How many vascular plant species are there in a local hotspot of biodiversity in Southeastern Brazil ?

Scientific information about the distribution of species richness and diversity is necessary for full comprehension of our evolutionary heritage forming a powerful tool for the development of nature conservation strategies. The aim of this article was to estimate the vascular plant species richness of the campos rupestres from the Itacolomi State Park (ISP) in order to verify the park ́s classification as a local hotspot of biodiversity and to outline the status quo of knowledge about biodiversity in the region. For that, the species richness of two phytosociological surveys of 0.15 ha each were extrapolated using (a) the species-area relationship fitted by the power and the logarithmic model as well as (b) the taxon ratio model. The taxon ratio model estimates total vascular plant species richness to 1109 species using seven different taxa. Extrapolations of different fittings of the species-area relationships calculate the complete park’s richness to values between 241 and 386 (logarithmic model), and 3346 to 10421 (power model). These extrapolations are far beyond realistic: the logarithmic model underestimates the park ́s species richness, because more than 520 vascular plant species have already been registered in the park. On the other hand, it is not plausible that one-fourth or more of all species registered so far in the Atlantic Tropical Forest Biome occur in the ISP as indicated by the extrapolation with the power model. The most realistic estimation was achieved by the taxon ratio model. Although confidence intervals are large, the species richness estimated by this method legitimates the classification of the ISP as a local hotspot of biodiversity, but this value also outlines the lack of knowledge about biodiversity in this region and the urgent need for more taxonomical and ecological research.


Introduction
Species richness and diversity are intriguing subjects that ecologists seek to understand in the context of evolutionary and ecological processes (e.g., Hubbell, 2001;Webb et al., 2002;Cavender-Bares et al., 2009;HilleRisLambers et al., 2012).For full comprehension of our evolutionary heritage (May, 2011), mankind is trying to answer questions such as "How many species are there on earth?" or even in a specific country or site (May, 1988;Scotland and Wortley, 2003;Mora et al., 2011).Such information is relevant because the knowledge about the distribution of extant species richness and diversity as phenomena provides powerful tools for the prediction of extinction rates caused by human impacts or climate change (Lewis, 2006), the development of conservation policies (Brooks et al., 2006;Killeen and Solórzano, 2008;Pardini et al., 2009;Ribeiro et al., 2009) and the declaration of superregional or local biological hotspots (Myers et al., 2000;Orme et al., 2005).As a piece of scientific information, the number of species (e.g., vascular plant) within a region is not only an important figure for nature conservation and landscape planning (Williams et al., 2004;Arponen et al., 2009;Gastauer et al., 2013), but it is also a measure for the state of current taxonomic and ecological research (Mora et al., 2011).
There are about 250,000 to 300,000 vascular plant species in the world (IUCN, 2007), nearly 20% of them occurring in Brazil (Giuletti et al., 2005;Lewinsohn and Prado, 2005).This is not a very precise evaluation, but its improvement requires enormous effort, that is far beyond the possibilities of many countries from the world.Achieving reliable numbers for species richness in a certain region is constrained by the fact that many species are neither registered for the region or -even worse -are not described yet.Lacking human resources to sample, identify, describe and archive all species from some regions form a bottleneck to evaluate all this diversity, especially in megadiverse and emergent countries as Brazil.This problem is aggravated by the fact that we are living in a period of extinction crises (Millenium Ecosystem Assessment, 2005;Lewis, 2006;Chen and Hui, 2009), so that many species might disappear even before they have been censused and described.Therefore, the complete species richness from the world, for states, provinces or regions is rather estimated by extrapolation than being the result of expensive, complete censuses (Raven, 1985;Grassle and Maciolek, 1992;Briggs and Snelgrove, 1999;Bebber et al., 2007;Joppa et al., 2010), although statistical methods to answer these questions have been discussed controversially (Mora et al., 2011;Colwell et al., 2012).The species-area relationship is one of the best examined patterns of ecology (Rosenzweig, 1995;Lomolino, 2001).Although area is not the only predictor of species richness (He and Hubbell, 2011), extrapolations based on this well accepted principle of macroecology are widely accepted (Condit et al., 1996;Pimm and Raven, 2000).The taxon ratio model proposed by Colwell and Coddington (1994), on the other hand, is little explored in scientific literature.As it uses the ratio of well-studied taxa within a small survey and a geographical entity to estimate whole species richness, it seems to be independent of sample effects.Extrapolations by both tools should reveal important insights about the distribution of and the knowledge about biodiversity in megadiverse Brazil.This country possesses two global hotspots of biodiversity on its territory, the Cerrado and the Atlantic Rainforest (Myers et al., 2000), which are characterized by high species richness and a high degree of endemism (Mendonça et al., 2008;Stehmann et al., 2009).Campos rupestres (rock-field), azonal montane ecosystems, replace vegetation of both hotspots usually above 900 m ASL in Southeastern Brazil (Magalhães, 1966;Ribeiro and Walter, 1998).The ecosystem is characterized by an open vegetation with a negligible tree layer (Alves and Kolbeck, 2010) and is bound to a mosaic of rocky outcrops and shallow white sands in the mountain ranges from Minas Gerais, Bahia and Goiás states of Brazil (Alves et al., 2007).Also showing similarity to other azonal outcrop vegetations from the Neotropics (Alves and Kolbeck, 2010), this species-rich and diverse ecosystem with many endemics (Conceição et al., 2007;Menini Neto et al., 2007;Gastauer et al., 2012) is composed by specific plant communities (Alves et al., 2007;Conceição and Pirani, 2005).The aim of this paper was to estimate the total number of vascular plants of all campos rupestres from the Itacolomi State Park, Minas Gerais, Brazil, by the taxon ratio model and by extrapolations of the species-area relationship.Both tools were stressed as much as possible in order to combine a suitable level of certainty with the lowest possible effort to match the challenge of biodiversity evaluation with available human effort.

Methods
This study was carried out in the Itacolomi State Park (ISP), Minas Gerais, Brazil.This park comprises 7543 ha in the South of the Espinhaço Mountain Range that are covered mainly by Seasonal Semidecidual Forests and campos rupestres (Drummond, 2005), the latter corresponding to 44% (3319 ha) of the area of the park (Fujaco et al., 2010).Within the park, two phytosociological surveys of 0.15 ha in Calais (20°25' S, 43°30' W) and Lagoa Seca (20°26' S, 43°29' W), each divided in 15 plots of 10 x 10 m, were carried out.Distance between both surveys is 2.7 km (Figure 1, Gastauer et al., 2012).To estimate vascular plant species richness of campos rupestres from the ISP, the taxon ratio model proposed by Colwell and Coddington (1994) was applied.For that, we assumed that the average ratio, i.e. S TT :S TS (S TT = total number of species from a single, wellstudied taxon from the ISP, and S TS = number of species from that taxon registered within a smaller survey), is constant across different taxa.From a triage of seven well-studied taxa from the park, an average taxon ratio was estimated for extrapolation purpose.From the distribution of taxon ratios, 95% confidence intervals (95% CI) were computed to evaluate statistical support of analysis.Furthermore, extrapolations of the species-area relationship from both surveys were generated.For that, we calculated the average number of species of groups of two, four, seven, and finally all the 15 plots of each survey.Only neighboring plots were grouped.Then, both speciesarea relationships were fitted and extrapolated to the whole area of campos rupestres from the ISP using the power model S = c * A z proposed by Arrhenius (1921) as well as the logarithmic function S = a * log(A) + b (Gleason, 1922;Rosenzweig, 1995;Dengler, 2009).Here, S is the number of species, A is the area, c, z, a, and b are the fitting parameters.The rarefactions were tested to reliability by (i) downscaled to one m 2 , (ii) by correlation coefficient, and (iii) by F statistics after linearization due to logtransformation of area (logarithmic model) and all values, respectively (power model).As sampled plots do not form a continuous area, but are arranged in three rows with distances within and between rows of ten meters (for details see Gastauer et al., 2012), we computed the speciesarea relationship considering two different areas as upper and lower limit for extrapolation.The upper limit is computed correlating observed species richness with the real sample area (100 m 2 per plot and 1500 m 2 per survey), while for lower limit the real sampled area is boosted by the area between plots resulting in an area of 400 m 2 per plot and 6000 m 2 per survey).For the calculation of fitting parameters, a complete linear least square fitting routine was run for linearized values (LINEST function from the Microsoft Excel TM package) that returns the mean values as well as their standard deviations (SD) for the fitting parameters.Inserting mean values in the equations of each model returns the expected number of species.95% CI for this extrapolation were calculated using mean values ±1.96 × SD of each fitting parameter.For lower bound of 95% CI, 1.96 × SD was subtracted from each fitting parameter, for upper bound, it was added.

Results
A total of 161 species were detected in both study sites, 23 of them occurring in both surveys (Table 1).Seventy-six species were sampled in Lagoa Seca and 107 in Calais.The taxon ratios for seven well studied taxa vary from 3.5 to 9.5 and amount 6.88 on average (95% CI 5.47, 8.30, Table 2).Under these assumptions, the total number of species within the park´s campos rupestres is estimated to 1109 species with 95% confidence intervals ranging from 881 to 1336 (Table 2).With correlation coefficients above 0.99, species-area relationships from both surveys were fitted significantly by the power as well as by the logarithmic model (Table 3).Downscaling the SAR to one m 2 produces negative species richness using the logarithmic model.According to the power model, the average species richness of one m 2 is between 2 or 3 (Lagoa Seca, lower and upper limit of estimation) and 2 to 4 (Calais), respectively.Extrapolating the logarithmic model to the whole park's campos rupestres area results in an overall plant species richness of 241 (from Lagoa Seca, lower limit) and 386 species (Calais, upper limit).
Using the power model, the number of vascular plant species within the park's campos rupestres is estimated between 3346 and 6109 (Lagoa Seca) or 5569 to 10421 (Calais, Table 3).

Discussion
Considering the studies cited in Table 2 plus the 161 species found within our two surveys, 520 species have already been detected within the park's campos rupestres.Further surveys (e.g.Peron, 1989;Casarino et al., 2009) might enrich the list.This fact discards the extrapolations from the species accumulation curve using the logarithmic model because they underestimate the vascular plant species richness of the campos rupestres from the ISP below the number of already registered species.This interpretation is supported by the fact that downscaling of this model leads to negative species richness in one m 2 -this is definitely not plausible.But the extrapolations of the species-area relationships by the power model should be analyzed with caution as well, although this model is well studied and generally well accepted (Brown et al., 2002;Martin and Goldenfield, 2006).Good fittings and plausible results for downscaling lead into temptation to upscale these models.Nevertheless, it is not likely that one-fourth of all species from the Atlantic Tropical Forest (Stehmann et al., 2009) or more coexist in the small ISP.This overestimation by the power model as well as the underestimation by the logarithmic model are due to the fact that the true species-area relationship in a log-log space is triphasic and of sigmoid form (Rosenzweig, 1995) with scale-dependent inclination (Condit et al., 1996;Conner and McCoy, 2001;Hubbell, 2001;Condit et al., 2004;Dolnik and Breuer, 2008).For extrapolations from a very small survey to very large areas as in our case, we need models that reduce the inclination in dependence of scale, perhaps in function of similarity as proposed by Krishnamani et al. (2004).To apply their calculations, we need more study sites within the ISP, which contradicts our goal to extrapolate vascular plant species richness of the ISP's campos rupestres from minimal sampling effort.
Another possibility to avoid the problem of scale dependent inclination is the application of fix inclinations that have been found in other studies comprising more magnitudes.Rosenzweig (1995) suggests an inclination of 0.25 for linearized species area relationships of terrestrial plant communities, which is generally well accepted (Pimm and Raven, 2000).The insertion of this value in the extrapolation equation estimates the overall vascular plant species richness of the ISP to 11163 , the same magnitude as the taxon ratio model.But this must be coincidence.What remains is the taxon ratio model.With that, the vascular plant species richness of all campos rupestres Table 2. Taxon ratios between total number of species found within the Itacolomi State Park, Minas Gerais State, Brazil, and number of species found in two phytosociological surveys including estimation of the park's plant species richness as proposed by Colwell & Coddington (1994).CI: confidence interval.from the ISP is estimated to the magnitude of 1100 species.Although the 95% confidence intervals are high, this seems to be a realistic value, as more than 520 species have already been registered and censuses of many species-rich and abundant families like Poaceae, Cyperaceae, Velloziaceae and Rubiaceae are still lacking.Surely, there are many first records for the ISP and not yet described species waiting for their discovery (Almeida, 2008;Almeida and Carvalho-Okano, 2010).Giulietti et al. (1997) guessed that there might be found about 4000 plant species within the Espinhaço Mountain Range, where most of Brazilian campos rupestres are situated.On the other hand, Stehmann et al. (2009) compile a -still incomplete -list of 14,000 plant species that occur in the Atlantic Rainforest biome.From that, 3,740 species occur in rocky formations.Further species are still to be registered or to be described.Low similarity of nearby study sites (Lemes, 2009;Gastauer et al., 2012) indicate that overall campos rupestres vascular plant species richness should be higher than ISP's richness and give confidence and plausibility to our extrapolations of the taxon ratio model.We estimated that there might be found about 1100 vascular plant species in the campos rupestres of the ISP on an area of only 33 km 2 .Except some studies not considered in our count (Peron, 1989), the occurrence of only 520 species has already been published.This shows that there is still a lot of taxonomic work to be done to achieve a complete list of vascular plants of the ISP.The expected species richness of the ISP corresponds to about 2% of the Brazilian flora and about 0.004% of all vascular plants from the world.It also corresponds to about one-fourth of all species from the campos rupestres ecosystem.This emphasizes the exceptional species richness of campos rupestres in general, but especially in the Southern Espinhaço Moutain Range (Drummond, 2005), and underlines the importance of the ISP for the conservation of biological resources.Furthermore, these figures justify the park's declaration as a (local) hotspot of biodiversity.Beside an evaluation of the species richness in an important reserve and the declaration of the ISP as a local hotspot of biodiversity, our findings provide a better assessment of extant species richness and diversity by a well-known tool.Nevertheless, there still remain some uncertainties, so that the precision of this estimation and the knowledge about biodiversity and its underlying evolutionary and ecological processes in the region should still be improved.

Figure 1 .
Figure 1.Geographic location of the study sites in the Itacolomi State Park (ISP), Minas Gerais, Brazil, and its position in South America (upper left) and Southeastern Brazil (lower left).

Table 1 .
Species found in phytosociological surveys of 0.15 ha carried out in Lagoa Seca and Calais, Itacolomi State Park, Minas Gerais State, Brazil.For geographical position of study sites, see Figure1.Species are sorted by their affiliation to families.

Table 3 .
Power and logarithmic fittings of the species accumulation curves of the surveys from Lagoa Seca and Calais and their extrapolations to the overall vascular plant species richness of the park's campos rupestres, Itacolomi State Park (ISP), Minas Gerais State, Brazil.# (1m 2 ) is the species richness expected in one m 2 .Values in brackets are lower and upper bounds of 95% confidence intervals, SD is standard deviation for each fitting parameter retrieved from a complete linear last square curve fitting routine.