ArticlePDF Available

Abstract and Figures

The rocky montane savannas of South America, known as campos rupestres in Brazil, where they largely occur, represent a megadiverse habitat housing c.15% of the Brazilian vascular flora in less than 1% of the Brazilian territory. Amongst other factors, the remarkable plant diversity in campos rupestres has been attributed to its occurrence as many isolated patches and to floristic influences from surrounding habitats, including lowland woody savannas (cerrado), Atlantic rain forests, seasonally dry woodlands and Amazonian rain forests. However, no study has assessed the degree to which the putative floristic influence from surrounding habitats drives compositional variation in campos rupestres. Here, we used a dataset on the composition of South American woody plant communities (>4,000 community surveys, with >100 representing campos rupestres), combined with environmental data, with the aim of characterising and explaining compositional variation of the campos rupestres woody flora. Our results showed that all campos rupestres, including the sites occurring in Amazonian ironstone formations, are more similar to cerrado woody savannas than to any other South American vegetation formations covered in our dataset. Also, multiple campo rupestre floristic groups may be recognized based on distinct species composition and environmental conditions, primarily related to substrate and climate. We stress the importance of considering this floristic heterogeneity in conservation, management and research planning.
Content may be subject to copyright.
Elsevier Editorial System(tm) for Flora
Morphology, Distribution, Functional Ecology of Plants
Manuscript Draft
Manuscript Number: FLORA-D-17-00006R1
Title: Lack of floristic identity in campos rupestres - a hyperdiverse
mosaic of rocky montane savannas in South America
Article Type: SI: Plant life on campo rupestre
Keywords: campos rupestres, Espinhaço range, cangas, floristic
composition, environmental heterogeneity, cluster analysis
Corresponding Author: Dr. Danilo M. Neves, PhD
Corresponding Author's Institution: University of Arizona
First Author: Danilo M. Neves, PhD
Order of Authors: Danilo M. Neves, PhD; Kyle G Dexter, PhD; R. Toby
Pennington, PhD; Marcelo L Bueno, PhD; Pedro S Miranda, MSc; Ary T
Oliveira-Filho, PhD
Abstract: The rocky montane savannas of South America, known as campos
rupestres in Brazil, where they largely occur, represent a megadiverse
habitat housing c.15% of the Brazilian vascular flora in less than 1% of
the Brazilian territory. Amongst other factors, the remarkable plant
diversity in campos rupestres has been attributed to its occurrence as
many isolated patches and to floristic influences from surrounding
habitats, including lowland woody savannas (cerrado), Atlantic rain
forests, seasonally dry woodlands and Amazonian rain forests. However, no
study has assessed the degree to which the putative floristic influence
from surrounding habitats drives compositional variation in campos
rupestres. Here, we used a dataset on the composition of South American
woody plant communities (> 4,000 community surveys, with > 100
representing campos rupestres), combined with environmental data, with
the aim of characterising and explaining compositional variation of the
campos rupestres woody flora. Our results showed that all campos
rupestres, including the sites occurring in Amazonian ironstone
formations, are more similar to cerrado woody savannas than to any other
South American vegetation formations covered in our dataset. Also,
multiple campo rupestre floristic groups may be recognized based on
distinct species composition and environmental conditions, primarily
related to substrate and climate. We stress the importance of considering
this floristic heterogeneity in conservation, management and research
Response to Reviewers: Dear Drs. Silveira and Morellato,
The whole manuscript was revised in the light of the editor and
reviewers' recommendations. Our detailed revision is attached to the
cover letter as responses to the editor and reviewers’ comments.
Yours sincerely,
Danilo M. Neves
Article type: Original Article
Lack of floristic identity in campos rupestres a hyperdiverse
mosaic of rocky montane savannas in South America
Danilo M. Neves1, Kyle G. Dexter2,3, R. Toby Pennington3, Marcelo L. Bueno4,
Pedro L. S. de Miranda2, Ary Teixeira de Oliveira-Filho4
1Department of Ecology and Evolutionary Biology, University of Arizona,
Tucson, AZ 85721, USA. 2School of Geosciences, The University of Edinburgh,
Edinburgh EH9 3JN, UK. 3Royal Botanic Garden Edinburgh, Edinburgh EH6
5LR, UK. 4Programa de Pós-Graduação em Biologia Vegetal, Universidade
Federal de Minas Gerais, Belo Horizonte 31270090, Brazil.
Campos rupestres do not represent a single floristic group;
Influence from surrounding cerrados drives compositional differentiation
in campos rupestres;
Environmental conditions can predict the differentiation amongst campo
rupestre floristic groups;
Conservation units fail to protect important parts of the campo rupestre
floristic space;
campos rupestres and their surrounding lowland cerrados merit
simultaneous conservation attention.
*Highlights (for review)
Article type: Original Article 1
Lack of floristic identity in campos rupestres a hyperdiverse 3
mosaic of rocky montane savannas in South America 4
Danilo M. Neves1*, Kyle G. Dexter2,3, R. Toby Pennington3, Marcelo L. Bueno4, 6
Pedro L. S. de Miranda2, Ary Teixeira de Oliveira-Filho5
1Department of Ecology and Evolutionary Biology, University of Arizona, 9
Tucson, AZ 85721, USA. 2School of Geosciences, The University of Edinburgh, 10
Edinburgh EH9 3JN, UK. 3Royal Botanic Garden Edinburgh, Edinburgh EH6 11
5LR, UK. 4Laboratório de Ecologia e Evolução de Plantas, Departamento de 12
Biologia Vegetal, Universidade Federal de Viçosa, Viçosa 36570-000, Minas 13
Gerais, Brazil. 5Programa de Pós-Graduação em Biologia Vegetal, 14
Universidade Federal de Minas Gerais, Belo Horizonte 31270090, Brazil. 15
*Correspondence: Danilo M. Neves, Department of Ecology and Evolutionary 18
Biology, University of Arizona, Tucson, AZ 85721, USA. 19
E-mail: and 20
Keywords: campos rupestres, Espinhaço range, cangas, floristic composition, 47
environmental heterogeneity, cluster analysis. 48
1. Introduction 50
The rocky montane savannas of South America, known as campos 52
rupestres in Brazil, where they largely occur, are found on quartzite, sandstone 53
and ironstone formations, mostly above 900m (a.s.l.) and up to 2,033m (Giulietti 54
et al., 1997, Fernandes et al., 2014; Silveira et al., 2016). Its core area is spread 55
along the highlands of eastern Brazil (Giulietti et al., 1997; Hughes et al., 2013; 56
Silveira et al., 2016). Disjunct areas also occur along mountain ranges in 57
central-western Brazil (Frisby and Hind, 2014; Mews et al., 2014; Silveira et al., 58
2016), eastern Bolivia (Saravia, 2008) and in the Amazon forest (Silveira et al., 59
2016). The campos rupestres are a growing focus of attention because they 60
have been recently proposed as one of the world’s old climatically-buffered 61
infertile landscapes (OCBILs, e.g., the fynbos of the Cape Floristic Region; 62
Silveira et al., 2016). 63
In Brazil, the campos rupestres contain c.5,000 vascular plant species 64
(Reflora, 2016), corresponding to a remarkable c.15% of the Brazilian vascular 65
flora in less than 1% of the Brazilian territory (Fernandes et al., 2014; Silveira et 66
al., 2016). Amongst other factors, this outstanding floristic diversity in campos 67
rupestres a pattern common to OCBILs (Hopper et al., 2009) has been 68
attributed to its high levels of local endemism (Hensold, 1988; Echternacht et 69
al., 2011a) as well as to the geographically disjunct distribution of campo 70
rupestre sites and, hence, the associated floristic influence from distinct habitats 71
(Giulietti et al., 1997), namely cerrado woody savannas, Atlantic rain forests, 72
seasonally dry woodlands and Amazonian rain forests. However, to our 73
knowledge, no study has assessed the degree to which this alleged floristic 74
influence from surrounding habitats drives compositional differentiation of 75
campos rupestres. 76
Attempts to address this knowledge gap could be of importance for 77
effective conservation strategies. If the floristic variation of campos rupestres is 78
high, with multiple distinct floristic groups, future conservation assessments 79
could highlight, for instance, that a large number of separate conservation areas 80
are needed to fully protect campos rupestres diversity. Here we go a step 81
further in data refinement and analysis by using a large dataset on the 82
composition of South American woody plant communities (> 4,000 community 83
surveys, with > 100 representing campos rupestres), combined with 84
environmental data, in order to elucidate the spatial floristic patterns of campos 85
rupestres. We address the following hypotheses stemming from the literature 86
(Giulietti et al., 1997; Echternacht et al., 2011b): (h1) multiple campo rupestre 87
floristic groups may be recognized based on distinct species composition; (h2) 88
community composition differentiation amongst campo rupestre floristic groups 89
can be predicted by variation in environmental conditions. 90
2. Material and Methods 92
2.1. Study area 94
The South American rocky montane savannas (henceforth campo 96
rupestre for a single site, and campos rupestres for multiple sites) cover 97
c.65,000 km2 (Fernandes et al., 2014) and occur primarily on quartzite and 98
sandstone formations in the highlands (mostly > 900 m a.s.l) of eastern Brazil 99
as well as in scattered mountain ranges in central-western Brazil (Frisby and 100
Hind, 2014; Mews et al., 2014, Silveira et al., 2016). These highlands border 101
three primary vegetation ‘Domains’ (IBGE, 1993; Ab’Sáber, 2003): the Atlantic 102
Domain to the east and south (known as Mata Atlântica in Brazil), the Caatinga 103
Domain to the north and the Cerrado Domain to the west (see Giulietti et al., 104
1997; Hughes et al., 2013; Silveira et al., 2016). The prevailing land cover of 105
these bordering Domains are rain forest in the Mata Atlântica, semi-arid thorn 106
woodlands in the Caatinga and woody savannas in the Cerrado. Campos 107
rupestres are also found in ironstone formations of south-eastern and central-108
western Brazil, eastern Bolivia and the south-eastern Amazon Forest (known as 109
cangas in Brazil; Jacobi and Carmo, 2011; Silveira et al., 2016). Campo 110
rupestre landscapes also comprise patches of transitional vegetation (e.g., 111
parkland savanas, riverine forests), but here we adopt the sensu stricto 112
definition of campos rupestres, which comprises only the grassy-shrubby 113
savannas on quartzite, sandstone or ironstone rock outcrops (Alves et al., 114
2014). Many campo rupestre sites comprised in this contribution were not 115
included in previous studies (Fernandes et al., 2014; Silveira et al., 2016), 116
especially those found in quartzite and sandstone outcrops across the Goiás 117
state (central-western Brazil; Mews et al., 2014) and the ironstone-associated 118
campos rupestres found in the Mato Grosso do Sul state, near the Brazil-Bolivia 119
border (Neves and Damasceno-Junior, 2011). Mountaintop grasslands (campos 120
de altitude), which are found nearer to the Atlantic coast (Ribeiro et al., 2007), 121
were not included in this contribution because their flora is distinct and more 122
closely related to that of the páramos in the Andes (Safford, 2007). 123
2.2. Dataset 125
We extracted the dataset from the NeoTropTree (NTT) database 127
(, which consists of checklists of woody, 128
freestanding (i.e., lianas excluded) plant species, compiled for geo-referenced 129
sites, extending from southern Florida (U.S.A.) and Mexico to Patagonia. NTT 130
currently holds 5,126 sites/checklists, 14,878 woody plant species and 920,129 131
occurrence records. A site/checklist in NTT is defined by a single vegetation 132
type, following the classification system proposed by Oliveira-Filho (2015), 133
contained in a circular area with a 10-km diameter. Where two or more 134
vegetation types co-occur in one 10-km area, there may be two geographically 135
overlapping sites in the NTT database, each for a distinct vegetation type. 136
The data were originally compiled from an extensive survey of published 137
and unpublished literature (e.g., PhD theses), particularly those comprising 138
floristic surveys and forest inventories. Moreover, new species occurrence 139
records obtained from major herbaria and taxonomic monographs have been 140
added to the checklists when they come from within the 10-km diameter of the 141
original NTT site, and within the same vegetation type. All species and their 142
occurrence records were checked regarding current taxonomic and 143
geographical circumscriptions, as defined by the team of specialists responsible 144
for the online project Flora do Brasil (available at 145
The compilation of NTT avoided, therefore, the inclusion of occurrence records 146
with doubtful identification, location or vegetation type. It also excluded 147
checklists with very low species richness (< 20 species), because this is often 148
due to low sampling/collecting efforts, which results in poor descriptive power. 149
The dataset extracted from NTT consisted of 4,637 South American 150
woody plant community surveys, of which 115 were campos rupestres from 151
eastern and central western Brazil, south-eastern Brazilian Amazon and eastern 152
Bolivia. The full species matrix contained presence/absence data for 11,954 153
woody plant species, with a total of 883,258 presences, and the campos 154
rupestres species matrix contained presence/absence data for 1,055 woody 155
plant species, with a total of 12,801 presences. 156
The NTT database also includes 24 environmental variables for all sites, 157
obtained from multiple sources. Elevation (m above sea level) at the site centre, 158
obtained from WorldClim 1.4 (Hijmans et al., 2005), was used as an integrative 159
environmental variable. Climatic variables consisted of isothermality, maximum 160
temperature of warmest month, mean annual temperature, mean annual 161
precipitation, mean daily temperature range, minimum temperature of coldest 162
month, precipitation of driest month, precipitation of wettest month, precipitation 163
seasonality, temperature annual range and temperature seasonality, obtained 164
from WorldClim 1.4 data layers (Hijmans et al., 2005); cloud interception (mm) 165
and frost frequency (days), obtained from modelling known values as response 166
variables (data obtained from 135 and 57 Brazilian Meteorological Stations 167
measuring frost frequency and cloud interception, respectively), and elevation, 168
latitude and the aforementioned WorldClim layers as predicting variables; 169
duration (days) and severity (mm) of water deficit, produced by interpolating 5-170
day intervals of monthly temperatures and precipitation (WorldClim 1.4; Hijmans 171
et al., 2005) to be plotted in, and then extracted from, Walter’s Climate 172
Diagrams (Walter, 1985); and two additional variables, potential 173
evapotranspiration (mm) and an aridity index (annual precipitation/potential 174
evapotranspiration), derived by Zomer et al. (2007, 2008) from WorldClim data. 175
Six variables were substrated-related: grass coverage (%), obtained by 176
direct observation of the site surface via Google Earth© images in five 177
100×100m areas, one at the central coordinates of the NTT site and four at 2.5 178
km away from it and towards the NE, SW, NW and SE; soil coarseness (% 179
sand), soil fertility (% base saturation) soil salinity (ds/m) and surface rockiness 180
(% exposed rock), obtained from the Harmonized World Soil Database v 1.2 181
(available at and then ranked by mid-182
class percentage (raw figures were unrealistic due to local soil heterogeneity); 183
and soil drainage classes, obtained following EMBRAPA’s protocol (Santos et 184
al., 2013), which combines soil type, texture and depth with land forms. 185
2.3. Analyses of community composition 187
We used Simpson distance as the dissimilarity metric and unweighted 189
paired groups as the linkage method in a hierarchical clustering analysis 190
(McCune & Grace, 2002). We built 1000 clusters, with each cluster being built 191
after randomising the row order in the species composition matrix (species per 192
site), following the procedure proposed by Dapporto et al. (2013). The final 193
cluster is assembled following the majority consensus rule: if a given group is 194
represented in at least 50% of the trees built using a given set of samples, that 195
group is represented in the final consensus tree (Omland et al., 2008). This 196
analysis was conducted using the recluster package (Dapporto et al., 2015) in 197
the R Statistical Environment (R Development Core Team, 2016). 198
We assessed the overall patterns of floristic identity in campos rupestres 199
by (i) analyzing species occupancy (i.e., with species incidences rather than 200
abundances), and (ii) performing an ordination of campo rupestre woody plant 201
communities (115 sites) by non-metric multidimensional scaling (NMDS) of their 202
species composition (McCune and Grace, 2002) using Simpson distance as the 203
dissimilarity metric. Following methods similar to those proposed by Kreft and 204
Jetz (2010), the colours blue, green, yellow and red were assigned to the four 205
corners of the two-dimensional ordination plot in clockwise order from the origin. 206
NMDS scores were then mapped in geographic space by assigning a colour to 207
each site according to its position in the two-dimensional ordination space. 208
Beforehand, the ordination was rescaled to axes ranging from 0 to 1. Rescaling 209
is possible with NMDS results since ordination axes as such have no meaning 210
and only the relative position of points in ordination space matters. The NMDS 211
and the colour assignment were conducted in the statistical packages vegan 212
(Oksanen et al., 2016) and recluster (Dapporto et al., 2015), respectively, both 213
in the R Statistical Environment (R Development Core Team, 2016). 214
We tested whether variation in environmental conditions can predict 215
differentiation in campos rupestres community composition, and then visually 216
explored the results by (i) plotting the NMDS scores in ordination and 217
geographic space, and (ii) fitting the values of the most important environmental 218
variables by generalized linear models (GLM) and generalized additive models 219
(GAM), respectively. This routine follows methods similar to those proposed by 220
Blanchet et al. (2008) and Legendre et al. (2012), which comprise (i) the 221
exclusion of 261 singletons (species found at a single site), as they commonly 222
increase the noise in most analyses without contributing information (Lepš & 223
Šmilauer, 2003); (ii) the Hellinger transformation of the binary 224
presence/absence data (Legendre & Gallagher, 2001), which reduces the 225
weight of widespread species and their inherent effect in ordination analyses; 226
(iii) the independent compilation of significant spatial and environmental 227
variables through a forward selection method for redundancy analysis (RDA), 228
after first checking that the respective global models were significant (Blanchet 229
et al., 2008); (iv) an additional and progressive elimination of collinear variables 230
based on their variance inflation factor (VIF) and ecological relevance, until 231
maintaining only those with VIF < 10 (Quinn & Keough, 2002); and (v) variation 232
partitioning of the community composition matrix with respect to the significant 233
spatial and environmental variables. As spatial variables, we used principal 234
coordinates of neighbour matrices (PCNMs; Borcard et al. 2004), which 235
represent the spatial structure of the sampling units at multiple spatial scales 236
without considering any environmental variation (Borcard et al., 1992; Legendre 237
et al., 2002; Borcard et al., 2004). We tested the overall significance of the 238
environmental fraction (controlled for spatial autocorrelation) by applying 239
ANOVA permutation tests (999 permutations) for RDA (Peres-Neto et al., 240
2006). The variable selection, variation partitioning and GLM/GAM analyses 241
were conducted using the fields (Nychka et al., 2015), spacemakeR (Dray et al., 242
2010) and vegan (Oksanen et al., 2016) packages in the R Statistical 243
Environment. 244
Finally, we conducted an assessment of the conservation status of 245
campos rupestres by overlaying the distribution of our 115 sites on to the 246
coverage of protected areas across South America. We used conservation units 247
from the Cadastro Nacional de Unidades de Conservação (Ministério do Meio 248
Ambiente - Brazil, and World Database on Protected 249
Areas (IUCN & UNEP - WCMC, All maps were 250
designed using the package maptools (Lewin-Koh and Bivand, 2012) in the R 251
Statistical Environment. 252
3. Results 254
3.1. Floristic patterns 256
The hierarchical clustering showed that campos rupestres, including the 258
sites occuring in Amazonian ironstone formations, are more similar to cerrado 259
woody savannas than to any of the other South American phytogeographical 260
domains (Fig. 1). These campo rupestre sites share 95% of their woody plant 261
species with other habitats in our community dataset, being 74% with Cerrado 262
woody plant formations and 53% with other phytogeographical regions (i.e., 263
Amazon, Atlantic Forest and Caatinga). Campo rupestre sites do not form a 264
single cluster but are scattered within a broad cerrado cluster (Fig. 1). The 265
assessment of species incidences revealed that 80% of species have relatively 266
low occupancy across campos rupestres (dashed line in Fig. 2). The distribution 267
of campos rupestres in the ordination space yielded by NMDS suggests a 268
compositional segregation into four relatively distinct floristic units (Fig. 3), 269
namely northern (blue spectrum), south-eastern (brown-yellow spectrum), 270
north-eastearn (green spectrum) and central-western (red-purple spectrum) 271
campos rupestres. Bolivian campos rupestres are floristically related to the 272
central-western group (see similarity in Fig. 3b), and the foristic differentiation 273
between eastern groups is comparatively more subtle. 274
The conservation status of campo rupestre sites is uneven across these 275
four floristic groups (Fig. 4a). Many central-western campos rupestres are 276
unprotected, while northern campos rupestres are better protected (Fig. 4a). 277
The conservation assessment revealed that the current network of protected 278
areas does not cover the entire floristic space of campo rupestre woody plant 279
communities (Fig. 4b). 280
3.2. Environmental drivers of community turnover 282
The forward selection procedure retained six PCNMs for modelling 284
variation in campo rupestre community composition (adjusted R2 = 0.158, which 285
is fairly close to the value for all 38 PCNMs without any selection, adjusted R2 = 286
0.159). These selected spatial vectors are amongst the first PCNMs, which 287
represent broad-scale, positive spatial autocorrelation. Regarding 288
environmental variables, the forward selection retained 13 environmental 289
variables (adjusted R2 = 0.271, which is near the value for all 24 environmental 290
variables without any forward selection, adjusted R2 = 0.304) for modelling 291
variation in campo rupestre community composition (Table 1). Altitude, aridity 292
index, isothermality, maximum temperature of hottest month, mean daily 293
temperature range, minimum temperature of coldest month, potential 294
evapotranspiration, precipitation of wettest month, salinity, temperature annual 295
range, and water deficit duration and severity were the excluded environmental 296
variables. 297
When partitioning the variation explained by the retained environmental 298
and spatial predictors, we found that the environmental fraction explained 27% 299
of the variation, 15% of which was independent of spatial autocorrelation (P < 300
0.01). The environmental predictors could not account for 3% of the spatially 301
structured variation (P < 0.01), and 70% of the variation remained unexplained. 302
By fitting the values of the most important environmental variables in ordination 303
and geographic space (Fig. 3a and b, respectively), we observed a strong east 304
to west gradient related to decreasing surface rockiness (Fig. 3a-b), a proxy for 305
soil water deficit, thus segregating eastern campos rupestres from northern and 306
central-western campos rupestres. A south-east to north gradient was related to 307
increasing mean annual temperature (MAT) and decreasing temperature 308
seasonality (TempSeas), with northeastern and central-western campos 309
rupestres occurring in intermediate MAT and TempSeas (Fig. 3a-b). Mean 310
annual precipitation (MAP) was the third most important variable and was 311
associated with the floristic differentiation of north-eastern from northern 312
campos rupestres, with south-eastern and central-western occurring in 313
intermediate MAP (Fig. 3b). 314
4. Discussion 316
4.1. Floristic identity of campos rupestres 318
The first hypothesis was clearly supported by our results. Multiple campo 320
rupestre floristic groups may be recognized, based on distinct woody plant 321
species composition. Instead of representing a single floristic group across 322
South America, the campos rupestres form several separate groups within a 323
wider cerrado savannas group. This is the first attempt to show the degree to 324
which the geographically disjunct distribution of campo rupestre sites, and its 325
associated environmental heterogeneity, is underpinning the outstanding 326
floristic diversity in campos rupestres. Despite the fact that our dataset only 327
comprises woody plants, we predict that subsequent studies focusing on herbs 328
(a large component of campos rupestres floristic diversity) and/or animals will 329
reinforce this claim; i.e., influence from surrouding habitats is an important 330
factor shaping overall species composition in campos rupestres. Also, we 331
hypothesize that the high level of local endemism found in the non-woody 332
component of campos rupestres (Hensold, 1988; Mello-Silva, 1989; 333
Echternacht et al., 2011a) leads to even greater floristic heterogeneity amongst 334
campos rupestres; i.e., floristic dissimilarity amongst campos rupestres is even 335
higher if considering the non-woody component. 336
Our results show that individual campo rupestre groups share more 337
species with surrounding lowland cerrados than they do with other campo 338
rupestre groups. On the other hand, the campo rupestre sites in southeastern 339
and central-western Brazil represent a large and relatively cohesive floristic 340
group of campos rupestres (larger red cluster in Fig. 1), in agreement with a 341
considerable degree of floristic similarity between campos rupestres from the 342
Espinhaço range (southeastern Brazil) and the disjunct mountain ranges from 343
central-western Brazil (Feres et al., 2009). 344
4.2. Environmental drivers of community turnover 346
The second hypothesis was also supported by our results. Community 348
composition differentiation amongst campo rupestre floristic groups can be 349
predicted by variation in environmental conditions. Our results show that 350
northern campos rupestres occur in wet and warm environments with lower 351
surface rockiness (i.e., low soil water deficit). The northeastern group occurs in 352
the driest extreme of the precipitation space occupied by campos rupestres, 353
whereas southeastern and central-western campos rupestres are found in 354
intermediate, moist enviroments. The later two groups diverge over two other 355
important gradients: the southeastern group occurs in environments with lower 356
mean annual temperature, higher temperature and higher surface rockiness, 357
while the central-western group has lower rockiness and intermediate mean 358
annual temperature and temperature seasonality. 359
Variation in environmental conditions across the geographically disjunct 360
distribution of campos rupestres seems to be the main factor leading to floristic 361
divergence of campo rupestre woody plant communities. However, 362
understanding species-environment relationships in campo rupestre woody 363
plant communities is complex, as it partly depends on understanding the floristic 364
relationships between campos rupestres and their surrounding lowland 365
cerrados. On the one hand, the environmental gradients found across campos 366
rupestres have given rise to a pattern of low species occupancy such that many 367
campo rupestre woody plant species inhabiting this gradient can be said to 368
belong to only one of the four floristic groups. On the other hand, most of these 369
species also occur in other habitats of our woody plant community database, 370
suggesting that the environmental similarity between campos rupestres and the 371
surrounding lowland cerrados has allowed a regular exchange of woody plant 372
species between these two habitats. This is in agreement with a previous study 373
showing that campos rupestres and lowland cerrados in Goiás state, a portion 374
of our central-western group, differ in population structure of their woody plant 375
species but not in composition (Mews et al., 2014). From an ecological 376
perspective, campos rupestres and their surrounding lowland cerrados are likely 377
to form a continuous metacommunity with spatial variation in woody plant 378
population sizes being mainly driven by source-sink dynamics (Pulliam and 379
Danielson 1991); i.e., species that are better adapted to lowland cerrados 380
(source habitat for this species) are also found in campos rupestres (sink 381
habitat for this species), though in smaller populations, since species better 382
adapted to rocky substrate and shallower soils will prevail in population size. 383
4.3. Spatial structure 385
The campo rupestre floristic groups are largely geographic, thus 387
suggesting that there may be a role for spatially structured dispersal limitation 388
and historical biogeography in driving floristic differentiation. Nevertheless, our 389
results indicate that environmental conditions are better predictors of community 390
turnover (a proxy for niche-based dispersal limitation) than are geographical 391
factors (i.e., community composition/differentiation of unsampled campos 392
rupestres is better predicted based on environmental similarity than by 393
geographic proximity). This is supported by the negligible unique variation 394
attributed to positive spatial autocorrelation in campos rupestres, a proxy for a 395
distance decay in community similarity (Nekola and White, 1999), and by the 396
fact that it is more parsimonious to attribute the spatially structured 397
environmental variation to niche-based controls (cf. Legendre et al., 2009; 398
Neves et al., 2015). The niche-based dispersal limitation in campos rupestres is 399
further supported by two other results: (i) the comparatively high compositional 400
variation in southeastern Brazil is most likely to be associated with the role of 401
environmental heterogeneity in underpinning the occurrence of three floristic 402
groups, regardless of geographic proximity; and (ii) 95% of woody plant species 403
in campos rupestres are also found in other habitats, but have restricted 404
distributions across campos rupestres, likely because environmental conditions 405
are more similar between campos rupestres and surrounding lowland cerrados 406
than between geographically distant campo rupestre groups (spatially 407
structured environmental variation). 408
4.4. Conservation implications 410
Threats to campo rupestre biodiversity are many, and include mining, 412
unplanned urbanisation, high frequency of anthropogenic fire, uncontrolled 413
harvesting of ornamental plants, eucalyptus plantations, selective logging and 414
unplanned tourism (Giulietti et al., 1997; Jacobi et al., 2007, 2011; Fernandes et 415
al., 2014; Silveira et al., 2016). Considering the pervasive nature of most of 416
these threats, conservation strategies for campos rupestres need to be urgent 417
and well-informed scientifically. We believe our findings fit the ‘well-informing418
criteria and are therefore of relevance for conservation planning. Here we show 419
that campos rupestres are in fact segregated into three or four compositionally 420
distinct floristic units, which dictates that each group deserves separate 421
conservation planning. In doing so, future assessments may call attention to the 422
distribution of protected areas within each of these campo rupestre groups. 423
Recent studies have shown that at smaller geographic scales (e.g., Espinhaço 424
Range in eastern Brazil), several areas could be distinguished based on 425
taxonomic and evolutionary uniquiness of plants (Echternacht et al., 2011b; 426
Bitencourt and Rapini, 2013; Souza et al., 2013; Echternacht et al., 2014). 427
We also showed that campo rupestre floristic groups are unevenly 428
protected and that geographical gaps in the distribution of conservation units 429
result in a failure to protect important parts of the campo rupestre floristic space. 430
More specifically, campos rupestres found at intermediate values of the floristic 431
space summarized by the first NMDS axis are largely unprotected. These are 432
campos rupestres occurring under intermediate mean annual precipitation 433
(c.1,500 mm) in western Goiás state (central-western group) and southern 434
Minas Gerais state (southeastern group). We also call attention to the campos 435
rupestres found in Mato Grosso do Sul state, near the Bolivia border. These 436
campos rupestres, occurring in the ironstone formations of the Urucum plateau, 437
are largely unprotected, poorly studied and highly threatened by opencast 438
mining (Neves and Damasceno-Junior, 2011). 439
An alternative, and important, route forward in conservation planning lies 440
in addressing the evolutionary history of these campo rupestre groups. Previous 441
studies indicate that plant lineages from multiple biogeographical origins have 442
colonized campos rupestres many times over evolutionary history. For instance, 443
some bromeliad (Versieux et al., 2012) and orchid species (Gustafsson et al., 444
2010) found in campos rupestres seem to have an Atlantic rain forest origin 445
(i.e., sister taxa are mainly found in Atlantic rain forests), while some legume 446
species seem to have an origin in seasonally dry woodlands (Souza et al., 447
2013). In addition, others have stressed the idea that campos rupestres have 448
acted as ‘species pump’ for the surrounding lowland habitats (Simon et al., 449
2009; Silveira et al., 2016). Either way, future studies intending to quantify these 450
evolutionary shifts could shed light into the historical assembly of the campo 451
rupestre flora and, potentially, emphasize the necessity of conservation 452
strategies aiming to protect distinct campo rupestre groups along with 453
associated surrounding habitats. 454
5. Conclusion 456
We found an overall lack of compositional identity across the campos 458
rupestres woody flora, which is driven by their geographically disjunct 459
distribution with its associated environmental heteregoneity and floristic 460
influence from surrounding habitats. Therefore, we stress the necessity of 461
considering such floristic and environmental heterogeneity in conservation, 462
management and research planning and emphasize the need for multiple 463
protected areas across the separate floristic groups of campos rupestres. Our 464
findings also indicate that campos rupestres and their surrounding lowland 465
cerrados exchange woody plant species regularly and, therefore, merit 466
simultaneous conservation attention. Conservation units aiming to protect 467
campo rupestre biodiversity should not be limited to campo rupestre areas. 468
Rather, effective protected areas should function as ecological corridors 469
connecting multiple campos rupestres through lowland cerrados. We predict 470
that future studies will confirm that lowland cerrados are linking geographically 471
distant woody plant populations, thus improving ecological functionality of 472
campos rupestres; such as pollen flow between campo rupestre sites. 473
Acknowledgements - AOF and MLB were supported by the Conselho 475
Nacional de Desevolvimento Científico e Tecnológico - Brazil (CNPq) (grants 476
301644/88-8 and 151002/2014-2, respectively). D.M.N. and R.T.P. were 477
supported by the National Environmental Research Council (grant 478
NE/I028122/1). PLSM thanks the Coordenação de Aperfeiçoamento de Pessoal 479
de Nível Superior - Brazil (CAPES) for supporting a full PhD at the University of 480
Edinburgh under the Science Without Borders Programme (grant BEX 13197-481
13-4). We are indebted to Fernando A.O. Silveira, Fabio R. Scarano and one 482
anonymous referee for their valuable contributions to the manuscript. 483
6. References 485
Ab’Sáber, A.N., 2003. Os domínios de natureza no Brasil: potencialidades 487
paisagísticas. Ateliê Editorial, São Paulo. 488
Alves, R.J.V., Silva, N.G., Oliveira, J.A., Medeiros, D., 2014, Circumscribing 489
campo rupestre - megadiverse Brazilian rocky montane savannas. Braz. 490
J. Biol. 74, 355362. 491
Bitencourt, C., Rapini, A., 2013. Centres of endemism in the Espinhaço Range: 492
identifying cradles and museums of Asclepiadoideae (Apocynaceae). 493
Syst. Biodivers. 11, 525536. 494
Blanchet, F.G, Legendre, P, Borcard, D., 2008. Forward selection of 495
explanatory variables. Ecology 89, 2623-2632. 496
Borcard, D., Legendre, P., Drapeau, P., 1992. Partialling out the spatial 497
component of ecological variation. Ecology 73, 1045-1055. 498
Borcard, D., Legendre, P., Avois-Jacquet, C., Tuomisto, H., 2004. Dissecting 499
the spatial structure of ecological data at multiple scales. Ecology 85, 500
1826-1832. 501
Dapporto, L., Ramazzotti, M., Fattorini, S., Talavera, G., Vila, R., & Dennis, 502
R.L.H., 2013, Recluster: An unbiased clustering procedure for beta-diversity 503
turnover. Ecography 36, 10701075. 504
Dapporto, L., Ramazzotti, M., Fattorini, S., Vila, R., Talavera, G., Dennis, 505
R.H.L., 2015. recluster: Ordination methods for the analysis of beta-506
diversity indices. R package version 2.8. http://CRAN.R-507 508
Dray, S., 2010. spacemakeR: spatial modelling. R package version 473 0.0509
5/r101. 510
IBGE, 1993. Mapa de Vegetação do Brasil. Fundação Instituto Brasileiro de 511
Geografia e Estatística, Ministério da Agricultura, Rio de Janeiro. 512
Echternacht, L., Sano, P.T., Trovó, M., Dubuisson, J., 2011a, Phylogenetic 513
analysis of the Brazilian microendemic Paepalanthus subgenus Xeractis 514
(Eriocaulaceae) inferred from morphology. Bot. J. Linn. Soc. 167, 137515
152. 516
Echternacht, L., Trovó, M., Oliveira, C.T., Pirani, J.R., 2011b, Areas of 517
endemism in the Espinhaço Range in Minas Gerais, Brazil. Flora 206, 518
782791. 519
Echternacht, L., Sano, P.T., Bonillo, C., Cruaud, C., Couloux, A., Dubuisson, J-520
Y., 2014. Phylogeny and taxonomy of Syngonanthus and Comanthera 521
(Eriocaulaceae): evidence from expanded sampling. Taxon 63, 4763. 522
Feres, F., Zucchi, M.I., Souza, A.P., Amaral, M.C.E., Bittrich, V., 2009. 523
Phylogeographic studies of Brazilian ‘campo-rupestre’ species: 524
Wunderlichia mirabilis Riedel ex Baker (Asteraceae). Biotemas 22, 17525
26. 526
Fernandes, G.W., Barbosa, N.P.U., Negreiros, D., Paglia, A.P., 2014. 527
Challenges for the conservation of vanishing megadiverse rupestrian 528
grasslands. Nat. Conservacao 12, 162-165. 529
Frisby, S., Hind, D.J.N., 2014. Ichthyothere sasakiae, (Compositae: 530
Heliantheae: Milleriinae), a new species from the Amazonian campo 531
rupestre of northern Mato Grosso State, Brazil. Kew Bull. 69, 9504. 532
Giulietti, A.M., Pirani, J.R., Harley, R.M., 1997. Espinhaço Range region, 533
eastern Brazil, in: Davis, S.D., Heywood, V.H., Herrera-Macbryde, O., 534
Villa-Lobos, J., Hamilton, A.C. (Eds.), Centres of plant diversity: a guide 535
and strategy for their conservation. IUCN Publication Unit, Cambridge, 536
UK, pp. 397-404. 537
Gustafsson, A.L.S., Verola, C.F., Antonelli, A. (2010) Reassessing the temporal 538
evolution of orchids with new fossils and a Bayesian relaxed clock, with 539
implications for the diversification of the rare South American genus 540
Hoffmannseggella (Orchidaceae: Epidendroideae). BMC Evol. Biol. 10, 541
177. 542
Hensold, N., 1988, Morphology and systematics of Paepalanthus subgenus 543
Xeractis (Eriocaulaceae). Systematic Botany Monographs 23, The 544
American Society of Plant Taxonomists, Michigan. 545
Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very 546
high resolution interpolated climate surfaces for global land areas. Int. J. 547
Climatol. 25, 19651978. 548
Hopper, S.D., 2009. OCBIL theory: towards an integrated understanding of the 549
evolution, ecology and conservation of biodiversity on old, climatically 550
buffered, infertile landscapes. Plant Soil 322, 4986. 551
Hughes, C.E., Pennington, R.T., Antonelli, A., 2013. Neotropical plant evolution: 552
assembling the big picture. Bot. J. Linn. Soc. 171, 118. 553
Jacobi, C.M., Carmo, F.F., Vincent, R.C., Stehmann, J.R., 2007. Plant 554
communities on ironstone outcrops - a diverse and endangered Brazilian 555
ecosystem. Biodivers. Conserv. 16, 21852200. 556
Jacobi, C.M., Carmo, F.F., 2011. Life-forms, pollination and seed dispersal 557
syndromes in plant communities on ironstone outcrops, SE Brazil. Acta 558
Bot. Bras. 25, 395-412. 559
Jacobi, C.M., Carmo, F.F., Campos, I.C., 2011. Soaring extinction threats to 560
endemic plants in Brazilian metal-rich regions. AMBIO 40, 540543. 561
Kreft, H., Jetz, W., 2010. A framework for delineating biogeographical regions 562
based on species distributions. J. Biogeogr. 37, 20292053. 563
Legendre, P., Gallagher, E.D., 2001. Ecological meaningful transformations for 564
ordination of species data. Oecologia 129, 271-280. 565
Legendre, P., Dale, M.R.T., Fortin, M.J., Gurevitch, J., Hohn, M., Myers, D., 566
2002. The consequences of spatial structure for the design and analysis 567
of ecological field surveys. Ecography 25, 601-615. 568
Legendre, P., Mi, X., Ren, H., Ma, K., Yu, M., Sun, I., He, F., 2009. Partitioning 569
beta diversity in a subtropical broad-leaved forest of China. Ecology 90, 570
663674. 571
Legendre, P., Borcard, D., Roberts, D.W., 2012. Variation partitioning involving 572
orthogonal spatial eigenfunction submodels. Ecology 93, 12341240. 573
Lepš, J., Šmilauer, J.P., 2003. Multivariate analysis of ecological data using 574
CANOCO. Cambridge University Press, Cambridge, UK. 575
Lewin-Koh, N.J. & Bivand, R. (2012) maptools: tools for reading and handling 576
spatial objects. R package version 0.817/ r238. Available at: http://R-577
Forge.R 578
McCune, B., Grace, J.B., 2002, Analysis of ecological communities. MjM 579
Software Design, Gleneden Beach, Oregon. 580
Mello-Silva, R., 1989, Velloziaceae de Grão-Mogol, Minas Gerais, Brazil. MSc 581
Thesis. Universidade de São Paulo. 582
Mews, H.A., Pinto, J.R.R., Eisenlohr, P.V., Lenza, E., 2014. Does size matter? 583
Conservation implications of differing woody population sizes with 584
equivalent occurrence and diversity of species for threatened savanna 585
habitats. Biodivers. Conserv. 23, 1119-1131. 586
Nekola, J.C., White, P.S., 1999. The distance decay of similarity in 587
biogeography and ecology. J. Biogeogr. 26, 867-878. 588
Neves, D.M., Damasceno-Junior, G.A., 2011. Post-fire phenology in a campo 589
sujo vegetation in the Urucum plateau, Mato Grosso do Sul, Brazil. Braz. 590
J. Biol. 71, 881-888. 591
Neves, D.M., Dexter, K.G., Pennington, R.T., Bueno, M.L., Oliveira-Filho, A.T., 592
2015. Environmental and historical controls of floristic composition 593
across the South American Dry Diagonal. J. Biogeogr. 42, 1566-1576. 594
Nychka, D., Furrer, R., Paige, J., Sain, S., 2015. fields: Tools for Spatial Data. R 595
package version 8.3-5. 596
Oliveira-Filho, A.T., 2015. Um Sistema de classificação fisionômico-ecológica 597
da vegetação Neotropical, in: Eisenlohr, P.V., Felfili, J.M., Melo, 598
M.M.R.F., Andrade, L.A., Meira-Neto, J.A.A. (Eds.), Fitossociologia no 599
Brasil: métodos e estudos de casos, volume 2. Editora UFV, Viçosa, 600
Brazil, pp. 452-473. 601
Oksanen, J., Blanchet, F.G., Kindt, R., Legendre, P., Minchin, P.R., O’Hara, 602
R.B., Simpson, G.L., Solymos, P., Stevens, M.H.H., Wagner, H., 2016. 603
vegan: community ecology package. R package version 2.03. 604 605
Omland, K.E., Cook, L.G., Crisp, M.D., 2008. Tree thinking for all biology: the 606
problem with reading phylogenies as ladders of progress. BioEssays 30, 607
85467. 608
Peres-Neto, P.R., Legendre, P., Dray, S., Borcard, D., 2006. Variation 609
partitioning of species data matrices: estimation and comparisons of 610
fractions. Ecology 87, 2614-2625. 611
Pulliam, H.R., Danielson, B.J., 1991. Sources, Sinks, and Habitat Selection - a 612
landscape perspective on population-dynamics. American Naturalist 137, 613
50-66. 614
Quinn, G.P., Keough, M.J., 2002. Experimental design and data analysis for 615
biologists. Cambridge University Press, Cambridge, UK. 616
R Core Team, 2016. R: a language and environment for statistical computing. 617
Version 3.1.0. R Foundation for Statistical Computing, Vienna. 618 619
REFLORA, 2016. Lista de espécies da flora do Brasil. Jardim Botânico do Rio 620
de Janeiro, Brazil. (accessed 19.12.16). 621
Ribeiro, K.T., Medina, B.M.O., Scarano, F.R. 2007. Species composition and 622
biogeographic relations of the rock outcrop flora on the high plateau of 623
the Itatiaia, SE-Brazil. Revista Brasil. Bot. 30, 623-639. 624
Safford, H.D., 2007, Brazilian Páramos IV. Phytogeography of the campos de 625
altitude. J. Biogeogr. 34, 1701-1722. 626
Santos, H.G., Jacomine, P.K.T., Anjos, L.H.C., Oliveira, V.A., Lumbreras, J.F., 627
Coelho, M.R., Almeida, J.A., Cunha, T.J.F., Oliveira, J.B., 2013, Sistema 628
brasileiro de classificação de solos, 3rd edn. Embrapa, Brasília. 629
Saravia, E.S.R., 2008. El costo de la conservación de los bosques tropicales. 630
PhD Thesis - Universidad Autónoma G. René Moreno, Santa Cruz de la 631
Sierra. 632
Silveira, F.A.O., Negreiros, D., Barbosa, N.P.U., Buisson, E., Carmo, F.F., 633
Carstensen, D.W., Conceição, A.A., Cornelissen, T.G., Echternacht, L., 634
Fernandes, G.W., Garcia, Q.S., Guerra, T.J., Jacobi, C.M., Lemos-Filho, 635
J.P., Le Stradic, S., Morellato, L.P.C., Neves, F.S., Oliveira, R.S., 636
Schaefer, C.E., Viana, P.L., Lambers, H., 2015. Ecology and evolution of 637
plant diversity in the endangered campo rupestre: a neglected 638
conservation priority. Plant Soil 403, 129-152. 639
Souza, E.R., Lewis, G.P., Forest, F., Schnadelbach, A.S., van den Berg, C., 640
Queiroz, L.P., 2013. Phylogeny of Calliandra (Leguminosae: 641
Mimosoideae) based on nuclear and plastid molecular markers. Taxon 642
62, 12001219. 643
Versieux, L.M., Barbará, T., Wanderley, M.G.L., Calvente, A., Fay, M.F., Lexer, 644
C., 2012. Molecular phylogenetics of the Brazilian giant bromeliads 645
(Alcantarea, Bromeliaceae): implications for morphological evolution and 646
biogeography. Mol. Phylogenet. Evol. 64, 177-189. 647
Walter, H., 1985. Vegetation of the earth and ecological systems of the geo-648
biosphere, 3rd edn. Springer-Verlag, Berlin. 649
Zomer, R.J., Bossio, D.A., Trabucco, A., Yuanjie, L., Gupta, D.C., Singh, V.P., 650
2007. Trees and water: smallholder agroforestry on irrigated lands in 651
northern India. IWMI Research Report 122. International Water 652
Management Institute, Colombo, Sri Lanka. 653
Zomer, R.J., Trabucco, A., Bossio, D.A., van Straaten, O., Verchot, L.V., 2008. 654
Climate change mitigation: a spatial analysis of global land suitability for 655
clean development mechanism afforestation and reforestation. Agric. 656
Ecosyst. Environ. 126, 67-80. 657
Table and Figure Captions 659
Table 1 Variables selected for the study of environmental drivers of community 661
turnover across campos rupestres of South America. The variables shown are 662
ordered by the amount of explanation in species composition variation across 663
the campos rupestres. Goodness-of-fit of the predictor variables were assessed 664
through adjusted coefficients of determination, Akaike Information Criterion 665
(AIC), F-values and significance tests (p-value). VIF, variance inflation factor, 666
was obtained using the r-squared value of the regression of one variable 667
against all other explanatory variables. 668
Figure 1. Hierarchical clustering for 4,637 South American woody plant 670
communities based on their species composition. The dissimilarity measure 671
and linkage methods used were Simpson and unweighted group average, 672
respectively. The woody plant communities are discriminated by different 673
colours: black, 621 cerrado woody savannas; red, 115 campos rupestres; gray, 674
3,901 other South American woody plant communities. 675
Figure 2. Rank occupancy of campo rupestre woody plant species. Each 677
gray circle represents a campo rupestres species in our dataset. Darker shades 678
of gray indicate overlapping circles (i.e., two or more species have similar 679
occupancies). Circles below the dashed line occur in 20 or less campo 680
rupestres sites (80% of the 1,055 species). 681
Figure 3. Ordination of 115 sites of campo rupestre inferred from non-683
metric multidimensional scaling (NMDS) of their woody plant species 684
composition (a), and geographical variation of species composition and 685
mean annual precipitation (b). NMDS scores were plotted in the ordination 686
diagram after assigning a colour to each site according to its position in the two-687
dimensional ordination space (axes 1 x 2). Variation in surface rockiness, mean 688
annual temperature and temperature seasonality were fitted in ordination space 689
by generalized linear model. Colours of circles plotted across geographic space 690
are identical to the colours of circles in the NMDS scatter plot. Variation in mean 691
annual precipitation was fitted across geographic space by generalized additive 692
model. Dashed lines in (b) represent Brazilian state borders. AF = Atlantic 693
Forest Domain (white along the Atlantic coast); Am = Amazon (white in 694
northwestern South America); Ca = Caatinga (light gray); Ce = Cerrado Domain 695
(dark gray). 696
Figure 4. Conservation assessment of campo rupestre woody plant 698
communities. (a) Distribution of protected and unprotected campos rupestres 699
in South America. Grey areas represent the current network of protected areas 700
across South America. Dashed lines represent Brazilian state borders. (b) 701
Conservation status of the two-dimensional floristic space of campo rupestre 702
woody plant communities. Circles represent the position of campo rupestre sites 703
in ordination space inferred from non-metric multidimensional scaling (NMDS 704
axes 1 x 2) and are identical to the position of campos rupestres in Fig. 3a. 705
Variation in surface rockiness, mean annual temperature and temperature 706
seasonality were fitted in ordination space by generalized linear model. 707
Contours representing mean annual precipitation were fitted in ordination space 708
by generalized additive model. Dashed lines in (a) represent Brazilian state 709
borders. 710
Table 1.
cumulative adjusted R2
Mean annual temperature
Mean annual precipitation
Temperature seasonality
Precipitation seasonality
Grass coverage
Soil drainage
Soil fertility
Days of frost
Mean daily temperature range
Precipitation of driest month
Cloud interception
Figure 1
Figure 2
Figure 3
Figure 4
... A series of endemism studies (e.g., Rapini et al., 2002;Costa et al., 2008;Zappi and Taylor, 2008;Echternacht et al., 2011;Bitencourt and Rapini, 2013;Neves et al., 2017;Colli-Silva et al., 2019) has been conducted in the campo rupestre over the years, even though there is still a huge vegetation diversity to be scientifically investigated. Echternacht et al. (2011;p. ...
... Also, Kamino et al. (2008) presented floristic data that indicate areas of transition between cerrado and campo rupestre at the slopes of the ER in Minas Gerais state. Additionally, Neves et al. (2017) reinforced that the campo rupestre sites do not form a single cluster but are dispersed within a cerrado cluster, being more similar to cerrado woody savannas than to any other South American vegetational formation. All these floristic aspects characterize the vegetative mosaic that covers the ER, as well as reinforcing the need to investigate areas of endemism in the context of different vegetation types. ...
... Although well-known, this latitudinal pattern of endemism was not discussed in important endemism studies (Bitencourt and Rapini, 2013;Colli-Silva et al., 2019). Yet, Kamino et al. (2008) and Neves et al. (2017) commented on the latitudinal variation of species distribution and related it analytically to climatic and edaphic factors, whereas Echternacht et al. (2011) linked with the influence of the phytogeographic domains that cover the ER and the presence of the Para una River valley as a barrier to broad species distribution. ...
Using Lauraceae as a study case, we aimed in this article to: (i) delimit areas of endemism in the Espinhaço Range, Brazil; (ii) compare these areas of endemism with those previously delimited, as well as with the centres of endemism; (iii) evaluate the association between areas of endemism and vegetation types; and (iv) classify the areas of endemism according to the International Code of Area Nomenclature (ICAN). Based on a recent survey of 99 species from the Espinhaço Range, our dataset consisted of 34 endemic species belonging to nine genera. Following previous studies, we performed parsimony analysis of endemicity (PAE) using a grid square size of 0.5º × 0.5º. We delimited four areas of endemism of Lauraceae: (i) Antônio dos Santos, (ii) Conceição do Mato Dentro, (iii) Itambé do Mato Dentro, and (iv) Rio de Contas; and confirmed six previously delimited areas of endemism: (i) Southern MG, (ii) Southern Mountains Complex, (iii) Conceição do Mato Dentro, (iv) Diamantina Plateau, (v) Serra do Cabral, and (vi) Chapada Diamantina. The areas of endemism Conceição do Mato Dentro, Serra do Cabral, Diamantina, and Serra do Cipó were classified as subdistricts in the Diamantina Plateau district of the Southern Espinhaço province. We summarized and mapped all areas of endemism corresponding to the provinces, districts, and subdistricts that cover the Espinhaço Range. Areas of endemism and centres of endemism are contrasted. Finally, we highlight that the biogeographic studies along this mountain range should embrace higher taxa with representative species in different types of vegetation in order to enrich the majority of the endemism studies mainly concentrated on the campo rupestre. Unusual distribution patterns, diversity of vegetation types, and the presence of restricted species and monophyletic groups open up opportunities to carry out integrative studies concerning the biogeographic regionalization of the ER at multiple spatio-temporal scales.
... Rainer (Maas et al., 2011), were classified as ecoregion specialists, geographically restricted, with scarce population (Form 7), and have previously been classified as vulnerable (IUCN, 2019). Thus, our results agree with those of other authors who found that the South American network of protected areas has a conservation gap that fails to conserve a part of the biodiversity (Cardoso Da Silva and Bates, 2002;Forero-Medina and Joppa, 2010;Jenkins and Joppa, 2009;Neves et al., 2018;Ribeiro et al., 2018;Rodrigues et al., 2004). Similar to the threatened and endemic plant species (DRYFLOR et al., 2016;Neves et al., 2018;, we found that part of SAS rare species is also not protected in the present network of protected areas. ...
... Thus, our results agree with those of other authors who found that the South American network of protected areas has a conservation gap that fails to conserve a part of the biodiversity (Cardoso Da Silva and Bates, 2002;Forero-Medina and Joppa, 2010;Jenkins and Joppa, 2009;Neves et al., 2018;Ribeiro et al., 2018;Rodrigues et al., 2004). Similar to the threatened and endemic plant species (DRYFLOR et al., 2016;Neves et al., 2018;, we found that part of SAS rare species is also not protected in the present network of protected areas. However, the conservation gap concerning rare species ...
... Nevertheless, other ecoregions, such as Campos rupestres, Maranhão babassu forests and Dry Chaco also contain large numbers of rare species, but they are not part of the Global200. For example, Campos rupestres have a high floristic diversity that is driven by many factors (Neves et al., 2018) and ...
Full-text available
The interest in quantifying rare species has been increasing, but less attention has been paid to analysing their conservation status. Here, we used the Rabinowitz method based on geographical range, habitat specificity and population size to classify 2,203 tree species of South American savannas (SAS). We considered species with narrow (stenochoric) or wide (eurychoric) geographic range respectively occurring in up to 10% or over of the latitudinal belts in the SAS, specialist (stenoecious) or generalist (euryoecious) species occurring in one or more ecoregions, respectively, and locally scarce or abundant populations of species with low or high numbers of records within a grid cell of 1° latitude by 1° longitude, respectively. We then quantified species which only occur inside protected areas (PAs), in and outside PAs, or only outside PAs. Of the 2,203 species, 49.25% were rare, 40.35% stenochoric, 14.12% stenoecious and 56.15% had a scarce population. The Cerrado (433) followed by Caatinga (259), Campos Rupestres montane savanna (256), Bahia interior forests (189) and Guiana savanna (154) had the highest number of rare species. Only 6.44% species only occurred inside PAs, 83.79% in and outside PAs and 9.75% only outside PAs. Unprotected species are mostly geographically restricted with ecoregion specialists, and have small population sizes. The number of protected species are is positively correlated with the number of records within them. The establishment of PAs should focus on both core and disjunct savannas to encompass all ecoregions in order to ensure the conservation of species and the range of evolutionary processes in SAS.
... In this sense, the disjunct geographic distribution of the campo rupestre, associated with large latitudinal, altitudinal, and topographic variations, allows a large number of species to coexist. At the same time, biomes adjacent to the campo rupestre greatly influence the species composition and high species turnover (Giulietti et al., 1997;Silveira et al., 2016;Neves et al., 2018;Rapini et al., 2021). Similarly, at the local scale, microclimatic, topographic, and edaphic variations support the formation of a heterogeneous floristic mosaic Fernandes, 2016;Silveira et al., 2016). ...
Full-text available
The increase in rates of habitat loss requires an understanding of how biodiversity is distributed. Campo rupestre is an old, climatically buffered, and infertile landscape located in Brazil. Considered a biodiversity hotspot, the campo rupestre is mainly threatened by mining activity that requires a large operating area. Campo rupestre is known for its restricted distribution area and high abiotic heterogeneity, which modulates species coexistence and richness. To recognise the association between habitat type and plant communities, we propose to describe the floristic composition of herbaceous and shrub components in four habitats of the campo rupestre comprising quartzite and ferruginous substrate. We classified habitat types by the main surface soil features. In each habitat, we sampled ten 100-m2 plots to access information on the shrub and ten 1-m2 plots for the herbaceous component. Altogether we sampled 153 species, belonging to 38 families. The cluster analysis ordered by Sorensen metric indicates a clear distinction of species composition in the shrub component in the four habitats. However, the floristic composition of the herbaceous component was similar between the four habitats but showed a distinction when contrasting with the substrate type. Our results highlight the local taxonomic distinction between habitat types and substrates, indicating that the ecological distinction among substrate types of the campo rupestre cannot be overlooked in conservation and restoration actions.
... Although the largest area of campo rupestre is especially characteristic of the mountains of Espinhaço Range, a mountain chain that extends through Bahia and Minas Gerais state in Eastern Brazil (Harley 1995), this vegetation can be found in other areas, such in the Brazilian Central Plateau, especially in Goiás state, in the Chapada dos Veadeiros region and the Serra dos Pireneus (Rapini et al. 2008, Vasconcelos 2011. Campo rupestres has long been recognised as areas with noteworthy species richness and endemism (Harley 1988, Giulietti et al. 1997, Bitencourt & Rapini 2013, Neves et al. 2018). ...
Full-text available
Oocephalus foliosus was described in the first half of 19 th century, based on a collection from central Goiás state, Brazil, being collected again only three times in surrounding areas. Although this species seems to be rare and endemic to a narrow area, it has never been listed on any threatened list or had its conservation status assessed. Recently, we recorded a small population of O. foliosus in the Pireneus peak, an area of campo rupestre located in the municipality of Pirenópolis, Goiás, allowing us to improve the species description, assess its extinction risk and comment on its taxonomy. Also, a second step lectotypification was needed to the species and is here proposed.
... Also, there was an association between the campo rupestre and the cerrado, since both share floristic elements (Rapini et al., 2008). Indeed, Neves et al. (2018) highlighted this association in the survey evaluating the woody plants in the campo rupestre and showed that the campo rupestre did not form a single cluster, but they are scattered in a wide cluster of the Cerrado. In fact, the campo rupestre has been traditionally classified as a vegetation type of the Cerrado domain (Ribeiro and Walter, 1998;Ratter, 2002, 2008). ...
Our study is the most comprehensive checklist of the species of Lauraceae in the Espinhaço Range, including 99 species that belong to 13 genera. This number of species represents about 22% of the total estimated in Brazil and the endemism 8%. Thirty-four endemic species are identified, of which 15 are potential new taxa to be yet described, and 15 are categorized as micro-endemic. The Espinhaço Range presents four centers of richness and endemism of Lauraceae: (i) Serra do Cipó, (ii) South Espinhaço Range in Minas Gerais (Quadrilátero Ferrífero), (iii) Chapada Diamantina, and (iv) the Diamantina Plateau. There is a high sharing of species between South Espinhaço Range in Minas Gerais and Serra do Cipó, and South Espinhaço Range in Minas Gerais, Serra do Cipó, and the Diamantina Plateau. A geographical congruence between the centers of species richness and species endemism is found, and both Serra do Cipó and Chapada Diamantina present the highest levels of endemism. The predominant vegetation type associated with the centers of species richness and endemism is the seasonal semideciduous forest, except in the Diamantina Plateau region that is mostly associated with the cerrado. In the Espinhaço Range, the seasonal semideciduous forest is the richest vegetation type throughout the latitudinal variation of the range. The high number of species widely distributed in different vegetation types highlights the association among the cerrado, campo rupestre, and seasonal semideciduous forest.
... Thus, we tested that there is a positive or negative effect of environmental variability on plant communities, which probably can promote higher beta diversity. Indeed, many studies from elsewhere in Brazilian mountaintop rupestrian ecosystems support this ecological pattern (Porembski 2007;Londoño et al. 2014;Le Stradic et al. 2015;Nunes et al. 2015;Silveira et al. 2016;Campos et al. 2018;Neves et al. 2018;Cordeiro and Neri 2019). ...
The Guayana Highlands (GH) represents a major center of Neotropical plant diversity and endemism. Considering the lack of information about the ecology of plant communities in this region, we analyzed the fine-scale pattern of vegetation diversity. Thus, we tested the fine-scale effects of abiotic filters (i.e. soil physicochemical properties and depth) on species richness and community composition in complex rocky outcrop landscapes at the Roraima table mountain summit. This tepui is located on the triple boundary shared by Brazil, Venezuela and Guyana. A total of 60 plots were allocated in three specific previously identified geoenvironments: Peaty Rupestrian Grassland, Bonnetia-Shrubby Rupestrian Grassland and Sandy Rupestrian Grassland. Overall, 4318 individuals were sampled, distributed among 60 species, especially belonging to the families Asteraceae and Bromeliaceae. Of this total, 27 (45%) species are endemic to GH. There were significant differences among the geoenvironments regarding species richness, community composition and soil attributes. The tested models demonstrated that species richness was influenced by variations in potential acidity; however, community composition was explained mainly by soil texture and depth effects. Our study revealed abiotic filters exerted crucial fine-scale effects on plant community diversity on the Roraima table mountain.
... Martius, autor do primeiro mapa de divisão da flora brasileira foi um dos precursores desse tema, cujos domínios na atualidade constituem os biomas do Brasil. No contexto da RBSE, as porções mais elevadas onde encontram-se os Campos Rupestres, abrigam em torno de 15% da flora vascular do Brasil em menos de 1% do território nacional (Fernandes et al., 2018;Neves et al., 2018;Silveira et al., 2016). ...
Full-text available
Este trabalho apresenta a exposição (Re)visitando a Paisagem Reserva da Biosfera da Serra do Espinhaço (RBSE) com Martius e Spix (1818-2018), realizada em Diamantina entre 2018/2019, como uma prática de ensino em ambiente não formal. A exposição consistiu em banners com informações das paisagens descritas pelos naturalistas Spix e Martius pela RBSE e em práticas sensoriais imergiram os visitantes ao século XIX. Temas abordados como conservação da natureza, degradação ambiental, geodiversidade e biodiversidade regional estão previstos na Base Nacional Comum Curricular nas áreas de Geografia, Ciências, História (ensino fundamental) e Ciências Humanas (ensino médio). Cerca de 2.300 pessoas visitaram a exposição, sendo 450 estudantes da educação básica e 180 do ensino superior. Desenvolvemos práticas como pinturas, dobraduras em papel e o plantio de sementes de espécies nativas do Cerrado descritas na obra Flora Brasiliensis aos finais das visitas. Essas atividades lúdicas desenvolveram-se com entusiasmo por professores e estudantes visitantes.
In Brazil, the country with the highest plant species richness in the world, biodiverse savannas and grasslands – i.e., grassy ecosystems, which occupy 27% of the country – have historically been neglected in conservation and scientific treatments. Reasons for this neglect include misconceptions about the characteristics and dynamics of these ecosystems, as well as inconsistent or regionally restricted terminology that impeded a more adequate communication about Brazil's savannas and grasslands, both within the country and internationally. Toward improved communication and recognition of Brazil’s diversity of ecosystems, we present the key drivers that control the main types of grassy ecosystems across Brazil (including in regions of the country where forests dominate). In doing so, we synthesize the main features of each grassy ecosystem in terms of physiognomy and ecological dynamics (e.g., relationships with herbivores and fire). We propose a terminology both for major grassland regions and for regionally relevant vegetation physiognomies. We also discuss terms associated with human land management and restoration of grassy ecosystems. Finally, we suggest key research needs to advance our understanding of the ecology and conservation values of Brazil’s grassy ecosystems. We expect that a common and shared terminology and understanding, as proposed here, will stimulate more integrative research that will be fundamental to developing improved conservation and restoration strategies.
Full-text available
The Espinhaço Range has an expressive relevance of natural and cultural attributes. This work aimed to draw a parallel between the geological and ecological importance of Espinhaço Range, a weaving between 19th-century travel literature and the relevance of the region's geodiversity and biodiversity today. Bibliographical research-based study. Geodiversity comprises a variety of natural geological and geomorphological aspects of the potential of a region's geological heritage. Biodiversity encompasses species that develop in an ecosystem and can eventually subsidize planning on protecting conservation and maintenance of specific areas. A more holistic look at Espinhaço Range indicated natural resource richness present in citations from naturalist books and by the listing of geodiversity sites. In addition, the Biosphere Reserve confirms the importance of the region in terms of biodiversity.
Full-text available
Payments for environmental services (PES) consist of a way of promoting conservation by providing a financial reward to landowners. We propose an approach to determine the potential of a region for implementing PES based on the land potential for agriculture. PES is a powerful mechanism able to contribute for at least ten United Nations sustainable development goals. The study area consists of the headwaters of two important basins in Brazil: the Doce and the Jequitinhonha rivers. The potential for agricultural use was determined based on lithology, soil class, and slope of the terrain layers. The land use was classified using Sentinel images between August 2019 and March 2020. The potential of environmental services were mapped overlapping the land use, the protected areas, and the potential for agricultural. We found that 42.9% of the land have low potential; 31.6% have moderate potential, 23.6% have high potential and 1.9% have very high agricultural. Native vegetation and rocky outcrops accounted together for 75% of the area. Pasture occupies another 22.3%, urban area 1.38%, mining 0.83% and agriculture 0.04%. We found that 87.3% of the land classified as low agricultural potential are still covered by native vegetation or rocky outcrops, and are natural candidates to enter a PES program. Livestock farming or agriculture developed in low potential areas are candidates for land retirement and restoration. The livestock farming is the dominant economic activity in the region, with annualized present value of US$ 106.77 per hectare, and should be the reference for the annual payment for natural areas included in a PES program. The annual budget for the PES program implementation aiming at the preservation of the current natural vegetation is US$ 10.8 millions.
Full-text available
Na presente edição, o Sistema Brasileiro de Classificação de Solos (SiBCS) mantém a mesma estrutura geral, incorpora mudanças, redefinições e correções, está liberado para o uso e pode ser citado e correlacionado com outros sistemas. Esta obra será aperfeiçoada ao longo de anos futuros, conforme determinado pelo uso efetivo em levantamentos de solos, estudos de correlação de solos e pesquisas na área de Ciência do Solo. As alterações aqui apresentadas foram disponibilizadas para testes e validação pelos usuários (Santos et al., 2016, 2017) e parcialmente apresentadas no Congresso Brasileiro de Ciência do Solo de 2017. Abrangem desde definições e conceitos básicos até reestruturações de classes em todos os níveis categóricos. Tais mudanças são reflexos das sugestões e críticas recebidas de usuários do SiBCS e, sobretudo, das ideias e propostas emanadas das últimas sete reuniões de classificação e correlação de solos (RCCs) realizadas nas regiões Sul, Sudeste e Norte do País (Reunião..., 2000, 2005, 2008, 2010a, 2010b, 2012, 2013, 2015, 2017). As RCCs tradicionalmente têm permitido a validação e o aperfeiçoamento do SiBCS, bem como a uniformização de critérios, o intercâmbio interinstitucional e a transferência de informações entre profissionais da Ciência do Solo. Dentre os aperfeiçoamentos, destacam-se ajustes, correções e redefinições de conceitos básicos relativos a definição de solo; caracteres argilúvico, alumínico e crômico; contato lítico; contato lítico fragmentário; constituição esquelética do solo; horizonte hístico e horizonte A antrópico. Alterações de redação, redefinição da seção de controle, de eliminação ou incorporação de classes de solos são propostas nos níveis categóricos de ordem (Gleissolos, Nitossolos, Organossolos, Vertissolos); de subordem (Argissolos Vermelhos, Cambissolos Hísticos, Chernossolos Argilúvicos, Neossolos Litólicos, Neossolos Regolíticos, Planossolos Nátricos); de grande grupo (inclusão dos Psamíticos nos Neossolos Regolíticos, redefinição dos Psamíticos nos Neossolos Flúvicos, exclusão dos Distroúmbricos e Eutroúmbricos nos Neossolos Litólicos e Neossolos Regolíticos, exclusão dos Alíticos em várias subordens, criação dos Hísticos e exclusão dos Húmicos nos Gleissolos Tiomórficos); e de subgrupo (inclusões de inúmeras classes de solos, exclusão dos úmbricos, redefinição dos chernossólicos, criação dos espesso-húmicos – Latossolos e Neossolos Quartzarênicos – e dos leptofragmentários – Argissolos, Cambissolos, Chernossolos, Gleissolos, Neossolos Regolíticos, Neossolos Quartzarênicos, Nitossolos, Organossolos e Plintossolos –, e substituição do termo êndicos por mésicos no 4º nível categórico dos Planossolos); bem como no nível categórico de família (como a criação dos ândicos) e série. São também apresentadas as classes de profundidade dos solos, classes de reação dos solos e uma proposta de designação dos tipos de terreno. Os subgrupos existentes e já definidos no SiBCS podem ser utilizados em outros grandes grupos, em que não constem suas ocorrências, devendo ser enviada uma justificativa e cópia do perfil para avaliação e validação da nova classe. A proposição de novas classes em qualquer nível categórico deve ser enviada ao Comitê-Executivo de classificação de solos (CE), contendo uma justificativa para a sua inserção e uma cópia do perfil correspondente para avaliação e validação a fim de que essa nova classe possa ser incorporada oficialmente ao sistema. Ao classificar um determinado perfil de solo, é permitido ao classificador fazer combinações de qualificativos para o 4º nível, desde que já definidos no SiBCS paraqualquer grande grupo de solo. Admite-se a utilização de no máximo três qualificativos de 4º nível categórico, por exemplo, Argissolo Vermelho Eutrófico solódico abrúptico plintossólico (ver Capítulo 5, Argissolo...). Esta edição substitui a classificação de solos que vinha sendo utilizada na Embrapa Solos (Camargo et al., 1987; Sistema..., 1999; Santos et al., 2006, 2013, 2014) e todas as aproximações anteriores (Sistema..., 1980, 1981; Camargo et al., 1988a; Carvalho et al., 1997). Objetivando que o SiBCS seja continuamente aprimorado, em decorrência da evolução científica e aumento do conhecimento dos solos brasileiros, solicita-se aos usuários o envio periódico de críticas e sugestões, que deverão ser encaminhadas ao CE para o endereço eletrônico <>, sendo que as atualizações realizadas poderão ser acessadas permanentemente na página do SiBCS na internet2. Cabe também esclarecer que o SiBCS, que vem sendo paulatinamente construído pela comunidade científica brasileira há décadas, hoje se encontra estruturado na forma de chave taxonômica até o 4º nível categórico, com recomendações de características/propriedades a serem empregadas na classificação de solos no 5º nível categórico (família). Ainda assim, na forma em que se encontra, o SiBCS já atende a praticamente todas as demandas atualmente conhecidas no Brasil acerca dos solos. As sugestões de características/propriedades para o 6º nível categórico são de caráter preliminar. A sua implementação demandará um volume de informação específico e suporte organizacional no País para a sua validação. Notas (2) Disponível em: <>
Full-text available
Studies on the herbaceous and sub-shrub layer of cerrado showed the occurrence of modifications in its composition between different regions, demonstrating sensibility to changes in climate, soil and intensity of fires, among other factors. The aim of this study was to describe the phenological variation in a campo sujo vegetation in the Urucum plateau. We established eight transects of 250 m each, 50 m apart. We sampled all flowering and fruiting species located at least three meters from each side of the transect. Beginning in October 2007, one month after an accidental fire occurred in the study site, we analysed flowering and fruiting plants in the transects' area. The intensity of the flowering and fruiting phenophases was not uniformly distributed. This study provide us information about the possible fire influence on the reproductive patterns of the community, presenting flowering peaks in October and November, two months after this event. Regression analysis with monthly rainfall also provides us information about the influence of climate data on the flowering and fruiting peaks.
Full-text available
Background Botanists, ecologists and evolutionary biologists are familiar with the astonishing species richness and endemism of the fynbos of the Cape Floristic Region and the ancient and unique flora of the kwongkan of south-western Australia. These regions represent old, climatically-buffered infertile landscapes (OCBILs) that are the basis of a general hypothesis to explain their richness and endemism. However, few ecologists are familiar with the campo rupestre of central and eastern Brazil, an extremely old mountaintop ecosystem that is both a museum of ancient lineages and a cradle of continuing diversification of endemic lineages. Scope Diversification of some lineages of campo rupestre pre-dates diversification of lowland cerrado, suggesting it may be the most ancient open vegetation in eastern South America. This vegetation comprises more than 5,000 plant species, nearly 15% of Brazil's plant diversity, in an area corresponding to 0.78% of its surface. Reviewing empirical data, we scrutinise five predictions of the OCBIL theory, and show that campo rupestre is fully comparable to and remarkably convergent with both fynbos and kwongkan, and fulfills the criteria for a classic OCBIL. Conclusions The increasing threats to campo rupestre are compromising ecosystem services and we argue for the implementation of more effective conservation and restoration strategies.
1. Introduction 2. Estimation 3. Hypothesis testing 4. Graphical exploration of data 5. Correlation and regression 6. Multiple regression and correlation 7. Design and power analysis 8. Comparing groups or treatments - analysis of variance 9. Multifactor analysis of variance 10. Randomized blocks and simple repeated measures: unreplicated two-factor designs 11. Split plot and repeated measures designs: partly nested anovas 12. Analysis of covariance 13. Generalized linear models and logistic regression 14. Analyzing frequencies 15. Introduction to multivariate analyses 16. Multivariate analysis of variance and discriminant analysis 17. Principal components and correspondence analysis 18. Multidimensional scaling and cluster analysis 19. Presentation of results.