.09), and not with subjective classification of forest areas as either dry or wet (Table 1).

Do dry and humid forests in northeastern Brazil represent a single vegetational unit?
rodal & sampaio (2002) observed that the biome concept is based on being able to identify "nuclear areas", in which a greater number of the characteristics considered basic to that biome can be found, as well as "marginal areas", in which these characteristics progressively decrease until arriving at the limits of another area where the vegetative and environmental features characterize another biome.These authors pointed out, however, that this approach has two basic requirements: identifying the basic features in each biome, and selecting the minimum number necessary to establish biome boundaries.however, no consensus about these criteria exists as yet among researchers studying the largest biomes in brazil.
The separation of serra negra and other montane dry forests (Correia, 1996;moura, 1997), specifically at 700 m asl rodal et al., in press) from dry forests that are more related to coastal forests confirms the hypothesis proposed by pennington et al. (2004) concerning the older, and now totally fragmented, seasonally dry neotropical forest biome, which includes that of the Caatinga.for these authors, the existence of strong floristic links between well-separated areas of seasonally dry forest in south America suggests that these areas are in fact fragments of an older and much larger forest formation.They argue that the drier conditions of the pleistocene (coinciding with glacial periods) favored expansion of dry over wet forest.To support this theory, they cite data from molecular biogeographical studies suggesting that, in many cases, the speciation of trees and shrubs in dry forest of south America pre-dates the pleistocene.

Fig. 2 -
Fig. 2 -Distribution of number of stems (a); basal area (b); and number of species (c) per height class in montane dry forest of the serra negra biological reserve.n = number of stems, n. spp.= number of species.

TABLE 2 Tree species sampled in montane dry forest in the Serra Negra Biological Reserve, Pernambuco state, Brazil. NS = number of trunks; RD = relative density (%); RDo = relative dominance (%); RF = relative frequency (%); and IC = importance cover. Names after families indicate names of those who identified the plant species and their herbaria.
4principal component analysis between serra negra and other forests sites in pernambuco state, brazil.1-2=lowlandhumidforest; 3 = montane humid forest; 4 = lowland dry forest; 5-10 = montane dry forest; and 11-12 = deciduous thorn woodland.seeinformationaboutthenumber of each area in Table1.andferrazetal., 2003)registered the presence of myrtaceae in areas of the Caatinga near the mountains and plateaus, as well as in the transitional zones between the humid forest and inland DTW. in these areas, again this family lacked structural importance.Vegetation within the Caatinga (sensu stricto) at 700(ferraz et al., 2003)and 500 m asl(rodal  et al., accepted), and dry forest at 900(ferraz  et al., 2003)and 700 m asl (rodal et al., accepted) changes with altitude increase in the eastern part of the Caatinga biome(rodal & sampaio, 2002).increases in precipitation, humidity, plant size, and species number per area from the lowest in the center of the inland depression to the higher mountain areas result in the gradual substitution of DTW by dry forest.inthe lowest areas of the Caatinga biome (ca.500 m asl) spiny, deciduous, and highly-branched plants typical of deciduous thorn woodland (DTW or Caatinga sensu stricto) and averaging between 3 and 4 m in height predominate.Anacardiaceae (Schinopsis brasiliensis, Myracrodruon urundeuva, and Spondias tuberosa), Cactaceae (Pilosocereus spp.and Opuntia spp.), euphorbiaceae (Cnidoscolus spp., Croton spp., and Jatropha spp.), mimosaceae (Mimosa spp.and Piptadenia spp.), and Caesalpiniaceae (Caesalpinia spp.) are all typical of DTW