ARBOREOUS SPECIES POPULATION IN A TOPOGRAPHIC GRADIENT OF MIXED OMBROPHILOUS FOREST IN RIO GRANDE DO SUL

This study was part of the Long Term Ecological Project (PELD) conducted in the São Francisco de Paula National Forest, RS where Mixed Ombrophilous Forest is predominant. The aim was to identify distribution of arboreous species depending on different soil conditions occurring in the low, medium and upper slopes of the studied area. 180 sample units of 100 m distributed along the topographical gradient were installed, with identification of the totality of arboreous species with 3 cm ≤ CBH < 30 cm and height exceeding 1.30 meters. In addition, thirteen environmental variables were collected in each sub-unit. The Canonical Correspondence Analysis model (CCA) indicated that only the slope, the topographic position, the stoniness and the soil color (up to 50 cm in depth) were the variables that best distinguished the sample groups of the slope environments. Out of the 53 species included in the model, ten species showed best correlations with the typical environments of upper slopes and 12 species were most characteristics of the lower positions, while the remaining resulted well adapted to different hillside topographic conditions. Environmental preferences of the species could be considered in silvicultural and recuperation programs applied to hillside environments of the region.


INTRODUCTION
In Rio Grande do Sul, Mixed Ombrophilous Forest is characterized by tropical (Afro-Brazilian) and temperate (Austral-Antarctica and Andean) flora, with great physiognomic relevance of Coniferaceae and Lauraceae elements (MARCHIORI, 2002).This forest structure suffered great environmental degradation and reduction of its original area of distribution mainly due to forest resources extraction, in particular for the species Araucaria angustifolia (Bertol.)Kuntze, whose timber has great commercial value (LADEIRA, 2002).Thus, development of scientific researches in the area of ecology and silviculture is necessary to subside sustainable forest management programs and conservation of biodiversity in this type of forest.
The way a forest regenerates depends on mechanisms helping seed germination to establish new individuals and species (ÁVILA et al., 2011), like seed rain, soil seed bank and the seedlings bank.These mechanisms may be influenced by different types of microclimates within the forest, name "microsites", which act directly on the forest regeneration process.
Therefore, in the analysis of forest inventories, the application of multivariate techniques is very important, to demonstrate the relationship between sampled species and common microsites of the studied community, thus providing information about autoecology of forest species and so great technical bases to the development of a sustainable forest management (BARIK et al, 1996).The Canonical Correspondence Analysis (CCA) have been a frequently used multivariate technique in Brazilian forest studies, to describe relationships between plant species and various environmental variables (OLIVEIRA-FILHO et al., 1998;PINDER;ROSSO, 1998;BLANC et al., 2000;SCIPIONI et al., 2010;GRAF NETO, 2011).
In this way, Narvaes et al. (2008) analyzed the floristic composition of natural regeneration in the Mixed Ombrophilous Forest of the National Forest (FLONA) in São Francisco de Paula, RS and demonstrated, through grouping analysis, the presence of three floristic groups which could be related to the different environments encountered along a topographic gradient: lower, medium and higher hillsides.
Considering diversity and complexity of the microhabitats studied in this analysis, this study is based on the hypothesis that naturally regenerating species behave according to a preferential colonization/distribution standard, following the microsites conditions which depend on the topographic gradient variation.
In this context, this study aimed to identify the relationship between the analyzed sub-unities and sampled species, depending on the environmental factors encountered in the lower, medium and higher portions of a hillside, physically defined as permanent stands in the FLONA of São Francisco de Paula, RS.
Terrain relief is undulating in the north side with 930 m height, and in the south side is hilly and forms more than 100 m deep Canyons, characteristic of the area where the study was conducted (SANTOS et al., 2013).However, analyzing the location of the permanent stands, their height and corresponding mean declivity, variation was from 844 to 911 meters and from 4.7 to 40.3 % respectively, demonstrating great relief variation in the studied area.Height and declivity values were extracted using a digital topographic model, originated by refinement by kriging of SRTM -Shuttle Radar Topography Mission data, called TOPODATA, with a spatial resolution of 30 m (VALERIANO; ROSSETI, 2012).Furthermore, according to the classification by Santos et al. (2013), crossing the location of permanent stands with declivity classes proposed, using derivation generated from TOPODATA, it was highlighted that permanent stand 1 is located in a relief with 3-8% declivity (gentle slope-undulating); stand 2 in a relief with 20-45% declivity (steep-undulating); stands 3 and 4 in a relief with 8-20% declivity (undulating); while stands 5 and 6 in a relief with 20-45% declivity (steep-undulating).
More specifically, when micro reliefs within each stand are analyzed, sub-unities of the lower hillside have gentle declivity, while in the middle hillside the wide majority of stands vary between medium and steep declivity, although many gentle declivities were observed in the analyzed sub-unities (10 x 10 m).In the upper hillside almost all sub-unities presented steep declivity.

Collection and analysis of data
Inside the studied area, 6 permanent stands were previously chosen, with average linear distance of 1.200 meters each other, in a way that they initially include a wide variation of sites, characterizing lower, middle and upper positions of the analyzed hillsides.Those permanent stands (100 x 100 m unities) belong to the Long Term Ecological Project (PELD).
In 2002, thirty sub-unities 10 x 10 m (100 m 2 ) were drawn in each stand (total 1.8ha), distributed in 90 sub-unities in the lower position of the hillside, 60 sub-unities in the middle hillside and 30 subunities in the higher hillside.In each sub-unity, all arboreal individuals taller than 1.30m and with circumference at breast height between 3 cm ≤ CBH < 30 cm were sampled.In this way, the evaluated arboreal component included young individuals (in regeneration) and adult specimens (established) of the community.Species with supporting and climbing habit were grouped as lianas.Those individuals were numbered and, when not identified in field, botanic material was collected to be compared to dried samples in the Herbarium of Forest Sciences Department (HDCF) of the Universidade Federal de Santa Maria.Taxonomic classification of species followed the Classification System of Angiosperm Phylogeny Group III (APG III, 2009).
To analyze correlations between environmental and vegetational gradients, the Canonic Correspondence Analysis (CCA) (TER BRAAK, 1987) was applied, using PC-ORD software version 3.17 for Windows (MCCUNE; MEFFORD, 1997).CCA is based on value of chi-squared distance, such that samples are weighted according to their totals, which provides heavy weights for species whose total density in data matrix is low, this way exaggerating distinction of samples containing various rare species (MC CUNE;MEFFORD, 1995;1997).Thus, the matrix of density of the species was formed by the number of individuals per stand of species with ten or more individuals detected.Following recommendations of Ter Braak (1995), density values (x) were transformed by the expression ln (x+1) to compensate deviations caused by some high values.Dead individuals, unidentified (three species) and low density individuals (< 10 ind.) and with similar densities in the three hillside positions (eight individuals), were discarded since they did not contribute to CCA calculations.A total of 57 species contributed to establish correlations with the environmental variables (Table 1).
Humidity was introduced in the environmental variables as a form to assess the hypothesis that water and dissolved mineral nutrients could be determining the distribution of forest species (BOTREL et al., 2002).Also variables as basal area and frequency distribution of growing stock were assessed.Original from the PELD project, these two variables could present direct influence on composition and distribution of the natural regeneration, through seed rain, directly responsible for the appearance of seedlings in the remaining natural forest (CHAMI et al., 2011).
Following the statistical procedure established by Botrel et al. (2002), a preliminary CCA was applied, discarding eight environmental variables strongly correlated or highly redundant with other variables.Of all the sampled species, the ones which presented similar density values in two or more topographic positions were discarded from the analysis, since they could cause the increase of residual variance that is not explained in the final model (TER BRAAK, 1995;MC CUNNE;MEFFORD, 1997).
Final CCA was processed with the five most representative variables, more strongly correlated with the ordination axes, in decreasing order of importance: topographic position, soil color up to 50 cm depth, humidity, declivity, stoniness and soil depth.Significance between matrixes was verified by the Monte Carlo permutation method, using 1,000 permutations, aiming to increase precision in calculation of probability.

Floristic composition
Database of species considered in the CCA included 53 arboreous species belonging to 22 botanic families.

Ordination of environmental variables and species.
CCA eigenvalues for the three ordination axes were 0.233 (axis 1), 0.159 (axis 2) and 0.086 (axis 3), representing respectively 6.6% (axis 1), 4.7% (axis 2) and 2.5% (axis 3) of their variability.On the whole, the three axes explained 14% of the global data variance, indicating high "noise" or unexplained residual variance.Nevertheless, significance of relationships species-environment was not impaired, since CCA produced considerable correlations in the three axes: 0.786 (axis 1), 0.706 (axis 2) and 0.699 (axis 3), highlighting balanced contribution of the three ordination axes.Those values of the coefficient of the generated model can be considered similar to the study of Botrel et al., (2002), for ten chemical and textural soil variables and two topographic variables, and inferior to another study conducted by Martins et al., (2003), for eleven chemical and textural variables, where texture classes were assessed depending on different contents of coarse sand, thin sand, silt and clay (SANTOS et al., 2013).Both studies had the objective to ultimately assess the effects of topographic variations on the floristic composition of semievergreen seasonal forest environments, in different regions.
Monte Carlo permutation test, for the three ordination axes, demonstrated that correlations between density of species in the sampling sub-units and environmental variables were significant at 5% level of probability.These results indicate the existence of an environmental gradient, with some species confined to specific sampling sub-units along this gradient.
The more strongly correlated environmental variables, that is to say the ones with greater adjusted correlation values (TER BRAAK, 1987) with the first axis were, in decreasing order, topographic position and declivity, while soil color (≤50 cm) and stoniness emerged with the second axis.The third ordination axis had soil humidity and soil color as the environmental variables with the best correlation.Fávero et al. (2009); Gotelli and Ellison (2011) explain that these weighted correlations are obtained through a linear relationship, elaborated by correlation or variance/covariance matrixes, which reflect and maximize only linear relationships between two groups of variables, both metric and not metric.
Table 2. Canonical correspondence analysis (CCA) with canonical coefficients and internal correlations ('intraset') in the two first ordination axis and weighted correlation matrix for the environmental variables used in the analysis.Tabela 2. Análise de correspondência canônica (ACC) com os coeficientes canônicos e as correlações internas ('intraset') nos dois primeiros eixos de ordenação e a matriz de correlações ponderadas para as variáveis ambientais utilizadas na análise.Weighted correlations showed strong interrelations between some variables, particularly between topographic position and declivity, and between topographic position and soil color (<50 cm).This confirms the proposition verified by Cattanio et al., (2002), that physical and chemical properties of soil are in relationship with the variation of water content.
Ordination of sampling sub-units in the two main axes of the model demonstrated the existence of sub-unities groups which represented very well the variability of the three hillside environments (topographic position) and of declivity, and the consequent edaphic variation between the stands, mainly in terms of soil color and stoniness (Figure 1).
Considering the environmental characteristics, the darker soil tones observed in sub-unities of the lower hillside portion (sub 61 to sub 90 = sub-unities sequentially numbered, samples collected inside each sampling unit of PELD) indicate possible higher organic matter contents.According to Vieira et al., (1988), the redox environment, with poorly drained lowlands, generally present less stony and darker soils, due to organic matter content.Moreno and Schiavini (2001) also demonstrated that forest microenvironments subjected to soil flooding allow accumulation of organic matter and thus soil color tends to be darker.
With the increase of topographic position and declivity, sub-units of the middle hillside (sub 121 to sub 180) suffer great alterations in their edaphic characteristics and hydric availability (reduction of water content), and present a varied range of soil color conditions and stoniness (Figure 1).In these terms, their behavior brought changes in soil color from grayish in the lower hillside regions to reddish and dark red in the middle hillside.Hence, the great variation in stones and boulders quantity exposed to soil surface (stoniness), in middle hillside sampling sub-unities, is due to different declivities which define distinct movement intensities of stone blocks falling from the upper hillside.
In the ordination graphic, as can be seen, some sampling sub-units of the lower and middle hillside appear spatially close each other (Figure 1), which indicates environmental similarity of these two topographic positions.The sampling sub-units of the upper hillside topographic position (sub 31 to sub 60) were distributed in the positive quadrant of the first ordination axis and were positively influenced by topographic position (0.835) and declivity (0.715), causing changes of soil color in the surface layer (-0.771) and superficial stoniness (-0.552) (Table 2).This behavior was expected, because according to Streck et al. (2008); Santos et al. (2013), elevation of the topographic position is associated to environments with lower organic matter content and reddish tones, caused by presence of hematite.
In the lower hillside position, eleven forest species showed relationship with this topographic condition, and these too are registered in other studies as having the same environmental preferences associated to more humid sites: Stillingia oppositifolia (BACKES; NARDINO, 1998), Picramnia parvifolia Engl.(CARVALHO et al. 2009), Myrcia oligantha O. Berg (LÓPEZ et al. 2006), Myrceugenia myrcioides (Cambess.)O. Berg, Calyptranthes concinna, Campomanesia rhombea, Eugenia subterminalis DC., Eugenia uruguayensis Cambess.(LEGRAND; KLEIN, 1969KLEIN, , 1970KLEIN, , 1978)), Myrceugenia cucullata, Myrciaria floribunda (West ex Willd.)O. Berg, Myrrhinium atropurpureum Schott (MARCHIORI; SOBRAL, 1997) and Lonchocarpus nitidus (Vogel) Benth (SILVA; TOZZI, 2012).This standard of ecological colonization adapted to humid environments is proved by the great density of individuals (Table 1).A great number of species did not show positive correlations with the edaphic characteristics demonstrated in stands/positions of the hillside (Figure 2).This lack of correlation may be partly explained by edaphic similarity registered between the lower and middle hillside sub-unities, demonstrated by their spatial proximity in the bi-dimensional ordination graphic (Figure 1).On the other hand, lack of correlation may indicate that forest species may present great establishment and growth adaptability to these edaphic variables, measured and selected in the model, thus not limiting their colonization only in the lower or middle hillside positions Indication of the existence of three floristic groups in the same studied area (species typical of the lower, middle and upper hillside positions), demonstrated by Narvaes et al. (2008) by cluster analysis (Twinspan), revealed similar relationship to results of the Canonical Correspondence Analysis (Figure 1).
The preferential colonization standard of few species to an existent topographic position was mainly determined by declivity, which directly affected stoniness and soil color in the first layer.The same kind of colonization standards of species restricted to determined environmental conditions, depending on different topographic positions, using CCA in different native forest formations, were also proved by studies of Oliveira-Filho et al (1998); Pinder and Rosso (1998); Moreno and Schiavini (2001);Botrel et al. (2002);Cardoso and Schiavini (2002), Martins et al. (2003) and Scipioni et al. (2010).
Thus, the model proved to be efficient, since it clearly distinguished groups of sampling subunities typical of each topographic position, although did not reveal clear distribution standards for the majority of species along the altitude gradient.This condition reveals that the majority of species presented wide edaphic adaptation to different environments in terms of establishment and growth.Meanwhile, the model proved that there is a group of species with specific environmental adaptations (declivity, stoniness and soil color) to the lower and upper hillside positions.

CONCLUSIONS
Basing on results obtained it was concluded that:  Sampling units of the forest inventory characterized the variations of the hillside topographic gradient.Darker and deeper soils (richer in organic matter) prevailed in the lower hillside portions when compared to upper/top hillside portions, which were thinner and with less organic matter content.Middle hillside portions presented intermediate environmental conditions, however, they showed more similarities with the lower hillside positions. Out of the 53 studied species, 22 (41.5%)presented a distribution along the topographic gradient which was conditioned by declivity variations, stoniness and soil color (50 cm depth).Species which presented greater correlations with the upper hillside conditions were: Mollinedia elegans, Myrciaria delicatula, Podocarpus lambertii, Ocotea pulchella, Ilex dumosa, Duranta vestita, Rudgea parquioides, Siphoneugena reitzii, Lamanonia ternata and Zanthoxylum petiolare.Species which presented greater correlations with the lower hillside conditions were: Stillingia oppositifolia, Picramnia parvifolia, Myrcia oligantha, Myrceugenia myrcioides, Calyptranthes concinna, Campomanesia rhombea, Eugenia subterminalis, Eugenia uruguayensis, Myrceugenia cucullata, Myrciaria floribunda, Myrrhinium atropurpureum and Lonchocarpus nitidus.All other considered species might be considered as widely adapted to the different hillside conditions. Preference of these species for determined soil condition in different hillside positions should be taken in consideration by the different recuperation/reforestation or silvicultural programs, when conducted in irregular relief areas of the Mixed Ombrophilous Forest in Rio Grande do Sul.

Figure 2 .
Figure 2. Ordination diagram of species in the first two axes held by the canonical correspondence analysis, based on density distribution of species under natural regeneration process in 180 sampling sub-units of the Mixed Ombrophilous Forest at São Francisco de Paula National Forest, RS, Brazil.Figura 2. Diagrama de ordenação das espécies nos dois primeiros eixos produzidos pela análise de correspondência canônica, baseada na distribuição da densidade de espécies da regeneração natural em 180 subunidades amostrais da Floresta Ombrófila Mista na FLONA de São Francisco de Paula, RS, Brasil.
Cod : code of the species in the CCA ordination axis.Low.: lower hillside, Mid: middle hillside and Upp.: upper hillside.