FLORISTIC COMPARISON BETWEEN TWO TREE COMMUNITIES ASSOCIATED WITH HABITAT DESCRIPTOR VARIABLES

ABSTRACT: The knowledge about the influence of habitat variables is essential to understand the underlying ecological patterns in vegetation. This study compared the floristic composition of two forest communities located in different altitudes. Associated with this comparison, we used a methodology where habitat descriptor variables were scaled and interpreted by the biotic set sampled. We constructed one matrix with scores given to physical, biotic, vegetation, and anthropogenic variables in the field and one matrix with the species sampled and performed multivariate analyses. We found that the floristic communities differ between the different altitudes and that the methodology used showed significant variables for the ecological characterization of the sampled habitat.


INTRODUCTION
The distribution of plant species is determined by physicochemical and biological factors, and by a combination of these factors (CONTI; FURLAN, 2009;COX;MOORE, 2011).At the macroscale, climate is the major determinant of plant distribution, as it provides different conditions of temperature and precipitation (WHITTAKER, 1975;WALTER, 1985).Conversely, variations in topography and soils are the main determinants of vegetation at smaller scales (NEWTON, 2007;FERREIRA JÚNIOR et al., 2009).Moreover, interactions between species can also affect plant distribution, as the establishment and growth of several species can be inhibited or favored depending on the distribution of other plant species (KENT; COKER, 1992;SANTOS et al., 2012).Natural or anthropogenic disturbance events also cause changes in vegetation and may cause changes in the structure and dynamics of plant communities and populations (BEGON et al., 2007;GHAZOUL;SHEIL, 2010).
The influence of environmental variables on plant species may result in different distribution patterns that vary over time and space (MEIRELES et al., 2008;SANTOS et al., 2012).Variation in species distributions requires that the combined effects of vegetation and environmental variables be evaluated together (TER BRAAK, 1987), taking into account that some variables are more responsive than others when a particular group of species is considered.Thus, it is often necessary to combine methods with floristic surveys and forest structure studies to elucidate ecological patterns and processes.Nevertheless, some of the information that characterizes the habitat remains unknown due to measuring difficulties or the high cost of quantification.
This study aimed to evaluate the floristic variation of two plant communities at different altitudes from a qualitative method for habitat characterization that uses a set of scaled variables.We expect that this method may serve as support information for the evaluated communities and provides additional parameters to help explain the ecological patterns of vegetation.

MATERIAL AND METHODS
We listed characteristics of native forest habitats based on a literature review, considering biotic and abiotic factors, arboreal physiognomy, and human disturbances.We selected 20 variables that were treated as habitat descriptor variables (HDV).These variables were grouped into the categories described below.A Abiotic descriptors: A1 Slope, A2 Rocky outcrops, A3 Leaf litter deposition, A4 Humidity, and A5 Wind exposure.B Biotic descriptors: B1 Bamboo density, B2 Bromeliad density, B3 Orchid density, B4 Bryophyte density, B5 Lichen density, and B6 Liana With the selected variables we constructed a semiquantitative matrix for survey in the field, as in the methods of phenological assessment (FOURNIER, 1974;WHEELWRIGHT, 1985).We used a interval scale (0 -4) to infer the influence of variables on sampling units, with 0 indicating the lack of influence, 1 indicating a 1 -25% influence, 2 to 26 -50% influence, 3 to 51 -75% influence, and 4 indicating a 76 -100% influence.
Two areas were sampled in the same region, in the municipality of Itamonte, on a hillside of the Mantiqueira range, southern Minas Gerais state, Brazil.The floristic composition was sampled at 1500 m and 1900 m above sealevel (a.s.l.), in 30 sample points (15 in each area) delimited by a circle with 10 m radius, totaling 0,942 ha of sampling.At each sample point, all arboreal and sub-arboreal species were sampled, considering all individuals over 1m high.Simultaneously to the vegetation survey, habitat descriptor variables were scaled using the habitat matrix.
Detrended Correspondence Analysis (DCA) (TER BRAAK, 1987) from a binary presence-absence species matrix and a categorical matrix of altitude was used to check the effect of altitude on the distribution of species independent of other habitat variables.Canonical Correspondence Analysis (CCA) (TER BRAAK, 1987) was used to infer the influence of HDVs on the distribution of species.In this case, in addition to the species matrix, we constructed a matrix with HDV data and used the Monte Carlo test to test the significance of correlations.CCA was performed on all habitat variables (CCA1) and on variables with correlation above 0.7 with the vegetation gradient (CCA2), considering the first two axes.The analyses were performed using PC-ORD 5.10 for Windows software (MCCUNE; MEFFORD, 2006).

RESULTS AND DISCUSSION
In total, 133 species were sampled, 98 at 1500 m and 78 at 1900 m.Of the total species sampled, 55 were unique to the lower altitude area and 35 were sampled in the higher altitude area only.These results are in accordance with the notion that species richness decreases with altitude in tropical forests (GENTRY, 1995;RAHBEK, 1995;ADAMS, 2010;SCHEER et al., 2011).The list of species is shown in Table 1.In addition to species surveys, understanding the influence of habitat variables on the organization of biological communities helps corroborate the patterns observed and provides the delimitation of groups and the knowledge of the relationships between these groups and their habitats.The survey of the variables (Table 2) provides a greater basis for Table 1 List of arboreal and sub-arboreal species in sampling points at two altitudes in the Mantiqueira range, municipality of Itamonte, Minas Gerais state, Brazil (0 = absence of the specie; 1 = presence of the specie).
A Ordination analysis showed that arboreal and sub-arboreal vegetation differed between altitudes (Figure 1).Axis 1, which represents the main fl oristic gradient, indicates that each altitude has a different group of species.The eigenvalue of this axis (0.558) shows a long gradient, indicating the occurrence of species unique to each altitude (TER BRAAK, 1995).Moreover, there is a greater cohesion in the group of species from 1900 m a.s.l.due to the presence of species typical of higher elevations such as Byrsonima laxifl ora, Drimys brasiliensis, and Symplocos celastrinea, in addition to the genus Ilex (OLIVEIRA FILHO;FONTES, 2000).variables (scored value 0), were used in CCA1, totaling 16 quantitative variables and one categorical variable (altitude).This analysis showed that the distribution of species correlated signifi cantly with HDVs (p < 0.01).The difference in species composition between the two areas sampled was due to the differential distribution of species along the altitudinal gradient.Nevertheless, altitudinal variations comprise a complex set of variables related to temperature, air humidity, water availability, wind exposure, in addition to soil depth and drainage classes (CARVALHO et al., 2005;HOMEIER et al., 2010;BLUNDO et al., 2012).
Canonical Correspondence Analyses showed signifi cant correlations between habitat variables and community organization (Table 3).All habitat descriptor variables (HDV), except for the four anthropogenic CCA2 was calculated using the fi ve variables that correlated with the vegetation gradient (r > 0.7, considering the fi rst two axes): humidity, wind exposure, bromeliad density, orchid density, and lichen density (Table 4).According to CCA2, axis 2 showed the correlation between humidity with the distribution of species sampled at 1500 m (Figure 2).
It should be noted that humidity, which is important for temperature regulation and nutrient transport in soil (CONTI; FURLAN, 2009), favors the establishment of species adapted to the microclimate provided by it.This effect was observed in the area at 1500 m, which was close to a water course.In fact, this variable probably favored the establishment of species such as Casearia obliqua, Endlicheria paniculata, Luehea divaricata, Magnolia ovata, Platycyamus regnellii, Protium spruceanum, and Vochysia magnifi ca, whose occurrences were previously associated with humid habitats (OLIVEIRA FILHO; FLUMINHAN FILHO, 1999;BOTREL et al., 2002;SOUZA et al., 2003;CARVALHO et al., 2005;ROCHA et al., 2005).
The variable wind exposure, most associated with the axis 1, is correlated to the distribution of species at 1900 m.This variable is associated with high altitudes and acts on the composition, structure and distribution of tree species in communities located in elevated areas (NEWTON, 2007;HOMEIER et al., 2010;BERTONCELLO et al., 2011).In study along a 2.5 km transect ranging in tropical montane cloud forests in north-western Costa Rica, Häger (2010) considered the wind exposure as a key variable to tree species turnover along environmental gradients on a local scale, therefore the constancy of the wind acts upon the air temperature and the cloudiness.Considering the ecology of cloud forests, the author also points out that these variables determine the occurrence of species, including endemic species, in habitat with strict conditions.(HÄGER, 2010).
Between the two communities evaluated the wind exposure should be acting in this way, contributing to the differentiation of the area located 1900 m a.s.l.In addition to the relationship with altitude wind exposure may also be associated with the occurrence of bamboos and canopy openness.Despite not entering in the inclusion criteria of CCA2, these two variables showed signifi cant correlations with axis 1 to the area located at 1900 m.
Another variables more strongly associated with species distributions at 1900 m with greater amplitude along axis 2 were orchid, bromeliad and lichen density, all of which typical of montane forests (ADAMS, 2010;GHAZOUL;SHEIL, 2010;SCHEER et al., 2011).The presence of these botanicals groups in montane forests is related to the constant water supply provided by the cloudness, which makes its typical species of this habitat type (ADAMS, 2010;ACHARYA, 2011;BLUM et al., 2011;KRÖMER et al., 2013;DYMYTROVA et al., 2014).Thus, some tree species characteristics of high altitude as Ilex paraguariensis, Croton piptocalyx, Leandra aurea, Ouratea fl oribunda, Roupala rhombifolia e Symplocos insignis, exclusive from community located at 1900 m and sampled in majority points, may have their local occurrence related to microclimate conditions.
The use of HDVs explained a large part of the variance of the data.The cumulative variance of the fi rst three axes in CCA2 was 22.7% for species data and 75.6% for the relationship between species and habitat variables (Table 3).Prado et al. (2002) argued that, in many cases, the fi rst two or three axes explain a large  part of the variance (60-90%), which enables to use the results to describe the system without significant loss of information.Thus, even when the first two axes only are considered, the percentage values obtained in this study are high.We consider that the use of HDVs provided parameters that allowed us to interpret and discuss the floristic composition of the sampled communities more widely, since in addition to the altitude, other habitat features were added.With the largest number of variables were obtained parameters allowed to extend the discussion.In addition to corroborate the influence of altitude the variables used showed characteristics of communities that extended their ecological interpretation.
understanding of the infl uence of the variables used on vegetation patterns.

Figure 1
Figure 1 Detrended Correspondence Analysis (DCA) diagram of the distribution of arboreal and sub-arboreal species sampled at two altitudes in the Mantiqueira range, municipality of Itamonte, Minas Gerais state, Brazil.A = points at 1500 m; and B = points at 1900 m.Figura 1 Diagrama gerado pela Análise de CorrespondênciaRetifi cada para a distribuição das espécies arbóreas e subarbóreas amostradas em duas cotas altitudinais da Serra da Mantiqueira no município de Itamonte, Minas Gerais, Brasil.

Figure 2
Figure 2 Canonical Correspondence Analysis (CCA) diagram of the distribution of arboreal and sub-arboreal species sampled at two altitudes in the Mantiqueira range, municipality of Itamonte, Minas Gerais state, Brazil.A = points at 1500 m; and B = points at 1900 m.Figura 2 Diagrama gerado pela Análise de Correspondência Canônica para a distribuição das espécies arbóreas e subarbóreas amostradas em duas cotas altitudinais da Serra da Mantiqueira no município de Itamonte, Minas Gerais, Brasil.

Table 3
Summary of Canonical Correspondence Analysis (CCA) results of arboreal and sub-arboreal species sampled at two altitudes in the Mantiqueira range, municipality of Itamonte, Minas Gerais state, Brazil.Tabela 3 Resumo dos resultados da Análise de Correspondência Canônica para as espécies arbóreas e subarbóreas amostradas em duas cotas altitudinais da Serra da Mantiqueira no município de Itamonte, Minas Gerais, Brasil.

Table 4
Correlations between habitat variables and the fi rst three axis in the Canonical Correspondence Analyses (CCA1) for communities located at two altitudes in the Mantiqueira range, municipality of Itamonte, Minas Gerais state, Brazil.Tabela 4 Correlações entre variáveis ambientais e os três primeiros eixos da Análise de Correspondência Canônica (CCA1) para comunidades localizadas em duas altitudes na Serra da Mantiqueira, no município de Itamonte, Minas Gerais, Brasil.