Floristic and vegetation structure of a grassland plant community on shallow basalt in southern Brazil

Few studies have adequately described the fl oristic and structural features of natural grasslands associated with shallow basalt soils in southern Brazil. Th is study was carried out on natural grazing land used for livestock production in the municipality of Santana do Livramento, in the Campanha region of the state of Rio Grande do Sul, Brazil. Th e aim of the study was to describe the fl oristic and structural diversity of the area. Th e fl oristic list obtained comprises 229 plant taxa from 40 botanical families, with a predominance of the families Poaceae (62), Asteraceae (28), Fabaceae (16) and Cyperaceae (12). Th e estimated diversity and evenness in the community were 3.00 and 0.874, respectively. Bare soil and rock outcrops accounted for 19.3% of the area, resulting in limited forage availability. Multivariate analysis revealed two well-defi ned groups among the sampling units. One group showed a high degree of internal aggregation, associated with deep soils, and was characterized by the presence of tussocks, whereas the other was less aggregate and was characterized by prostrate species growing on shallow soil. Ordination analysis indicated a gradient of moisture and of soil depth in the study area, resulting in diff erent vegetation patterns. Th ese patterns were analogous to the vegetation physiognomies described for Uruguayan grasslands. Overall, the grassland community studied is similar to others found throughout southern Brazil, although it harbors more winter forage species. In addition, the rare grass Paspalum indecorum Mez is locally dominant in some patches, behaving similarly to P. notatum Fl., a widespread grass that dominates extensive grassland areas in southern Brazil.


Introduction
Natural grasslands growing on shallow basalt soils cover approximately 4.5 million ha in southern Brazil (primarily in the Campanha region of the state of Rio Grande do Sul) and Uruguay (Hasenack et al. 2010).At present, most of this area is in relatively good conservation status and constitutes the main source of forage for regional beef cattle production (Bilenca & Miñarro 2004).Natural grasslands also provide many other environmental benefits, such as fixation of atmospheric carbon (Guterres et al. 2006); nutrient storage and cycling; regulation and maintenance of water cycles and water quality; prevention and control of soil erosion; and scenic beauty (Tornquist & Bayer 2009).In addition, due to the prevailing winds in the grassland-dominated ecosystems of southern Brazil and Uruguay, wind power plants have recently been installed in the region, contributing to clean energy generation.Because natural grasslands play such important roles, increasing our knowledge of these ecosystems and preserving them has become an ethical duty for farmers and researchers alike.
In livestock production systems that rely on heterogeneous and complex environments such as natural grasslands, identifying and understanding the biology of key plant species are prerequisites to inform decisions regarding management and sustainability, because species composition influences plant production dynamics at temporal and spatial scales (Soares et al. 2005).The result is the rise of a genuinely symbiotic integrative productive system, which combines the need to produce and the need to continue producing in the future.According to Hodgson (1990), the foundation of the management of natural grazing lands is to reconcile livestock nutritional demands with maintenance of the productive potential of grazed plants.In order to do so, it is imperative to gather knowledge of the composition and structure of the natural grazing land.
Despite the ecological and economic relevance of grasslands on shallow soils in Brazil, there have been only a few studies evaluating such areas, all of which have been carried out in neighboring Uruguay (Rosengurtt 1979;Berretta 1998;Lezama et al. 2006;Lezama et al. 2010).The objective of the present study was to characterize the floristic composition and structure of the plant community of natural grasslands on Lithic Ustorthents derived from basalt in southern Brazil, providing a representative floristic list for this formation.

Material and methods
The survey was carried out on private property in a rural area (Santo Antônio Farm, 31°03'29"S; 55°55'26"W, 327 m above mean sea level), in the Coxilha Rica district of the municipality of Santana do Livramento, in the state of Rio Grande do Sul, Brazil, in the physiographic region locally known as Campanha (Stammel 1996).Natural grasslands in the area are historically used as the main forage source for livestock production, as in most of the preserved natural grassland ecosystems in the region (Nabinger et al. 2009).According to the Köppen climate classification system, the climate in the study area is type Cfa, theoretically with no dry season, although there is a high probability of soil water deficit during the peak of summer (Leivas et al. 2006).The soil in the sampling area is a Lithic Ustorthents (Neossolo Litólico Eutrófico by the Brazilian Soil Classification System, Pedregal mapping unit; EMBRAPA 1999).Physically, the soil has low drainage and water storage capacity, featuring the frequent presence of small loose stones and larger rock outcrops.Grassland vegetation at the study site is heterogeneous, presenting different physiognomic patterns that are visually distinguishable in terms of the overall structure and species composition, as well as the presence of bare soil and rock outcrops.Based on these criteria, the studied site was visually classified a priori into four categories, or structural patterns: "soft", "open", "closed" and "indecorum".
In relief depressions with deeper soils, vegetation was dominated by tall erect species adapted to high humidity, which usually bear leaves with small amounts of silica and fibers, thus characterizing the "soft" pattern.At hilltops and on hillsides, overall vegetation height was smaller and encompassed three physiognomic patterns distributed in patches: the "closed" pattern, showing dense vegetation cover, with almost no bare soil or rocks; the "open" pattern, with large amounts of bare soil or rock outcrops; and the "indecorum" pattern, comprising vegetation patches dominated by the grass Paspalum indecorum.
Data collection for the structural survey was conducted in 90 square sampling units (SUs) of 0.25 m 2 each, distributed over 77.3 ha of natural grasslands.We used preferential SU allocation to sample the four physiognomic patterns proportional to their representation in the landscape (Matteucci & Colma 1982).An exhaustive floristic list was prepared prior to and during the structural survey by collecting plant specimens outside the SUs throughout the study area.The surveys were carried out during the spring of 2009.Specimens were identified by consulting the specific taxonomic literature (Barros 1960;Burkart 1969Burkart , 1974Burkart , 1979Burkart , 1987;;Delprete et al. 2004Delprete et al. , 2005;;Burkart & Bacigalupo 2005;Lourteig 1983;Rosengurtt et al. 1970; besides many localized taxonomic revisions and unpublished dissertations and theses) and by consulting specialists.The resulting a priori classification of the 90 SUs into one of the four physiognomic patterns was as follows: 36 "closed"; 44 "open"; 12 "soft"; and eight "indecorum".The robustness of this a priori classification was tested using multivariate analyses (see below).The sampled area has been historically managed with domestic herbivores.For all species, bare soil and rock present in each SU, cover was estimated using an adaptation of the decimal scale devised by Londo (Londo 1976).Respective scale value/real cover value were: 0.1/<1%, 0.5/1.1-5%,1/5.1-10%, 2/10.1-20% and so forth in decimal intervals up to 10/ 90.1-100%.
For each species, we calculated absolute and relative frequency; absolute and relative cover; and importance value (IV).For the plant community as a whole, we calculated the Shannon diversity index (H' , using the natural logarithmic base) and Pielou's evenness index (J' -Pielou 1969;Müeller--Dombois & Ellenberg 1974;Magurran 1988).
The raw data matrix was submitted to multivariate analyses using MULTIV software (Pillar 1997).Cluster analysis was carried out with the sum of squares as the clustering criterion (Ward 1963;Orlóci 1967).After cluster analysis, we looked for well-defined groups among the SUs using the method proposed by Pillar (1999a).We employed principal coordinate analysis using chord distance between SUs.We evaluated stability and significance of the first five ordination axes after 10,000 iterations of bootstrap resampling (Pillar 1999a;1999b).The variables "bare soil" and "rock cover" were not included in these analyses but were inserted in the ordination diagram after computing the correlation between their vectors and the ordination scores of SUs (first two axes).

Results and discussion
The floristic survey identified 229 taxa, belonging to 138 genera and 40 families (Appendix 1).The most representative families were Poaceae, with 62 taxa, and Asteraceae, with 28 taxa, accounting for 26.8% and 12.1% of the total species richness, respectively, compared with 6.9% and 5.2%, respectively, for Fabaceae and Cyperaceae, both of which are worth mentioning because most of their taxa constitute important forage sources.Collectively, Poaceae, Asteraceae, Cyperaceae and Fabaceae accounted for 50.9% of the taxa recorded in the survey.Twenty-two other families presented two to seven taxa each, accounting for 43.1% of the total species richness, and another 14 families presented one taxon each, accounting for the remaining 6%.In the structural survey of the 90 SUs, we identified 178 plant taxa (see Appendix 2 for the complete list with frequency and cover values for each variable).
For the plant community studied, the H' was estimated at 3.00 and the J' was estimated at 0.874.In a study of a grassland community on sandy soils, along the banks of a seashore lagoon in southern Brazil, Boldrini et al. (2008) found similar diversity (H'=2.98)and slightly lower species richness (183).Working in that same region, Ferreira & Setubal (2009) found higher diversity (H'=3.63)and lower richness (126).In a grassland community on predominantly shallow soils on a granite hill, also in southern Brazil, Ferreira et al. (2010) found an H' of 4.5 and identified 282 plant taxa.All of those studies were carried out on sites historically submitted to grazing, burning or both, although some of them are not managed at the present.
Sandy soils and shallow stony soils are restrictive environments for the development of many species.In such environments, a small group of stress-tolerant taxa can dominate the community, thus reducing diversity and evenness.The estimated evenness for our studied community was similar to the 0.87 and 0.86 found by Ferreira & Setubal (2009) and Ferreira et al. (2010), respectively, and higher than the 0.68 found by Boldrini et al. (2008).The relatively similar diversity, evenness and overall richness among surveys conducted at different sites within the Pampa biome provide further evidence that it constitutes a discrete ecological unit (Ferreira & Boldrini 2011).
Bare soil and rocks accounted for 19.3% of the total sampled area.Similar values have been found for grasslands on shallow basalt soils in Uruguay (Lezama et al. 2006).Although this proportion is high, bare soil and rock cover are important, intrinsic features of Litholic Neosols (Stammel 1996) and should be taken into account when estimating forage availability.
Poaceae, the family with most taxa in our survey, accumulated 49.2% of the importance value (IV), compared with 15.35%, 1.57% and 5.46% for Asteraceae, Cyperaceae and Fabaceae, respectively (Fig. 1).The fact that Fabaceae, which ranked third in richness, accumulated more IV than did Cyperaceae, which ranked second, indicates that relative cover and frequency more accurately describe the importance of each botanical family in the community than do the raw values of richness.Similarly, Rubiaceae and Convolvulaceae, both of which had few taxa in our survey, showed relatively high IVs (Fig. 1).This is due to the fact that the mean frequency of taxa from these families was high in the community studied, as well as to morphological features such as leaf architecture, growth habit and plant size, which directly influence the amount of soil covered by each specimen.The remaining 32 families accumulated only 19.08% of the overall IV, although they encompassed a total of 75 taxa.
Taxa with an IV > 1.0 (n = 29) accounted for 63.9% of the total IV.Within this group were taxa defined as forage, inclu-ding five summer grasses (Paspalum notatum, P. indecorum, Mnesithea selloana, Andropogon lateralis, and Axonopus affinis); five winter grasses (Piptochaetium montevidense, P. stipoides, Vulpia bromoides, Briza minor and Danthonia cirrata); one winter legume (Trifolium polymorphum); and one summer legume (Galactia pretiosa).The co-existence of winter and summer forages enhances the potential of the natural grassland as a forage source for livestock production (Nabinger et al. 2009).Still among these taxa, four Asteraceae are worth mentioning (Micropsis spathulata, Soliva pterosperma, Chevreulia sarmentosa and Baccharis coridifolia), since they are especially found at sites with high proportional cover of bare soil and rock.Asteraceae species gain functional benefits from traits such as high seed production and adaptations for wind dispersal, which optimize colonization in such "open" micro-environments.
The four physiognomic patterns of grassland vegetation defined a priori were unevenly represented throughout the 77.3 ha of the sampling area (see methods), and mean percentages of bare soil, rock cover and grass species representativeness were variable among them (Tab.1).The "open" pattern was characterized by what is locally known as "campos duros" ("hard fields"), with considerable exposure of bare soil, usually accompanied by rock outcrops (Tab.1), suggesting shallow soil, a condition associated with hilltops and upper hillsides.The "closed" pattern was found mostly in association with the "open" pattern, on hilltops and hillsides, but on apparently deeper soils.Despite the spatial proximity, the two patterns are markedly different regarding the mean proportional cover of bare soil and rocks, and the representativeness of grasses (Tab.1).The "soft" pattern was mostly found on lower hillsides and in relief depressions near water courses and temporary bodies of water, where the soil is deeper.In this pattern, the mean proportional cover of bare soil is even lower, and the mean IV of grasses is strikingly high (Tab.1).The "indecorum" pattern comprised SUs in which there was a predominance of Paspalum indecorum and almost no rock outcrops but still lower grass representativeness in comparison with that of the "soft" pattern (Tab.1).In the "indecorum" pattern, P. indecorum is dominant even in the presence of species that dominate in other patterns, such as the grasses Andropogon lateralis and P. notatum, as well as the winter/spring legumes Trifolium polymorphum and Adesmia incana.
Cluster analysis identified two sharp groups among the 90 SUs (p>0.1).One group comprised all SUs previously included in the "soft" pattern, whereas the other one encompassed SUs included in the "closed" or "open" patterns.There were SUs of the "indecorum" pattern in both groups.In the principal coordinate analysis, the first ordination axis was significant (p<0.1) and the first two axes trended toward stability, denoting sampling sufficiency for pattern detection (Pillar 1998;1999a;1999b).In the resulting scatterplot, all 90 SUs were labeled according to the groups obtained in the cluster analysis (Fig. 2).The SUs  IV -importance value.
in cluster 1 are closer to each other, reflecting a high level of internal aggregation, whereas those in cluster 2 are more widely scattered, suggesting less internal aggregation.The same scatterplot is presented in Fig. 3 but with SUs labeled according to the a priori physiognomic patterns, as well as the variables with the strongest correlations (i.e., taxa) and the environmental variables bare soil and rock cover.As can be seen, the "soft" SUs on the upper left quadrant are gradually replaced clockwise by "closed" SUs, which in turn are replaced by "open" SUs.This ordination pattern not only reflects the transition between the predetermined physiognomic patterns but also probably reflects a moisture and soil depth gradient, to which grassland vegetation responds by establishing the patchy pattern observed.On the left-hand side of the scatterplot, there are SUs inserted on deeper and more humid soils, which are replaced clockwise by SUs on shallower and drier soils, up to the "open" SUs located at hilltops.The "indecorum" SUs presented no distinguishable pattern in the scatterplot.There are two plausible explanations for this behavior: apart from the marked dominance of Paspalum indecorum, these SUs encompassed plant taxa that were also present in the other patterns; or, unlike those of the other patterns, "indecorum" SUs were not associated with a specific topology.The latter argument suggests that the dominance of P. indecorum in some vegetation patches is independent of the observed moisture/soil depth gradient and must result from unobserved biotic/abiotic variables or naturally clumped population distribution.Paspalum indecorum is a relatively rare species, which is cited here for the first time in an ecological survey conducted in southern Brazil.It shows striking biological similarities to P. notatum, a widespread and typically dominant grass species.Apparently, the two species play similar functional roles, although why they alternately dominate vegetation patches, even at a local scale, is an unanswered question that merits further investigation.The distribution of variables best correlated with the ordination axes (Fig. 3) corroborates the idea that there is a moisture/soil depth gradient in the study area.Andropogon lateralis, Vulpia bromoides and Steinchisma hians, clustered near "soft" SUs, are species usually associated with deeper soils and more humid conditions, and the last two are C3 grasses.According to Lezama et al. (2010), this vegetation pattern constitutes "meso-hydrophytic grazing lands" in Uruguay, usually found in plain-concave portions of the terrain, with similar cover of winter and summer grasses.In our survey, however, the summer/winter grass ratio was 2:1 in the "soft" pattern.This might be explained by the  geographic position of southern Brazilian grasslands, which are farther north than are the Uruguayan grasslands and, consequently, encompass more tropical summer grasses (Boldrini & Longhi-Wagner 2011).
The "closed" pattern was characterized by Paspalum notatum, Piptochaetium montevidense, Richardia humistrata and Trifolium polymorphum (Fig. 3).The first three species are common in natural grasslands throughout southern Brazil (Overbeck et al. 2007), as well as in central Uruguay, especially in grazed areas (Altesor et al. 2005), and have bro-ad ecological amplitude, thriving under variable conditions, whereas T. polymorphum is a pampas species restricted to southernmost Brazil, Uruguay, Argentina and Paraguay, typically found in moderately drained soils.
The "open" SUs were characterized by the presence of Microchloa indica, Evolvulus sericeus, Ayenia mansfeldiana, Micropsis spathulata and Trachypogon montufarii var.montufarii (Fig. 3).All of those species are typically found in open grasslands throughout South America.In Uruguay, this vegetation is referred to as "lithophytic steppe" and is characterized by higher cover of summer species than winter species, usually in high relief positions (Lezama et al. 2010).Therefore, our "open" pattern seems to be analogous to this Uruguayan physiognomy, once again providing evidence that supports the classification of South American natural subtropical grasslands as a discrete ecological unit (Ferreira & Boldrini 2011).The gradual substitution of SUs belonging to different patterns and of taxa along the gradient explains why our cluster analysis identified only two well-defined groups.
The variables "bare soil" and "rock cover" were strongly correlated with the ordination axes (Fig. 3).Correlation values were positive for axis 1 and negative for axis 2, so that both variables were plotted towards the "open" (or "well--drained", considering moisture) extremity of the observed gradient.These high correlation values strengthen the interpretation of the observed pattern as a moisture/soil depth gradient.Alternatively, the high correlation values could be interpreted as evidence of a "community openness" gradient, although the two interpretations are not mutually exclusive.
Natural grasslands on basalt-derived Litholic Neosols show high floristic richness and plant diversity, similar to those found at other sites throughout southern Brazil but with greater representation of winter forage species.The presence and local dominance of Paspalum indecorum, a species scarcely found in other grassland communities, are also worth mentioning.
We detected an apparently natural structural variability in the vegetation, resulting in patterns that probably arose in function of soil conditions.These patterns match vegetation physiognomies described for Uruguayan grasslands, providing further evidence of an ecological unit comprising South American subtropical grasslands.
Finally, bare soil and rock cover represent a significant area in grasslands growing on Litholic Neosols.Because livestock farming is one of the most important economic activities in the Pampa biome, these variables should be considered when estimating forage availability per area on lands where this soil type is represented in the landscape.

Figure 1 .
Figure 1.Richness and importance value (IV) of plant families sampled in the structural survey.

Figure 2 .
Figure 2. Scatterplot obtained in the principal coordinate analysis of 90 sampling units (SUs) described by 178 plant taxa.SUs are labeled according to sharp groups resulting from a cluster analysis.

Table 1 .
Proportions of bare soil and rock cover and representativeness of grasses in each a priori physiognomic pattern.
Appendix 1. Floristic list obtained for the natural grassland in the municipality of Santana do Livramento, Brazil.Plant taxa found in the vegetation structural survey in 90 sampling units, with respective absolute and relative frequency and cover (AF, RF, AC and RC, respectively).