Biodivers Conserv (2018) 27:2587–2603 https://doi.org/10.1007/s10531-018-1556-4 1 3 ORIGINAL PAPER The deadly route to collapse and the uncertain fate of Brazilian rupestrian grasslands G. Wilson Fernandes1,2 · N. P. U. Barbosa1 · B. Alberton3 · A. Barbieri4 · R. Dirzo2 · F. Goulart5,6 · T. J. Guerra7 · L. P. C. Morellato3 · R. R. C. Solar1 Received: 18 August 2017 / Revised: 28 March 2018 / Accepted: 2 May 2018 / Published online: 8 May 2018 © Springer Science+Business Media B.V., part of Springer Nature 2018 Abstract Rupestrian grasslands are biodiverse, evolutionary old vegetation complexes that harbor more than 5000 species of vascular plants and one of the highest levels of plant endemism in the world. Growing on nutrient–impoverished soils and under harsh environ- mental conditions, these mountaintop ecosystems were once spared from major human interventions of agriculture and intensive cattle ranching. However, in Brazil, rupestrian grasslands have experienced one of the most extreme land use changes among all Brazilian ecosystems, suffering from ill policies leading to intense mining activities, uncontrolled tourism, and unplanned road construction. Indeed, the discovery of large mineral reserves, the adoption of ineffective conservation policies, and, going forward, climate change, are Communicated by David Hawksworth. Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s1053 1-018-1556-4) contains supplementary material, which is available to authorized users. * G. Wilson Fernandes gw.fernandes@gmail.com 1 Departamento de Biologia Geral, Universidade Federal de Minas Gerais, Belo Horizonte, MG 30161-970, Brazil 2 Department of Biology, Stanford University, Stanford, CA 94305, USA 3 Laboratório de Fenologia, Departamento de Botânica, Instituto de Biociências, Universidade Estadual Paulista UNESP, Rio Claro, SP 13506-900, Brazil 4 Departamento de Demografia, Universidade Federal de Minas Gerais, Belo Horizonte, MG 30161-970, Brazil 5 Departamento de Zoologia, Instituto de Ciências Biológicas, Universidade Federal de Minas Gerais, CP 486, Belo Horizonte, MG 30161-970, Brazil 6 Instituto de Geociências, Pós-Graduação em Análise e Modelagem de Sistemas Ambientais, Universidade Federal de Minas Gerais, Belo Horizonte, MG 31270-900, Brazil 7 Departamento de Botânica, Instituto de Ciências Biológicas, Universidade Federal de Minas Gerais, Belo Horizonte, MG 30161-970, Brazil http://crossmark.crossref.org/dialog/?doi=10.1007/s10531-018-1556-4&domain=pdf https://doi.org/10.1007/s10531-018-1556-4 https://doi.org/10.1007/s10531-018-1556-4 2588 Biodivers Conserv (2018) 27:2587–2603 1 3 threatening this hyper-diverse ecosystem. Here, we shed light on the severe threats imposed by land-use changes in this ecosystem, modeling its future distribution under different sce- narios. We uncover a catastrophic forecast that, if not halted, will lead to the loss of 82% of this unique ecosystem in the future, impacting ecosystem services at regional scales, including water and food security potentially affecting more than 50 million persons. Keywords Biodiversity · Campo rupestre · Cerrado · Espinhaço mountains · Mining · Sustainability Introduction Conservation efforts in Brazil should be a matter of discussion across all its biomes, but it is undeniable that these efforts have tended to focus on forested ecosystems (Santos et al. 2011; Overbeck et  al. 2015; Fernandes 2016a). Non-forest areas, such as the rupestrian grasslands (“campo rupestre”), remain underappreciated and, consequently, under severe anthropogenic threat (Fernandes et al. 2014; Fernandes 2016b). These areas harbor a diver- sity of organisms that rivals forests in terms of number of species and exceeds them in the proportion of endangered flora. Distributed mainly within the Cerrado domain (savanna) in the upper parts of the Espinhaço mountain range (the second largest mountain chain in South America after the Andes), this highly heterogeneous herbaceous/shrubby vegeta- tion is a mosaic of rocky outcrops embedded in a matrix of sandy and stony grasslands, seasonal wet grasslands, springs, and occasional forest patches on mountaintops (Fig. 1) (for a review see Fernandes 2016a). Rupestrian grasslands evolved in ancient landscapes (Barbosa and Fernandes 2016) shaped by quartzitic and ferruginous rocks (Schaefer et al. 2016). These geologically old landscapes and evolutionary ancient vegetation complexes cover an area of ca. 83,000 km2 and harbor great biodiversity, which includes more than 5000 species of vascular plants and one of the highest levels of endemism in the world (Alves et al. 2014; Silveira et al. 2016). Brazilian rupestrian grasslands’ unique biota has long-attracted the attention of several scientists. Early in the XIX century, many prominent European naturalists, including Mar- tius, Spix, Langsdorff, Saint-Hilaire, Lund, and Warming, explored the region covered by this ecosystem (Warming 1892; Vasconcelos et al. 2008; Fernandes 2016a; Mügge et al. 2016; Lüttge 2017). The work by Eugen Warming in the transitional area between cerrado and the rupestrian grasslands—a seasonally dry forest transition region (1863–1866) at the base of the Espinhaço mountain range—provided the scenario whereby the first books on plant ecology (Warming 1892, 1895) were forged. Biodiversity in this ecosystem began to be catalogued more intensively by biologists in the 1960s and, although scattered in space and time, their assessments compiled a large number of species for the fauna and flora of the Espinhaço mountain range, with hundreds of new species being described over the past 50 years. The most recent species compilations for the Espinhaço range highlight this region as an important center of biological diversity and plant endemism (Giulietti et  al. 1997; Echternacht et  al. 2011; Fernandes 2016a; Silveira et  al. 2016), fish (Alves et al. 2008), frogs (Leite et al. 2008), birds (Vasconcelos et al. 2008; Chaves et al. 2015), galling insects (Fernandes and Santos 2014), and arbuscular mycorrhizal fungi (Carvalho et al. 2012; Coutinho et al. 2015), among many other groups. Even after decades of study, the rate of new species discovery in Brazilian rupestrian grasslands is still very high (Fig. 2). For instance, an average of 12 new plant and 4 new 2589Biodivers Conserv (2018) 27:2587–2603 1 3 animal species were described per year in the Espinhaço mountain range from 2005 to 2014. At least 118 new plant species in 27 families, and 26 new vertebrate species, com- prising 11 frogs, 8 lizards, 4 birds, 2 snakes and one mammal, were described within a single decade (Tables S1, S2). Two recently described species, the rodent Calassomys api- calis (Pardiñas et al. 2014) and the ovenbird Cinclodes espinhacensis (Freitas et al. 2012) (Fig. S1a), illustrate the singularity of most of the new species described. Also, among the 11 new arthropods described, the iridescent blue spider, Pterinopelma sazimai (Fig. S1b), and the flesh-fly mimicking weevil, Timorus sarcophagoides (Fig. S1c), are particularly remarkable. All these species are often rare, associated to mountaintop grasslands, patchily distributed and have become known to science under some degree of threat. Recent stud- ies also indicate that rupestrian grasslands are home to ca. 25% of the world’s described species of mycorrhizae (Carvalho et al. 2012) and that 19 new species of these fungi are in line to be described to science (Coutinho et al. 2015). Such high levels of diversity and endemism in rupestrian grasslands have been argued to be the result of strong environmental filters, such as nutrient–deprived soils, pronounced seasonality, and climatic variability related to wide altitudinal and latitudinal gradients (Fernandes et al. 2014; Negreiros et al. 2014; Fernandes 2016c). Indeed, not only new spe- cies, but also novel interactions are being found in these ecosystems, since harsh conditions and poor soils have driven striking adaptations of the endemic organisms. For example, the carnivorous plant Philcoxia minensis exhibits a unique prey-capture strategy among plants, Fig. 1 Aerial view of rupestrian grasslands on quartzitic soils in Serra do Cipó, southeastern Brazil. The landscape is composed of a mosaic of habitats, mostly sandy grasslands, rough terrain with stony grass- lands, rocky outcrops, and forests on humid slopes and valleys (a). Rocky outcrops are perhaps the most prominent habitat with large rocks interspersed with woody vegetation. Their erosion gives rise to the adja- cent stony or sandy grasslands, dominated by different species (b). Detailed view of the speciose rocky grassland habitats, where highly endemic Velloziaceae, Melastomataceae, Eriocaulaceae, forbs and grasses thrive (c). A patch of the endemic Paepalanthus robustus (Eriocaulaceae) (d) 2590 Biodivers Conserv (2018) 27:2587–2603 1 3 being able to trap and digest nematodes in its underground adhesive leaves (Figs. S1d, S1e) (Pereira et al. 2012; see also Oliveira et al. 2016). Additionally, the protocarnivorous plant Paepalanthus bromelioides (Fig. S1f) evolved the capacity of its leaves to derive nitro- gen from both spiders’ prey carcasses and feces, while roots uptake nitrogen from associ- ated termite mounds (Nishi et  al. 2013). Clearly, mechanisms and processes responsible for high species diversity in the Espinhaço range are numerous and vary across taxa (see Morellato and Silveira 2018  and references therein). The synergy among environmental filters, geographic barriers related to altitudinal variation and species interactions represent the most important forces leading to the vast array of adaptations and, likely, speciation in its fauna and flora (Fernandes 2016c). Furthermore, the high endemism and species rich- ness intrinsic to rupestrian grasslands could have been shaped by interglacial microrefugia, which might have maintained buffered microclimates derived from topographic idiosyncra- sies, which prevented extinctions (Barbosa et al. 2015). Therefore, while representing less than 1% of the Brazilian territory, rupestrian grass- lands shelter about 17% of the country’s estimated plant biodiversity and ca. 46% of that of the Cerrado, the second largest biome in South America (Fernandes et al. 2016a). How- ever, the unfortunate reality is that, despite its exuberant biodiversity, the rupestrian grass- lands are experiencing one of the most intense land use alterations among all Brazilian ecosystems (Fernandes et al. 2014; Sonter et al. 2014a; Fernandes 2016b; Pena et al. 2017). The loss of species and habitats, particularly on mountaintops, is of major relevance due to the potentially dire cascading effects that would result in the loss of species interactions, ecosystem connectivity and environmental services (Epps et al. 2006; Spehn et al. 2010). Given the continental relevance of this mountain region and the lack of appropri- ate protection of its natural heritage, our aim in this paper is to cast some light  on and draw attention to the major threats imposed by land use and climate change to rupestrian Fig. 2 Rate of new species discovery in Brazilian rupestrian grasslands between 2005 and 2014. See text and Table S1 for details 2591Biodivers Conserv (2018) 27:2587–2603 1 3 grasslands. By analyzing the prevailing threats to this ecosystem and then modeling its temporal dynamics, we aim to examine its projected distribution and to identify emergent patterns of potential future expansion or contraction and to determine climatically stable areas important for biodiversity conservation and ecosystem services (i.e., where condi- tions will remain suitable for the persistence of rupestrian grasslands). Furthermore, we analyze changes in landscape metrics to examine possible effects of changes in the habitat and on rupestrian grasslands’ biodiversity. Materials and methods Identification of priority regions To delineate priority regions for this study and conservation of rupestrian grasslands we used spatially-explicit data points of the location of rupestrian grasslands deemed as con- servation priority by experts who participated in the 1st Workshop on Biodiversity, Con- servation, Use and Public Policy in Rupestrian Grasslands, in Belo Horizonte (Brazil), in October 2011, focusing on the core area of the Espinhaço mountain range and the Canastra mountains (GWF, unpublished data). A total of 300 occurrence points for rupestrian grass- lands were generated covering the entire “core area” of the Espinhaço mountain range and the Canastra mountains. These two regions were chosen because of their strong represen- tation of this ecosystem (see Fernandes 2016a). Additional coordinates were kindly pro- vided by M Callisto, AA Conceição, and L Echternacht for localities in the northern part of the Espinhaço mountain range, the Canastra mountains and other isolated mountains in Goiás, Mato Grosso and Roraima states. In total, we registered 490 points of occurrence of rupestrian grasslands in Brazil, which resulted from an extensive discussion during the abovementioned workshop between experienced researchers about the distribution of the ecosystem. Environmental data Bioclimatic variables are commonly used predictors of the impacts of climate change on biodiversity (Schrag et  al. 2008). We assumed that such variables would satisfactorily determine the distribution of rupestrian grasslands in a historical context, given the strong associations of climate with the biotic environment (Carnaval and Moritz 2008; Garcia et  al. 2016). Bioclimatic and elevation variables, used in conjunction, were downloaded from the WorldClim platform at a spatial resolution of 0.0083° (~ 1  km2) (see Hijmans et al. 2005 for further details) for the period of 1950–1990. Aspect and slope data, derived from elevation, were generated using the Spatial Analyst Toolbox within the software ArcMap® (ESRI, California). Finally, we created a binary variable corresponding to areas where the predominant lithology is associated with rupestrian grasslands, such as quartz- ite formations, sandstone, silts, phyllites, meta-conglomerates, and iron ores (Schaefer et  al. 2016; see also Le Stradic et  al. 2014; Barbosa and Fernandes 2016). We assumed that this edaphic factor will remain largely unchanged under future climate scenarios. The shapefiles relating to the geological classifications for Brazil (scale 1:10,00,000) were downloaded from the Brazilian Geological Survey GEOSGB (http://geosg b.cprm.gov.br/) and processed through the software ArcMap® (ESRI, California) at a spatial resolution of 0.0083° (~ 1 km2). http://geosgb.cprm.gov.br/ 2592 Biodivers Conserv (2018) 27:2587–2603 1 3 Bioclimatic variables for year 2050 (averaged from 2041 to 2060) and year 2070 (aver- aged from 2061 to 2080) were also downloaded from WorldClim and followed the same spatial resolution of 0.0083° (Hijmans et al. 2005). We used data from five different cli- mate models (Table  1), to reduce the uncertainty of predictions resulting from a single global circulation model (GCM), in two concentration pathways: RCP 4.5 and RCP 8.5 (e.g., Khanum et al. 2013). Of these scenarios, the most pessimistic is the RCP 8.5, which assumes that global annual greenhouse gases continue to rise throughout the 21st cen- tury, and the more optimistic is RPC 4.5, which assumes that global annual greenhouse gas emissions peak around 2040 with emissions declining substantially thereafter (IPCC 2014). Other scenarios (e.g., RCP 2.6 and the RCP 6) were tested but not chosen, due to unrealistic predictions (a decrease in emissions at the end of this decade), and similar results of RCP 8.5, respectively. All models were used in combination, and the main result was the mean output. All variables were clipped to cover the Brazilian territory, encompassing the entire known distribution of rupestrian grasslands, including small relict areas. We built a Pear- son correlation matrix to detect the presence of multicollinearity between variables, and among those highly correlated variables (i.e., those with r ≥ 0.8), we opted to keep in the model those variables with higher biological relevance. Following this protocol, out of 19 bioclimatic variables initially selected, only eight were retained in the final model, in con- junction with the variables “elevation”, “aspect”, “slope” and “lithology” (Table S3). Predictive modeling Distribution models were built with the maximum entropy approach (MaxEnt) (Phillips et al. 2006; Elith et al. 2011). This algorithm can perform effective analysis even with small samples using only presence data (Hernandez et al. 2006). The algorithm was implemented by the software MaxEnt, version 3.3.3 (Computer Sciences Department—Princeton Uni- versity 2004), which was used for the generation of a logistic model for the distribution of rupestrian grasslands under baseline climate conditions for the period of 1950–1990 (e.g., Phillips et al. 2006). This model was then projected to the future scenarios. We used tenfold cross-validation for testing model performance (10 replicates) and averaged the results. We also used the jackknife procedure and permutation importance to estimate the relative influence of different predictor variables. The area under the ROC (receiver operating characteristic) curve (AUC) was used to evaluate model performance (Swets 1988). In addition, a threshold was selected which enabled us to obtain infor- mation about future suitable areas, which represent potential priority areas for the con- servation of rupestrian grasslands. The chosen threshold was the “10 percentile train- ing presence logistic threshold”, which shows the most realistic result and converts the Table 1 Climate models used in the predictions of the rupestrian grassland potential distribution models Model Name BCC-CSM1.1 Beijing climate center-climate system model 1.1 CCSM4 Community climate system model 4.0 GISS-E2-R NASA Goddard Institute for Space Studies modelE2 HadGEM2AO Hadley centre global environmental model version 2 (Atmosphere–Ocean) MIROC5 Model for interdisciplinary research on climate version 5 2593Biodivers Conserv (2018) 27:2587–2603 1 3 probability distribution derived from the model into a binary map of presence/absence (Liu et al. 2005). To avoid erroneous predictions of suitable habitat under future climate sce- narios for 2050 we used the “fade-by-clamping” option in MaxEnt, which removes heavily- clamped pixels from the final predictions (e.g., Khanum et al. 2013). Landscape metrics and land‑use threats/protection We conducted three analyses on how changes in environmental suitability and land-use will affect the biodiversity of rupestrian grasslands in current and future scenarios, given an increasingly complex set of assumptions. Firstly, we assessed how changes in environmen- tal suitability will affect landscape metrics, if all suitable areas are covered by native habi- tat (overall landscape metrics). The second analysis consisted of overlapping the current land-use with present and future areas of suitability under different scenarios, if land-use will not change (overall land-use threats and protection). The third analysis deals with the effect of mining only, overlapping areas of environmental suitability with areas directly and indirectly affected by mining in the current scenario and in a mining expansion scenario. The analysis assessed changes in area and connectivity for rupestrian grasslands’ spe- cies in current and future scenarios using Fragstats (McGarigal et  al. 2012). For this analysis, we assumed that all suitable areas are covered by native habitats. We assessed the following landscape metrics: total habitat area, number of patches, mean patch area, mean nearest neighbor (MNN), and proximity indexes. MNN is the average edge-to-edge Euclidean distance between a focal patch and the next nearest patch. Proximity is the asso- ciation between area and inter-patch connectivity. Increasing proximity indices means that the landscape is more structurally connected, while low values indicate a decrease in con- nectivity. We used a 5000 m threshold as a search radius, assuming this is a reasonable distance for the dispersion of plants and vertebrates in this time scale. To measure how much area is being (and will be) affected by intensive use, as well as how much area is (and will be) within protected areas, we overlapped current land-use with present and future areas of suitability. This analysis assumed that land-use would not change with time. Silviculture, intensive agriculture (soybeans, sugar cane, corn, cotton, rice, wheat, bean, coffee, orange, tobacco, cocoa, banana, and cassava) and urban areas for the entire Brazilian territory at a 30-m resolution were acquired from the Centro de Senso- riamento Remoto—CSR UFMG and kindly made available by Britaldo Soares-Filho. Maps of current mining activities were acquired from the Departamento Nacional da Produção Mineral (DNPM) website (http://sigmi ne.dnpm.gov.br/sad69 /UF.zip, April 2013) and com- prised areas under different phases of activity development: exploration, licensing, conces- sion, and exploitation. We overlapped the current mining maps with predicted areas of suitability in the present and future. Assuming that indirect impacts of mining reach 5 km (Durán et al. 2013), we built buffers around mines and calculated the proportion of the areas of suitability affected both directly and indirectly by mining. Additionally, we created mining expansion (ME) scenarios, in which all current areas under licensing or requesting mining authorization were considered; which were also overlapped with future areas of suitability. The ME sce- nario is the best approximation of where future mines will be placed, although this is a rather conservative approximation, as many mines could be implemented outside these areas. Finally, our models were built considering the following 12 scenarios: (1 and 2) current (considering current suitability and mining distribution), (3 and 4) 2050 RCP4.5 (climatic optimistic and current mine distribution), (5 and 6) 2050 RCP4.5 ME (climatic optimistic and mining expansion), (7 and 8) 2050 RCP8.5 (climatic pessimistic and current http://sigmine.dnpm.gov.br/sad69/UF.zip 2594 Biodivers Conserv (2018) 27:2587–2603 1 3 mining distribution), (9 and 10) 2070 RCP4.5 (climatic optimistic and current mining dis- tribution) and (11 and 12) 2070 RCP8.5 ME (climatic pessimistic and mining expansion scenario). Results Predictive modeling The current area of environmental suitability for rupestrian grasslands in Brazil is approxi- mately 83,000  km2 (South America Albers Equal Area Conic projection, “10 percentile training presence logistic threshold”), surpassing the area previously estimated by Fer- nandes et al. (2014) and Silveira et al. (2016). In our models, suitability was largely driven by topography and climate. Altitude, temperature seasonality, and annual precipitation were deemed the most important variables in the model (Table  S3). The model showed a satisfactory performance with an AUC value of 0.972 (± 0.003, SD), which indicates a high accuracy. The current distribution of rupestrian grasslands is largely affected by topography/lithol- ogy and climate and, as such, one could intuitively derive the effects climate change might have on its range and its conservation. In fact, under a more optimistic scenario (RCP4.5), we estimated a total loss of ca. 57,300 km2 (69%) of suitable areas for rupestrian grass- lands by 2050 and a loss of ca. 60,000 km2 (72%) of suitable area by 2070. This loss was concentrated in the regions of Chapada Diamantina, northeastern parts of the Espinhaço and in the south of Minas Gerais and Goiás states (Fig. 3a, c). Additionally, all the isolated mountains in northeastern and northern parts of Brazil will lose suitability for rupestrian grasslands. Under the more pessimistic scenario (RCP8.5), the loss of suitable areas would reach ca. 60,500 km2 (73%) by around 2050, and ca. 68,380 km2 (82%) in 2070, with areas only remaining in the southern part of the Espinhaço and fragmented areas in Chapada Diaman- tina, in the Iron Quadrangle, and in Serra da Canastra (Fig. 3b, d). More stable regions, likely to remain climatically suitable until the end of this century are the mountains of southern Minas Gerais, Ouro Preto, Caraça, Serra do Cipó, Canastra, part of the Diaman- tina plateau and fragments in the northern part of the Espinhaço range (Fig. 3). Currently, less than 10% of the distribution of this ecosystem is protected (ca. 7720 km2), which is below the 17% target according to the Convention on Biological Diversity (CBD; https :// www.cbd.int/sp/targe ts/). Our results show this scenario can worsen in the near future due to the loss of suitable areas in the face of climate change. Loss of suitable areas for rupes- trian grasslands could effectively shrink these conservation units by ca. 40% of the current size in 2050, reaching up to 55% in 2070 (Table 2). Landscape metrics and land‑use threats/protection Even assuming that rupestrian grasslands will cover all suitable areas predicted by the model, the average patch size will decrease by more than half in 2070. More disturbingly, patches would be getting smaller and their number will decrease up to four-fold in the worst scenario. Due to habitat fragmentation, patches will be more separated as shown by a slight increase in the mean nearest neighbor distance. Also, connectivity indices that con- sider patch area, such as proximity, will be halved in the 2070 pessimistic scenario (Fig. 4). https://www.cbd.int/sp/targets/ https://www.cbd.int/sp/targets/ 2595Biodivers Conserv (2018) 27:2587–2603 1 3 We found that environmentally suitable areas for rupestrian grasslands will decrease from 8.3 million ha to less than 2.6 million in any of the future scenarios (Fig. 5a). Assum- ing that current land-uses will be adhered to, the area under intensive use (i.e., agriculture, silviculture, urban and mining) will decrease from the currently 1.2 million hectares to a Fig. 3 Projection of environmental suitability for rupestrian grasslands centered around 2050 in an opti- mistic RCP 4.5 (a) and in a more pessimistic RCP 8.5 scenario (b). Projection of environmental suitability for rupestrian grasslands centered around 2070 in an optimistic RCP 4.5 (c) and in a more pessimistic RCP 8.5 scenario (d) 2596 Biodivers Conserv (2018) 27:2587–2603 1 3 number between 0.197 and 0.076 million ha in the future scenarios. This means that, as the total area of the rupestrian grasslands shrink, many areas currently under mining and agriculture pressure will not be adequate for the biota of rupestrian grasslands in terms of bioclimatic variables explaining the decline in total impacted area. Nevertheless, these esti- mates assume no changes in land-use, which is clearly unrealistic. Additionally, the pro- portion of the habitat area under intensive pressures is maintained in the future scenarios, especially if indirect impacts of mining are considered and in the mining expansion sce- narios. Therefore, the relative impact of intensive managements remains similar if land-use does not change, and probably increase if agricultural and silviculture practices expand. If the current network of protected areas is maintained, 1.2 million hectares (0.83 mil- lion hectares of strictly protected and 0.42 million hectares of sustainable uses reserves) will fall to a number between 0.71 and 0.53 million hectares in the future scenarios. Cur- rently, 0.29 million hectares of rupestrian grasslands are under direct effect of mining, which is about 3.5% of the total area of suitability (Fig. 5b). If both indirectly and directly Table 2 Loss of protected areas in rupestrian grasslands for the decades centered around 2050 and 2070, under two representative pathways: RCP 4.5 and RCP 8.5 Year Representative pathways Loss of protected areas (%) Loss of protected areas (km2) 2050 RCP 4.5 39.19 3025.09 RCP 8.5 40.62 3135.43 2070 RCP 4.5 45.10 3481.10 RCP 8.5 54.84 4232.98 Fig. 4 Landscape metric changes from the current situation to future scenarios according to different land- scape indices. It was assumed that all suitable areas are covered by natural habitats 2597Biodivers Conserv (2018) 27:2587–2603 1 3 affected areas are considered, this value rises to about 1.5 million ha, or 17.57% of the ecosystem’s distribution. Assuming all mines that are currently under licensing, or in any stage of authorization will become effective, the proportional area affected (directly and Fig. 5 Area under different land-uses in the current and future scenarios, assuming that land-use is kept in current conditions (a). Proportion of suitability areas for rupestrian grasslands indirectly and directly affected by mining and according to two mining scenarios (current, and mining expansion - ME) in differ- ent future scenarios (b) 2598 Biodivers Conserv (2018) 27:2587–2603 1 3 indirectly) may be more than 60% in the future scenarios. Because we expect that mines will increase beyond areas that are currently under licensing, the actual mined area can be even larger than the scenario of mining expansion predicts. Discussion While there has been no significant expansion of the distribution of rupestrian grass- lands since at least the Last Glacial Maximum (ca. 20,000 years B.P.) (Barbosa and Fer- nandes 2016), future scenario predictions are catastrophic for this ecosystem. All models applied here showed extremely negative results in terms of maintenance of suitable areas for conservation of rupestrian grasslands under climate change, reaching losses of up to 82% within the next 50 years. We also showed that a small proportion of protected areas become less effective due to the loss of suitability, land-use changes, mining and lack of adequate knowledge and policing for its conservation and management (e.g., Fernandes 2016b). Moreover, more than 50 million people that depend on ecosystem services pro- vided by rupestrian grasslands (Neves et al. 2016) could be directly or indirectly impacted, extending the pressure on an already overexploited land. Although 10 strictly protected areas were created in rupestrian grasslands in the last decade, safeguarding about 116,000 ha (Silva et al. 2008), unique elements of biodiversity are likely to be under-represented due to species turnover and high endemism (Fernandes 2016b; Monteiro et al. 2018). Although the mining industry has acquired several proper- ties in the Espinhaço range to offset the impacts of mining, most of them have not been officially transformed into protected areas yet and, therefore, not included in the analyses performed to date (GWF, unpub. data). Although in this region offsetting efforts has rarely reduced vegetation loss and are themselves often threatened by future mining (see Sonter et al. 2014b), we ought to build spaces and opportunities to create a pact that will lead to a more rational use and conservation of the rupestrian grasslands. The case of Brazilian rupestrian grasslands calls for profound scientific, conservation and political consideration due to their unparalleled biodiversity and levels of endemism, plus their strategic importance as a source of goods, including water, ornamental species, and scenic beauty for sustainable tourism, especially for the densely populated southeast- ern Brazil (Resende and Fernandes 2013; Resende et  al.  2017). Mountain environments in the Cerrado have until recently been relatively free from huge extensive cattle farming and habitat-aggressive plantations, such as soybean and sugar cane (Gibbs et  al. 2015). However, this situation is not likely to continue, as indicated by uncontrolled habitat con- version and the disturbances highlighted here. The worsening fate of rupestrian grass- lands could scale-up both spatially and temporally if we consider the fast pace of biologi- cal invasions (Barbosa et  al. 2010; Fernandes et  al. 2015), mining activities (Fernandes et al. 2014; Sonter et al. 2014a) and afforestation projects (Veldman et al. 2015; Fernandes et al. 2016b). The synergy of these strong drivers would result in major land-use change leading to strong impacts from which the low-resilience rupestrian grasslands might not recover, perhaps leading to novel and much more simplified ecosystems both botanically (Fernandes et al. 2014) and zoologically (Dirzo et al. 2014) that are deficient in the provi- sion of ecosystem services. Currently, ca. 15% of rupestrian grasslands are subjected to intensive anthropogenic impact (areas directly affected by mining, silviculture, urbanization and intensive agricul- ture). Considering the indirect impacts of mining reach 5 km beyond the mines themselves 2599Biodivers Conserv (2018) 27:2587–2603 1 3 (a very conservative estimate, Durán et al. 2013), as much as 18% of the ecosystem is cur- rently under the influence of mining alone (Fig. 5). The projected expansion of the mining sector in rupestrian grasslands would lead to an unprecedented cascade of impacts on the ecosystem (Fernandes et al. 2014; Fernandes and Ribeiro 2017; Pena et al. 2017). If the expansion of mining comes associated with climate change, the direct and indirect effects of mining would reach about 60% of the total future distribution of rupestrian grasslands under the pessimistic and optimistic 2045 and 2070 climatic scenarios (Fig. 3). This hap- pens since the distribution of future mines largely overlaps with suitable areas of rupestrian grasslands potential distribution. One anthropogenic threat that disrupts the rupestrian grassland dynamics, not yet men- tioned or included in our models and predictions, is the occurrence of anthropogenic fires. Although rupestrian grassland is a fire-prone vegetation, adapted to natural fires (Warming 1892; Silveira et al. 2016; Morellato and Silveira 2018), the time, intensity and frequency of human-induced fires impose additional stresses to plants, with unforeseen effects on the entire ecosystem (e.g, Figueira et  al. 2016; see also Bond and Keeley 2005). Rupestrian grasslands are moisture-dependent, fire regime ecosystems, with the ignition influenced mostly by the length of the dry season and the rainfall distribution along the season (Alva- rado et  al. 2017). The effects of changes in fire regime, caused by anthropogenic activi- ties or indirectly through climate change, are still unknown for this ecosystem and concur with mining and other threats. Monitoring of the vegetation recovering post-fire is needed (Figueira et al. 2016; Alberton et al. 2017) to evaluate the ecosystem’s potential resilience as well as whether the effects of anthropogenic fire regime are positive or negative in the areas disturbed by mining and other activities, integrating the management plan of conser- vation units. Concluding remarks and future perspectives Proactive and long-lasting actions are urgently needed to preserve this ecosystem and its irreplaceable ecosystem services for future generations. In the next few decades we will not just face loss of biodiversity and ecosystems services and processes, but we also risk losing an important cultural heritage (Neves et al. 2016; Fernandes and Ribeiro 2017). A first step towards effective conservation of rupestrian grasslands is the translation, to all sectors of society, of the knowledge and importance of this ecosystem and the magnitude of the loss due to its conversion and misuse. For instance, the consequences of the loss of environ- mental services and of the overexploitation of natural resources leading to deterioration of human wellbeing is an aspect that needs to become appreciated (see Biénabe and Hearne 2006; Resende et al. 2017). Conservation actions, however, are challenged by a number of basic limitations, including that the vast majority of the conservation units in the region lack planning, personnel, and the engagement of communities. The problem becomes even more complex and challenging as conservation plans developed for forested areas, can- not be transferred by governmental institutions to these grassland ecosystems, and instead need to be specifically developed for non-forest landscapes. This is an aspect that needs to be assimilated by local governmental institutions if they are to administer one of the most diverse and neglected ecosystems of Brazil (Fernandes 2016b). This comes at a time when increasing land-use change is accelerating. Last, but not least, while the Brazilian panel of climate change (Ambrizzi and Araujo 2014) predicts an increase of ca. 5 °C in temperature and a decrease of 35–45% in rainfall by the end of this century in central Brazil, no spe- cific information has been provided for rupestrian grasslands or the Espinhaço mountain 2600 Biodivers Conserv (2018) 27:2587–2603 1 3 range. The above scenario represents an enormous problem from a conservation perspec- tive, firstly because this mountain ecosystem holds unique elements of Brazilian biodiver- sity and ca. 50% of the species of the Cerrado (i.e., the most diverse savanna of the world and the second largest biome in South America: Fernandes 2016c), and secondly because mountaintops in this region hold the headwaters of major rivers and watersheds (river basins occupy an area of 1.21 million km2) that provide water to 50 million people and to mining, industry, agriculture, aquaculture, fishing, and transportation in southeastern and northeastern Brazil. Clearly, rupestrian grasslands represent a critical conservation agenda. Acknowledgements We thank the Conselho Nacional de Pesquisas (CNPq) for funding the Long-Term Ecological Research (PELD-CRSC-17), the Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG), Coordenadoria de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), PPBio/MCTIC, PRPq/UFMG and Neotropical Grassland Conservation (NGC) for financial support. We are also grate- ful to the Company Cedro Têxtil, Reserva Vellozia, Parque Nacional da Serra do Cipó, GSG and Pousada Serra Morena for logistical support. NPUB received a CNPq-PDJ scholarship (154664/2016-2), FFG, TJG received a CNPq-PNPD scholarship, BA received a PhD scholarship from FAPESP (Grant #2014/0215-0), RD was supported by Stanford University unrestricted funds, and LPCM was supported by FAPESP (Grant #2013/50155-0). RS received support from PRPq/UFMG 005/2016.  AFB received support from Rede Clima (FINEP/CNPq). LPCM and GWF received a research productivity fellowship from CNPq. References Alberton B, Torres RS, Cancian LF, Borges BD, Almeida J, Mariano G, Morellato LPC (2017) Introducing digital cameras to monitor plant phenology in the tropics: applications for conservation. Perspect Ecol Conserv. https ://doi.org/10.1016/j.pecon .2017.06.004 Alvarado ST, Fornazari T, Costola A, Morellato LPC, Silva TSF (2017) Drivers of fire occurrence in a mountainous Brazilian savanna: tracking long-term fire regimes using remote sensing. Ecol Indic. https ://doi.org/10.1016/j.ecoli nd.2017.02.037 Alves CBM, Leal CG, Brito MFG, Santos ACA (2008) Biodiversidade e conservação de peixes do Com- plexo do Espinhaço. Megadiversidade 4:177–196 Alves RJV, Silva NG, Oliveira JA, Medeiros D (2014) Circumscribing campo rupestre: megadiverse Brazil- ian rocky montane savannas. Braz J Biol. https ://doi.org/10.1590/1519-6984.23212 Ambrizzi T, Araujo M (2014) Base científica das mudanças climáticas—Contribuição do Grupo de Tra- balho 1 do Painel Brasileiro de Mudanças Climáticas ao Primeiro Relatório da Avaliação Nacional sobre Mudanças Climáticas. COPPE/UFRJ, Rio de Janeiro Barbosa NPU, Fernandes GW (2016) Rupestrian grassland: past, present and future distribution. In: Fer- nandes GW (ed) Ecology and conservation of mountaintop grasslands in Brazil. Springer, Switzerland, pp 531–544 Barbosa NPU, Fernandes GW, Carneiro MAA, Júnior LAC (2010) Distribution of non-native invasive spe- cies and soil properties in proximity to paved roads and unpaved roads in a quartzitic mountainous grassland of south-eastern Brazil (rupestrian fields). Biol Invasions. https ://doi.org/10.1007/s1053 0-010-9767-y Barbosa NPU, Fernandes GW, Sanchez-Azofeifa A (2015) A relict species restricted to a quartzitic moun- tain in tropical America: an example of microrefugium? Acta Bot Bras. https ://doi.org/10.1590/0102- 33062 014ab b3731 Biénabe E, Hearne RR (2006) Public preferences for biodiversity conservation and scenic beauty within a framework of environmental services payments. For Pol Econ. https ://doi.org/10.1016/j.forpo l.2005.10.002 Bond WJ, Keeley JE (2005) Fire as a global ‘herbivore’: the ecology and evolution of flammable ecosys- tems. Trends Ecol Evol. https ://doi.org/10.1016/j.tree.2005.04.025 Carnaval AC, Moritz C (2008) Historical climate modeling predicts patterns of current biodiversity in the Brazilian Atlantic forest. J Biogeogr. https ://doi.org/10.1111/j.1365-2699.2007.01870 .x Carvalho F, Souza FA, Carrenho R, Moreira FMS, Jesus EC, Fernandes GW (2012) The mosaic of habitats in the high-altitude Brazilian rupestrian fields is a hotspot for arbuscular mycorrhizal fungi. Appl Soil Ecol. https ://doi.org/10.1016/j.apsoi l.2011.10.001 https://doi.org/10.1016/j.pecon.2017.06.004 https://doi.org/10.1016/j.ecolind.2017.02.037 https://doi.org/10.1590/1519-6984.23212 https://doi.org/10.1007/s10530-010-9767-y https://doi.org/10.1007/s10530-010-9767-y https://doi.org/10.1590/0102-33062014abb3731 https://doi.org/10.1590/0102-33062014abb3731 https://doi.org/10.1016/j.forpol.2005.10.002 https://doi.org/10.1016/j.forpol.2005.10.002 https://doi.org/10.1016/j.tree.2005.04.025 https://doi.org/10.1111/j.1365-2699.2007.01870.x https://doi.org/10.1016/j.apsoil.2011.10.001 2601Biodivers Conserv (2018) 27:2587–2603 1 3 Chaves AV, Freitas GHS, Vasconcelos MF, Santos FR (2015) Biogeographic patterns, origin and specia- tion of the endemic birds from eastern Brazilian mountaintops: a review. Syst Biodivers. https ://doi. org/10.1080/14772 000.2014.97247 7 Coutinho ES, Fernandes GW, Berbara RL, Valério HM, Goto BT (2015) Variation of arbuscular mycor- rhizal fungal communities along an altitudinal gradient in rupestrian grasslands in Brazil. Mycorrhiza. https ://doi.org/10.1007/s0057 2-015-0636-5 Dirzo R, Young HR, Galetti M, Ceballos G, Isaac NJB, Collen B (2014) Defaunation in the Anthropocene. Science. https ://doi.org/10.1126/scien ce.12518 17 Durán AP, Rauch J, Gaston KJ (2013) Global spatial coincidence between protected areas and metal mining activities. Biol Conserv. https ://doi.org/10.1016/j.bioco n.2013.02.003 Echternacht L, Trovó M, Oliveira CT, Pirani JR (2011) Areas of endemism in the Espinhaço range in Minas Gerais, Brazil. Flora. https ://doi.org/10.1016/j.flora .2011.04.003 Elith J, Phillips SJ, Hastie T, Dudík M, Chee YE, Yates CJ (2011) A statistical explanation of MaxEnt for ecologists. Divers Distrib. https ://doi.org/10.1111/j.1472-4642.2010.00725 .x Epps CW, Palsbøll PJ, Wehausen JD, Roderick GK, McCullough DR (2006) Elevation and connectivity define genetic refugia for mountain sheep as climate warms. Mol Ecol. https ://doi.org/10.1111/j.1365- 294X.2006.03103 .x Fernandes GW (2016a) Ecology and conservation of mountaintop grasslands in Brazil. Springer, Switzerland Fernandes GW (2016b) The shady future of the rupestrian grassland: major threats to conservation and challenges in the Anthropocene. In: Fernandes GW (ed) Ecology and conservation of mountaintop grasslands in Brazil. Springer, Switzerland, pp 545–561 Fernandes GW (2016c) The megadiverse rupestrian grassland. In: Fernandes GW (ed) Ecology and conser- vation of mountaintop grasslands in Brazil. Springer, Switzerland, pp 3–14 Fernandes GW, Ribeiro SP (2017) Deadly conflicts: mining, people, and conservation. Perspect Ecol Con- serv. https ://doi.org/10.1016/j.pecon .2017.09.002 Fernandes GW, Santos JC (2014) Neotropical insect galls. Springer, Netherlands Fernandes GW, Barbosa NPU, Negreiros D, Paglia AP (2014) Challenges for the conservation of vanishing megadiverse rupestrian grasslands. Nat Conserv. https ://doi.org/10.1016/j.ncon.2014.08.003 Fernandes GW, Santos R, Barbosa NPU, Almeida HA, Carvalho V, Angrisano P (2015) Ocorrência de plantas não nativas e exóticas em áreas restauradas de campos rupestres. Planta Daninha. https ://doi. org/10.1590/S0100 -83582 01500 03000 09 Fernandes GW, Pedroni F, Sanchez M, Scariot A, Aguiar LMS, Ferreira G, Machado R, Ferreira ME, Pin- heiro SDR, Costa JAS, Dirzo R, Muniz F (2016a) Cerrado: Em Busca de Soluções Sustentáveis. Vert- ente, Rio de Janeiro Fernandes GW, Coelho MS, Machado RB, Ferreira ME, de Souza Aguiar LM, Dirzo R, Scariot A, Lopes CR (2016b) Afforestation of savannas: an impending ecological disaster. Nat Conserv. https ://doi. org/10.1016/j.ncon.2016.08.002 Figueira JEC, Ribeiro KT, Ribeiro MC, Jacobi CM, França H, Neves ACO, Conceição AA, Mourão FA, Souza JM, Miranda CAK (2016) Fire in rupestrian grasslands: plant response and management. In: Fernandes GW (ed) Ecology and conservation of mountaintop grasslands in Brazil. Springer, Switzer- land, pp 415–448 Freitas GHS, Chaves AV, Costa LM, Santos FR, Rodrigues M (2012) A new species of Cinclodes from the Espinhaço Range, southeastern Brazil: insights into the biogeographical history of the South American highlands. Ibis. https ://doi.org/10.1111/j.1474-919X.2012.01268 .x Garcia RA, Cabeza M, Altwegg R, Araújo MB (2016) Do projections from bioclimatic envelope models and climate change metrics match? Glob Ecol Biogeogr. https ://doi.org/10.1111/geb.12386 Gibbs HK, Rausch L, Munger J, Schelly I, Morton DC, Noojipady P, Soares-Filho B, Barreto P, Micol L, Walker NF (2015) Brazil’s soy moratorium. Science. https ://doi.org/10.1126/scien ce.aaa01 81 Giulietti AM, Pirani JR, Harley RM (1997) Espinhaço range region, eastern Brazil. In: Davis SD, Heywood VH, Herrera-MacBryde O, Villa-Lobos J, Hamilton AC (eds) Centres of plant diversity: a guide and strategy for their conservation. WWF/IUCN, Cambridge, pp 397–404 Hernandez PA, Graham CH, Master LL, Albert DL (2006) The effect of sample size and species character- istics on performance of different species distribution modeling methods. Ecography. https ://doi.org/10 .1111/j.0906-7590.2006.04700 .x Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. Int J Climatol 25:1965–1978 IPCC (2014) Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri RK, Meyer, LA (eds)]. IPCC, Geneva https://doi.org/10.1080/14772000.2014.972477 https://doi.org/10.1080/14772000.2014.972477 https://doi.org/10.1007/s00572-015-0636-5 https://doi.org/10.1126/science.1251817 https://doi.org/10.1016/j.biocon.2013.02.003 https://doi.org/10.1016/j.flora.2011.04.003 https://doi.org/10.1111/j.1472-4642.2010.00725.x https://doi.org/10.1111/j.1365-294X.2006.03103.x https://doi.org/10.1111/j.1365-294X.2006.03103.x https://doi.org/10.1016/j.pecon.2017.09.002 https://doi.org/10.1016/j.ncon.2014.08.003 https://doi.org/10.1590/S0100-83582015000300009 https://doi.org/10.1590/S0100-83582015000300009 https://doi.org/10.1016/j.ncon.2016.08.002 https://doi.org/10.1016/j.ncon.2016.08.002 https://doi.org/10.1111/j.1474-919X.2012.01268.x https://doi.org/10.1111/geb.12386 https://doi.org/10.1126/science.aaa0181 https://doi.org/10.1111/j.0906-7590.2006.04700.x https://doi.org/10.1111/j.0906-7590.2006.04700.x 2602 Biodivers Conserv (2018) 27:2587–2603 1 3 Khanum R, Mumtaz AS, Kumar S (2013) Predicting impacts of climate change on medicinal asclepiads of Pakistan using Maxent modeling. Acta Oecol. https ://doi.org/10.1016/j.actao .2013.02.007 Le Stradic S, Buisson E, Negreiros D, Campagne P, Fernandes GW (2014) The role of native woody species in the restoration of campos rupestres in quarries. Appl Veg Sci. https ://doi.org/10.1111/ avsc.12058 Leite FSF, Juncá FA, Eterovick PC (2008) Status do conhecimento, endemismo e conservação de anfíbios anuros da Cadeia do Espinhaço, Brasil. Megadiversidade 4:182–200 Liu C, Berry PM, Dawson TP, Pearson RG (2005) Selecting thresholds of occurrence in the prediction of species distributions. Ecography. https ://doi.org/10.1111/j.0906-7590.2005.03957 .x Lüttge U (2017) Die erkundung der vegetation Brasiliens im 19. Jahrhundert. Hoppea Denkschr Regensb Bot Ges 78:23–44 McGarigal K, Cushman SA, Ene E (2012) FRAGSTATS v4: spatial pattern analysis program for cat- egorical and continuous maps. Computer software program produced by the authors at the Uni- versity of Massachusetts, Amherst. http://www.umass .edu/lande co/resea rch/frags tats/frags tats.html Monteiro L, Machado N, Martins E, Pougy N, Verdi M, Martinelli G, Loyola RD (2018) Conservation priorities for the threatened flora of mountaintop grasslands in Brazil. Flora 238:234–243. https :// doi.org/10.1016/j.flora .2017.03.007 Morellato LPC, Silveira FAO (2018) Plant life on campo rupestre: new lessons from an ancient biodiver- sity hotspot. Flora 238:1–10. https ://doi.org/10.1016/j.flora .2017.12.001 Mügge FLB, Paula-Souza J, Melo JC, Brandão MGL (2016) Native plant species with economic value from Minas Gerais and Goiás: a discussion on the currentness of the data recovered by the French naturalist Auguste de Saint-Hilaire. Hortic Bras 34:455–462. https ://doi.org/10.1590/s0102 -05362 01604 02 Negreiros D, Le Stradic S, Fernandes GW, Rennó HC (2014) CSR analysis of plant functional types in highly diverse tropical grasslands of harsh environments. Plant Ecol. https ://doi.org/10.1007/s1125 8-014-0302-6 Neves ACO, Barbieri AF, Pacheco AA, Resende FM, Braga RF, Azevedo AA, Fernandes GW (2016) The human dimension in the Espinhaço Mountains: land conversion and ecosystem services. In: Fernandes GW (ed) Ecology and conservation of mountaintop grasslands in Brazil. Springer, Swit- zerland, pp 501–530 Nishi AH, Vasconcellos-Neto J, Romero GQ (2013) The role of multiple partners in a digestive mutual- ism with a protocarnivorous plant. Ann Bot. https ://doi.org/10.1093/aob/mcs24 2 Oliveira RS, Abrahão A, Pereira C, Teodoro GS, Brum M, Alcantara S, Lambers H (2016) Ecophysiol- ogy of campos rupestres plants. In: Fernandes GW (ed) Ecology and conservation of mountaintop grasslands in Brazil. Springer, Switzerland, pp 227–272 Overbeck GE, Vélez-Martin E, Scarano FR, Lewinsohn TM, Fonseca CR, Meyer ST, Müller SC, Ceotto P, Dadalt L, Durigan G, Ganade G, Gossner MM, Guadagnin DL, Lorenzen K, Jacobi CM, Weisser WW, Pillar VD (2015) Conservation in Brazil needs to include non-forest ecosystems. Divers Dis- trib. https ://doi.org/10.1111/ddi.12380 Pardiñas UFJ, Lessa G, Teta P, Salazar-Bravo J, Câmara EMVC (2014) A new genus of Sigmodon- tine rodent from eastern Brazil and the origin of the tribe Phyllotini. J Mammal. https ://doi. org/10.1644/13-MAMM-A-208 Pena JCC, Goulart F, Fernandes GW, Hoffmann D, Leite FSF, Santos NB, Soares-Filho B, Sobral-Souza T, Vancine MH, Rodrigues M (2017) Impacts of mining activities on the potential geographic distribution of eastern Brazil mountaintop endemic species. Perspect Ecol Conserv. https ://doi. org/10.1016/j.pecon .2017.07.005 Pereira CG, Almenara DP, Winter CE, Fritsch PW, Lambers H, Oliveira RS (2012) Underground leaves of Philcoxia trap and digest nematodes. PNAS. https ://doi.org/10.1073/pnas.11141 99109 Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distri- butions. Ecol Model. https ://doi.org/10.1016/j.ecolm odel.2005.03.026 Resende F, Fernandes GW (2013) Economic valuation of plant diversity storage service provided by Brazil- ian rupestrian grassland ecosystems. Braz J Biol. https ://doi.org/10.1590/S1519 -69842 01300 04000 05 Resende F, Fernandes GW, Andrade DC, Néder HD (2017) Economic valuation of the ecosystem ser- vices provided by a protected area in the Brazilian Cerrado: application of the contingent valuation method. Braz J Biol. https ://doi.org/10.1590/1519-6984.21215 Santos JC, Leal IR, Almeida-Cortez JS, Fernandes GW, Tabarelli M (2011) Caatinga: the scientific neg- ligence experienced by a dry tropical forest. Trop Conserv Sci. https ://doi.org/10.1177/19400 82911 00400 306 Schaefer CEGR, Corrêa GR, Candido HG, Arruda DM, Nunes JA, Araujo RW, Rodrigues PMS, Filho EIF, Pereira AFS, Brandão PC, Neri AV (2016) The physical environment of rupestrian grasslands https://doi.org/10.1016/j.actao.2013.02.007 https://doi.org/10.1111/avsc.12058 https://doi.org/10.1111/avsc.12058 https://doi.org/10.1111/j.0906-7590.2005.03957.x http://www.umass.edu/landeco/research/fragstats/fragstats.html https://doi.org/10.1016/j.flora.2017.03.007 https://doi.org/10.1016/j.flora.2017.03.007 https://doi.org/10.1016/j.flora.2017.12.001 https://doi.org/10.1590/s0102-053620160402 https://doi.org/10.1590/s0102-053620160402 https://doi.org/10.1007/s11258-014-0302-6 https://doi.org/10.1007/s11258-014-0302-6 https://doi.org/10.1093/aob/mcs242 https://doi.org/10.1111/ddi.12380 https://doi.org/10.1644/13-MAMM-A-208 https://doi.org/10.1644/13-MAMM-A-208 https://doi.org/10.1016/j.pecon.2017.07.005 https://doi.org/10.1016/j.pecon.2017.07.005 https://doi.org/10.1073/pnas.1114199109 https://doi.org/10.1016/j.ecolmodel.2005.03.026 https://doi.org/10.1590/S1519-69842013000400005 https://doi.org/10.1590/1519-6984.21215 https://doi.org/10.1177/194008291100400306 https://doi.org/10.1177/194008291100400306 2603Biodivers Conserv (2018) 27:2587–2603 1 3 (campos rupestres) in Brazil: geological, geomorphological and pedological characteristics, and interplays. In: Fernandes GW (ed) Ecology and conservation of mountain top grasslands in Brazil. Springer, Switzerland, pp 15–53 Schrag M, Bunn AG, Graumlich LJ (2008) Influence of bioclimatic variables on treeline conifer distribution in the Greater Yellowstone ecosystem: implications for species of conservation concern. J Biogeogr. https ://doi.org/10.1111/j.1365-2699.2007.01815 .x Silva JA, Machado RB, Azevedo AA, Drumond GM, Fonseca RL, Goulart MF, Júnior EAM, Martins CS, Neto MBR (2008) Identificação de áreas insubstituíveis para conservação da Cadeia do Espinhaço, estados de Minas Gerais e Bahia, Brasil. Megadiversidade 4:248–270 Silveira FAO, Negreiros D, Barbosa NPU, Buisson E, Carmo FF, Carstensen DW, Conceição AA, Cornelis- sen TG, Echternacht L, Fernandes GW, Garcia QS, Guerra TJ, Jacobi CM, Lemos-Filho JP, Le Stradic S, Morellato LPC, Neves FS, Oliveira RS, Schaefer CE, Viana PL, Lambers H (2016) Ecology and evolution of plant diversity in the endangered campo rupestre: a neglected conservation priority. Plant Soil. https ://doi.org/10.1007/s1110 4-015-2637-8 Sonter LJ, Moran CJ, Barrett DJ, Soares-Filho BS (2014a) Processes of land use change in mining regions. J Clean Prod. https ://doi.org/10.1016/j.jclep ro.2014.03.084 Sonter LJ, Barrett DJ, Soares-Filho BS (2014b) Offsetting the impacts of mining to achieve no net loss of native vegetation. Conserv Biol. https ://doi.org/10.1111/cobi.12260 Spehn EM, Rudmann-Maurer K, Körner C, Maselli D (2010) Mountain biodiversity and global change. GMBA-DIVERSITAS, Basel Swets JA (1988) Measuring the accuracy of diagnostic systems. Science 240:1285–1293 Vasconcelos MF, Lopes LE, Machado CG, Rodrigues M (2008) As aves dos campos rupestres da Cadeia do Espinhaço: diversidade, endemismo e conservação. Megadiversidade 4:221–241 Veldman JW, Overbeck GE, Negreiros D, Mahy G, Le Stradic S, Fernandes GW, Durigan G, Buisson E, Putz FE, Bond WJ (2015) Where tree planting and forest expansion are bad for biodiversity and eco- system services. Bioscience. https ://doi.org/10.1093/biosc i/biv11 8 Warming E (1892) Lagoa Santa: et Bidrag til den biologiske Plantegeografi: med en Fortegnelse over Lagoa Santas Hvirveldyr. Kongelige Danske Videnskabernes Selskabs Skrifter. Naturvidenskabelig og Math- ematisk Afdeling, 6. Rk. 6:153–488 Warming E (1895) Plantesamfund—Grundtræk af den økologiske Plantegeografi. P.G. Philipsens Forlag, Kjøbenhavn https://doi.org/10.1111/j.1365-2699.2007.01815.x https://doi.org/10.1007/s11104-015-2637-8 https://doi.org/10.1016/j.jclepro.2014.03.084 https://doi.org/10.1111/cobi.12260 https://doi.org/10.1093/biosci/biv118 The deadly route to collapse and the uncertain fate of Brazilian rupestrian grasslands Abstract Introduction Materials and methods Identification of priority regions Environmental data Predictive modeling Landscape metrics and land-use threatsprotection Results Predictive modeling Landscape metrics and land-use threatsprotection Discussion Concluding remarks and future perspectives Acknowledgements References