Caves pose significant challenges for mining projects, since they harbor many endemic and threatened species, and must therefore be protected. Recent discussions between academia, environmental protection agencies, and industry partners, have highlighted problems with the current Brazilian legislation for the protection of caves. While the licensing process is long, complex and cumbersome, the criteria used to assign caves into conservation relevance categories are often subjective, with relevance being mainly determined by the presence of obligate cave dwellers (troglobites) and their presumed rarity. However, the rarity of these troglobitic species is questionable, as most remain unidentified to the species level and their habitats and distribution ranges are poorly known. Using data from 844 iron caves retrieved from different speleology reports for the Carajás region (South-Eastern Amazon, Brazil), one of the world's largest deposits of high-grade iron ore, we assess the influence of different cave characteristics on four biodiversity proxies (species richness, presence of troglobites, presence of rare troglobites, and presence of resident bat populations). We then examine how the current relevance classification scheme ranks caves with different biodiversity indicators. Large caves were found to be important reservoirs of biodiversity, so they should be prioritized in conservation programs. Our results also reveal spatial autocorrelation in all the biodiversity proxies assessed, indicating that iron caves should be treated as components of a cave network immersed in the karst landscape. Finally, we show that by prioritizing the conservation of rare troglobites, the current relevance classification scheme is undermining overall cave biodiversity and leaving ecologically important caves unprotected. We argue that conservation efforts should target subterranean habitats as a whole and propose an alternative relevance ranking scheme, which could help simplify the assessment process and channel more resources to the effective protection of overall cave biodiversity.
Citation: Jaffé R, Prous X, Zampaulo R, Giannini TC, Imperatriz-Fonseca VL, Maurity C, et al. (2016) Reconciling Mining with the Conservation of Cave Biodiversity: A Quantitative Baseline to Help Establish Conservation Priorities. PLoS ONE 11(12): e0168348. https://doi.org/10.1371/journal.pone.0168348
Editor: Judi Hewitt, University of Waikato, NEW ZEALAND
Received: April 20, 2016; Accepted: November 30, 2016; Published: December 20, 2016
Copyright: © 2016 Jaffé et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The study was funded by the Vale Institute of Technology, a non-profit research institution.
Competing interests: Herewith we declare that co-authors Xavier Prous, Robson Zampaulo, and Iuri Viana Brandi are employed by the commercial company Vale. This does not alter our adherence to PLOS ONE policies on sharing data and materials. All authors declare no competing or financial interests associated with the publication of this manuscript.
Decisions about which lands to allocate for conservation often involve conflicting interests between industry, local communities, and environmental protection agencies . Such conflicts can be particularly acute in the mining sector, given that mineral resources are not uniformly distributed but rather spatially clustered, implying that the allocation of conservation areas may be constrained . A common alternative to solve these conflicts has been the use of offsets (conservation areas located outside development areas) to compensate for the residual and unavoidable impacts of development projects on biodiversity [3,4]. Offsets, however, are more difficult to apply in cases where endemic or threatened species occur inside areas containing rich mineral resources.
Caves pose significant challenges for mining projects, since they harbor many such endemic or threatened species, also known as short range endemics because of their small distribution area and assumed restricted dispersal . For instance, different troglobitic species (obligate subterranean dwellers which must complete their entire life cycle in caves ) have halted mining projects worldwide , resulting in economic losses rising to billions of dollars . Brazil has one of the most stringent cave protection regimes in the world, requiring extensive speleological surveys prior to the implementation of any development project . The consulting companies performing these surveys must then assign caves into one of four relevance categories (maximal, high, mid, or low), based on a complex set of biological, geological, and cultural attributes. Caves containing rare troglobitic species, for instance, are always defined as maximal relevance caves, which must be protected along with a buffer area of 250m . High relevance caves, on the other hand, can be impacted if appropriate compensation offsets are provided (i.e. preserving two similar caves).
Recent discussions between academia, environmental protection agencies, and industry partners , have identified problems with the current Brazilian legislation for the protection of caves (Federal Decree 6640/2008), highlighting the need for novel quantitative approaches to determine conservation priorities [8–11]. First, the criteria used to assign caves into relevance categories are often subjective, given that the factors influencing cave biodiversity remain largely unknown [12–16]. Second, the relevance classification process is long, complex and cumbersome . Any cavity with over 5m of extension must be considered in speleology surveys, so these take between one and two years to complete, and involve intensive field work and considerable resources. Third, identifying caves for compensation has become increasingly difficult, because most caves must be compensated . The great majority of caves are classified as high relevance caves, since caves exhibiting any one of a long list of criteria must be classified so. Finally, there are few taxonomists specialized in identifying cave fauna in the country, and they are asked to analyze thousands of specimens collected in areas of high biodiversity. For instance, most troglobites remain unidentified to the species level, some represent new undescribed taxa, and their distribution ranges are usually unknown . In consequence, the characterization of threatened troglobitic species is often questionable.
Here we aim to help optimize the current Brazilian cave protection regime by: 1) Filling part of the knowledge gap regarding the main drivers of cave biodiversity; 2) Examining how the current relevance classification scheme ranks caves with different biodiversity indicators; and 3) Proposing a new scheme to establish conservation priorities. We first perform a large-scale quantitative study relating iron cave biodiversity proxies to different cave characteristics. We analyzed data for 844 iron caves, which contain higher richness of troglomorphic species than caves of other lithologies . The data was retrieved from different speleology reports for the Carajás region (South-Eastern Amazon, Brazil), one of the world's largest deposits of high-grade iron ore . Several mines are already operating in the region and the world's largest iron-ore mine (project S11D) is about to begin operating, so there is now a pressing need to achieve a compromise between mining and the conservation of cave biodiversity in the area. We then examine how the current relevance classification scheme ranks caves with different biodiversity indicators, and discuss the implications of these results for the conservation of cave biodiversity. Finally, we propose a new scheme to establish conservation priorities and estimate environmental offsets, which could help simplify the assessment process and channel more resources to the effective protection of cave biodiversity.
Materials and Methods
We retrieved data contained in eight speleology reports for the region of Carajás, State of Pará, Brazil (see Table A in S1 File). Together, these reports surveyed a total of 844 caves distributed across Serra Norte, Serra Sul, Serra Leste and Serra da Bocaina (Fig 1). All surveys employed similar methods, and reports assigned cave relevance based on the same set of cave attributes (those specified in the current Brazilian legislation for the protection of caves). We gathered all relevant data available on cave attributes and biodiversity indicators, including cave ID, coordinates, altitude, horizontal projection (length), slope, area, volume, presence of percolating water and water reservoirs, cave lithology, presence of plant material, presence of plant detritus, presence of roots, presence of guano, presence of other feces, presence of regurgitation balls, presence of carcases, presence of troglobites and rare troglobites, presence of resident bat populations, and species richness. Cave attributes were surveyed twice a year, during the dry and wet seasons, and total species richness was calculated as the total number of species found in both seasons. Specialist taxonomists classified species as troglobites if they exhibited troglomorphic traits  unknown in phylogenetically related taxa occurring in above-ground habitats. While troglobites found in three or less caves were considered rare, only groups of 100 or more bats were considered resident populations. Stygofauna was not considered in any of the speleology reports. In addition we retrieved from the reports the overall cave relevance classification, determined considering a complex set of biological, geological, and cultural attributes , as well as the biological relevance classification of each cave, determined using biological attributes only. Information on the richness of troglobitic species (for caves containing troglobites) was unavailable in some reports, and in others was either vaguely mentioned in the text or presented in tables with varying formats and nomenclatures. We thus only retrieved data on the richness of troglobitic species when this information was clear and explicitly made available (in 329 caves). Detailed descriptions of the survey methods and taxa inventories for all caves are presented in the original reports (Table A in S1 File and S1 Dataset).
Caves are colored according to their relevance category (N = 843 caves), and the Carajás National Forest is shown in green. The digital elevation raster (SRTM, 1 arc-second) was obtained from USGS Earth Explorer, the National Forest shapefile from ICMBIO, and the South America shapefile from Thematicmapping.org. All layers are copyright-free.
We selected four proxies of cave biodiversity: Total species richness, presence of troglobites, presence of rare troglobites, and presence of resident bat populations. For each one, we constructed a complete-cases dataset (excluding observations containing missing data), to avoid fitting models containing a different number of observations. A principal component analysis showed that length, area, and volume were highly correlated (the first component explaining 95% of total variance). We therefore chose to use cave area as a proxy for cave size, because it reflects cave size more accurately than length alone, and its estimation is less susceptible to measurement errors than volume. All statistical analyses were implemented in R . The full dataset, including the geographic coordinates for the locations of all caves, can be found along with all R scripts as S2 Dataset.
Modeling species richness
To test if species richness was spatially aggregated , we quantified spatial autocorrelation in species richness. We used the R package spdep  to compute Moran's I, a standard measure of spatial autocorrelation ranging from -1 (indicating perfect dispersion) to +1 (perfect correlation, with zero indicating a random spatial pattern). Moran's I is affected by the spatial scale and the coding scheme style chosen to assign weights to neighbors. We thus quantified spatial autocorrelation across the full range of spatial scales of our data. To do so we computed Moran's I in networks of neighboring caves located within increasing distances, until we reached the maximal extent of our study region. We chose the “W” style, which standardizes rows and thus favors observations with few neighbors, but results did not change substantially when using different styles.
Species richness was modeled using a linear mixed model (LMM) with total species richness as response variable, all meaningful cave attributes as predictors (Table B in S1 File), and the report containing the original data as a random effect. This allowed us to account for variation between the different reports (arising due to different teams, timing, and geographic locations). We then followed a model-selection protocol, based on likelihood ratio tests (henceforth LRT), to find the best model describing species richness. We used the lme4 package  to compare a full model (containing all meaningful predictors) to reduced models where each predictor was eliminated one by one. A predictor was excluded if the reduced model (lacking that particular predictor) was not significantly different from the full model (using LRT, α = 0.05). Predictors were eliminated sequentially, beginning with those showing the highest p-values. Likelihood ratio tests were also used to compare the resulting model to models containing all possible combinations of first-degree interactions between the remaining predictors. We then used the nlme package  and restricted maximum likelihood estimation to test whether including a constant variance function to account for variance heterogeneity improved our model. Finally, we tested for spatial autocorrelation in the residuals of our final model, calculating Moran's I across different spatial scales as described above. The final model was validated by plotting residual vs. fitted values, residual vs. predictors, by looking at the distribution of residuals, and by checking for multicollinearity. To evaluate the relationship between total species richness and the richness of troglobitic species, we included the richness of troglobitic species as an additional predictor of our final model, using a data subset for which information on the richness of troglobitic species was available.
Modeling the presence of troglobites, rare troglobites, and resident bat populations
We first tested for spatial autocorrelation in the presence of troglobites, the presence of rare troglobites, and the presence of resident bat populations. We employed the Join Count Test of the spdep package to assess spatial autocorrelation across the full range of spatial scales of our data. We computed the Single Color Statistic for presence-presence in networks of neighboring caves located within increasing distances, until we reached the maximal extent of our study region. As in the case of species richness we chose the “W” style, but results did not change substantially when using different styles.
We then modeled these variables using generalized linear mixed models (GLMM) with Bernoulli distributed responses (logistic regressions). Presence/absence was thus set as response variable, all meaningful cave attributes as predictors (Table B in S1 File), and report as a random effect. As in the case of species richness we kept report as a random effect in all models to account for variation between the different speleology reports. Likewise, we used the lme4 package to compare full models (containing all meaningful predictors) to reduced models where each predictor was eliminated one by one, using likelihood ratio tests. We compared the resulting models to models containing all possible combinations of first degree interactions between the remaining predictors, and also tested for spatial autocorrelation in the residuals of our final models, calculating Moran's I across different spatial scales as described above. Best models were finally validated by plotting residual vs. fitted values, residual vs. predictors, by looking at the distribution of residuals, and by checking for multicollinearity. To evaluate the effect of species richness on the presence of troglobites, the presence of rare troglobites, and the presence of resident bat populations, we included species richness as an additional predictor of our final models.
Reassessing cave relevance for conservation
In order to quantitatively assess how the current relevance classification scheme ranks caves with different biological characteristics, we scored caves according to the number of biodiversity indicators they contained. These indicators comprised high species richness (higher than the upper quartile of species richness per cave, across all analyzed caves), presence of rare troglobites, presence of troglobites, and presence of resident bat populations. We assigned a value of 1 or zero for the presence or absence of each indicator, and then added all values to obtain a total score, representing the number of biodiversity indicators found in each cave.
While we found significant spatial autocorrelation in species richness across most of the analyzed spatial scales, the presence of rare troglobites, presence of troglobites, and presence of bat populations only showed significant spatial autocorrelation at small and intermediate spatial scales (Fig 2). In all cases spatial autocorrelation decreased with increasing spatial scale, and did not differ from the random expectation when reaching the maximal extent of our study area (Fig 2). No spatial autocorrelation was detected in the residuals of our final models (Fig A in S1 File).
While the solid lines show the value of estimates (Moran's I and Single Color Statistic), the gray area depict 95% confidence intervals. Dashed lines represent expected values under a null model of no spatial autocorrelation.
Cave area and the presence of organic material and water reservoirs were found to be key predictors of cave biodiversity (Table 1). Specifically, species richness increased with cave area, and this increase was more pronounced in caves containing guano (Table 2, Fig 3A). Species richness was also higher in caves containing detritus, roots, and resident bat populations, but lower in caves with water reservoirs than in caves without them (Fig 3B). Larger caves were also more likely to contain troglobitic species, rare troglobites, and resident bat populations (Table 2). However, caves with water reservoirs were less likely to contain rare troglobites. Species richness was positively associated to the presence of troglobites, rare troglobites, and bat populations (Table C in S1 File). Moreover, total species richness was positively associated to the richness of troglobitic species in a data subset for which this information was available (Table C in S1 File).
A: Relationship between species richness and cave area in caves with guano (triangles) and caves without guano (circles). B: Species richness in caves with and without water reservoirs (sample sizes are given in parentheses). Species richness is detrended for the effect of other predictors.
Large caves containing high species richness, were often classified as high relevance caves in the speleology reports (Fig B in S1 File), as were caves with a high total species richness and a high richness of troglobitic species (Fig C in S1 File). All caves of maximal relevance contained troglobites, and all caves containing rare troglobites were classified with maximal relevance (Fig D in S1 File), as stipulated in the current Brazilian legislation for cave protection. When assessing how the current relevance classification scheme ranks caves with different biological characteristics, we found that caves containing zero or one biodiversity indicator are often classified as high relevance caves (Fig 4). Moreover, many caves with two biodiversity indicators were classified as maximal relevance caves, while not all caves with three biodiversity indicators were (Fig 4).
While variation in cave area is represented in the X axis, density plots are shown besides each variable (the vertical dashed line represents the median area of caves containing three or more biodiversity indicators). The relevance categories shown are based on biological attributes only, following the current Brazilian legislation.
Our results reveal spatial autocorrelation in the four cave biodiversity proxies analyzed (species richness, presence of troglobites, presence of rare troglobites, and presence of resident bat populations), and underscore the importance of cave area and the presence of organic material as key predictors of cave biodiversity. Larger caves had higher species richness and were more likely to contain troglobitic species, rare troglobites, and bat populations. Additionally, our data shows that the current relevance classification scheme does not prioritizes the conservation of overall cave biodiversity.
Cave biodiversity is thought to be determined by an island biogeography dynamic , whereby small and isolated caves have higher extinction, lower colonization rates, and fewer available habitats than larger and inter-connected caves [12,24–26]. Indeed, species richness was found to be spatially autocorrelated, suggesting that caves containing many species facilitate the colonization of nearby caves. Previous studies have shown that cave size is a key predictor of cave biodiversity, because larger caves not only have higher colonization rates, but can also host larger and more diverse communities. Analyzing seven caves from Slovenia, Culver et al. 2004  found a higher richness of troglobitic species in larger caves, whereas Brunet and Medellín 2001  found a significant association between the number of bat species and cave area in 20 caves from central Mexico. Studying 55 limestone caves located in the Brazilian Savannah, Simões et al. (2015)  found a positive association between species richness, width of entrances and linear development of the caves. A similar relationship between cave size and species richness was found in a study analyzing 91 caves of different lithologies from Southern Brazil, with iron caves showing the strongest effect . Preliminary studies on iron caves also found higher species richness and higher richness of troglobitic species in larger caves [11,27]. Our results go in line with these previous findings, reaffirming the importance of cave area as a key predictor of species richness, the presence of troglobites, rare troglobites, and bat populations.
Our data also supports the idea that a higher availability of trophic resources facilitates colonization of the cave's interior [11,13,26,28], as species richness was higher in caves with detritus, roots, and resident bat populations, and the relationship between cave area and species richness was found to be stronger in caves containing guano. For instance, the cave's deep interior has been compared to a desert , since it is largely deprived of trophic resources [6,13,30]. The fauna that inhabit the cave's interior thus rely on external material that is washed into the cave or brought in by mobile species [13,28,31,32]. In our case, detritus might have been brought inside the caves by bats, as guano, or may have been washed in by water. Interestingly, our results reveal that roots also seem to constitute an important food resource for the fauna of iron caves. Indeed, cave root communities are known to include many rare and highly specialized species , and they are considered one of the main food resources in iron caves [11,14,34].
The importance of bat colonies for cave macro-invertebrate communities has been well documented. Bats provide nutrients and water resources through guano, respiration and urination, which can support diverse cave arthropod communities [13,35–38]. These resources are especially important in iron caves, which rarely receive additional resources from permanent streams [14,39]. Our findings highlight the importance of bat colonies as pillars of iron cave biodiversity. Additionally, bats are important providers of ecosystem services. As mobile agents  bats contribute to pollination [41,42] and seed dispersal [43–45], thus mediating plant reproductive success and enhancing the restoration of degraded habitats [46,47]. As predators of pest insects, bats provide biological pest control services to many commercial crops, including cocoa , coffee , cotton , and corn . Bats thus have a key ecological importance both inside and outside caves, so conservation efforts to preserve bat populations should not be undermined by the focus on rare troglobitic species.
We found lower species richness in caves containing water reservoirs than in caves without them. This result was unexpected, and contradicts earlier findings showing higher species richness in caves containing permanent water bodies [16,27]. One reason for this apparent discrepancy, could be that the water effect found by these previous studies was driven by the effect of cave size (i.e. caves containing water bodies were usually larger than caves without them). Indeed, the effect of water reservoirs on species richness was not found to be significant when we ran a simple mixed model (containing the presence of water reservoirs as the only predictor), although it was significant in our final model (which contained six other predictors, Table 2). One potential explanation for the observed effect of water reservoirs, is the fact that some can occupy a relatively large area inside the caves, thus reducing the effective cave area. As the speleology reports analyzed here focused exclusively in troglofauna (ignoring stygofauna), large water reservoirs are expected to undermine the effect of cave area on species richness. Alternatively, water bodies that are subject to flood pulses can eventually be detrimental for the establishment of troglobitic species, and thus imply a reduction in species richness .
The occurrence of troglobites and rare troglobites was found to be spatially aggregated (i.e. neighboring caves tended to either have or lack troglobites). Analyzing over 3000 records from more than 450 troglobitic species occurring across the eastern United States, Christman et al. 2005  also found significant spatial autocorrelation in the total number of troglobitic species, the number of non-endemics, and the number and occurrence frequency of single-cave endemics. Our results thus support these earlier findings, and suggests that troglobitic species can move between caves separated by up to 40 Km (Fig 2), although the dispersal mechanisms remain unknown. The presence of water reservoirs was negatively associated to the occurrence of rare troglobites, which suggests water is enhancing connectivity between caves. Indeed, Simões et al. (2015) found that the community of troglophile species was more similar between caves containing streams than between dry caves or caves with puddles. Similarly, Pipan and Culver 2007  identified lateral epikarst connections based on the distribution of copepods, indicating that water in shallow subterranean habitats facilitates dispersal in these arthropods.
The banded ironstone formations known as Canga, where our study caves occur, are constituted by highly porous rocks that form many micro-cavities and cracks [14,34,54]. Preliminary evidence indicates these are potential subterranean habitats that could serve as dispersal corridors to some troglobitic species . Alternatively, these cracks and crevices might actually represent the primary habitat for troglobitic species, with caves serving as convenient sampling sites. For instance, troglobites were also found in small caves (Fig D in S1 File), which were common in our dataset (median cave length = 16.40m; mean cave length = 31.01m; see distribution of cave area in Fig 4). Since small caves may not comprise suitable stable habitats for obligate dwellers of the dark zone, our results suggest that troglobites are using subterranean habitats other than caves. Our findings thus provide additional evidence for the permeable nature of banded ironstone formations, and indicate that iron caves should not be viewed as isolated entities but rather as components of a cave network immersed in the karst landscape [34,52]. Such broader perspective would imply shifting conservation efforts from individual caves to subterranean habitats as a whole.
We propose a simple but effective scheme to establish cave conservation priorities and estimate compensation offsets. Using a set of cave biodiversity indicators, including high species richness (in our case higher than the upper quartile of species richness per cave, across all analyzed caves), presence of troglobitic species, presence of rare troglobites, and presence of resident bat populations, we ranked caves according to the number of indicators they exhibited. We decided to assign an equal weight to all biodiversity indicators, but weights could be modified if particular indicators need to be prioritized. Conservation priorities and compensation offsets can then be established based on these ranks. In our case (Fig 4), caves containing three or four indicators are considered the most important caves for conservation (maximal relevance, 10% of all caves), followed by caves with two (high relevance, 17%), one (mid relevance, 35%), and no indicator (low relevance, 38%).
Our proposed scheme is appealing for three reasons. First, it constitutes a shift from a reactive approach focused on rare troglobites to one that supports a more complete suite of conservation priorities, as stressed out in the precautionary principle , and recommended by recent analyses of current conservation initiatives [4,57]. For instance, preserving ecosystem function is also important, so a cave containing a rare troglobite may be as important for conservation as a cave lacking rare troglobites but containing high species richness, many troglobitic species, and a resident bat population. Second, because it involves few biodiversity indicators, which are already measured during speleology surveys and constitute proxies of cave biodiversity, it implies a more efficient allocation of resources towards the preservation of the most relevant caves. Even though our approach is considerably less complex than some reserve selection algorithms , we believe that simple indicators may be more appropriate when the taxonomic framework (and hence species range information) is so poorly developed. Finally, it represents a better compromise between cave protection and the extraction of mineral resources than the current cave protection regime, by which an average of 70% of the surveyed caves are assigned high relevance (according to our dataset), and must then be compensated by offsets [8,9]. Following our scheme only 17% of caves would need to be compensated by other high relevance caves, a more realistic requirement that should be possible to fulfill.
Our relevance ranking scheme demonstrates that the current Brazilian cave protection regime does not allocate maximal conservation priority to caves exhibiting more biodiversity indicators (Fig 4). The strong emphasis on rare troglobites is leaving ecologically important caves unprotected, while protecting less relevant caves that contain rare troglobites (Fig D in S1 File). This is further aggravated by the fact that the rarity of these troglobitic species is questionable, as most remain unidentified to the species level, and their distribution ranges are poorly known [10,11]. For instance, many troglobites are considered rare in the absence of extensive surveys in the area, although sampling intensification often expands the distribution range of troglobitic species [7,59]. Our work thus suggests that the great emphasis on rare troglobitic species may be a poor allocation of scarce conservation resources, as shown for many other surrogate species . These findings highlight the need to develop more human resources to aid the taxonomic identification of cave fauna, unify databases containing the occurrences of these organisms , and implement the use of molecular bar coding techniques to increase the speed and accuracy of taxonomic identifications .
Our results also show that the most relevant caves for conservation are larger caves (see the vertical dashed line in Fig 4). For instance, larger caves not only exhibit a high overall species richness, but are also more likely to contain more troglobitic species, rare troglobites, and bat populations (Table 2, Table C in S1 File). A similar effect was reported by Christman et al. (2005), who found that more diverse communities usually contain a higher fraction of endemic troglobites. Large caves with high species richness should thus be viewed as key potential reservoirs of endemic troglobitic species, so they should be considered of maximal conservation priority.
Overall, our findings have important implications for the conservation of iron cave biodiversity. First, large caves were found to be important reservoirs of biodiversity, so they should be prioritized in conservation programs. Second, we found that species richness and the presence of troglobites and bat populations were spatially aggregated, implying that there are delimited areas containing groups of caves with high (or low) relevance for conservation. Third, spatial autocorrelation and the presence of troglobitic species in small caves suggest that troglobites can move between caves, and thus that iron caves should be treated as components of a cave network immersed in the karst landscape. This implies that conservation efforts should target subterranean habitats as a whole. Finally, our relevance ranking scheme demonstrates that the current Brazilian cave protection regime does not prioritizes the conservation of overall cave biodiversity. We believe that our proposed approach could help simplify the assessment process and channel more resources to the effective protection of cave biodiversity.
S1 File. This file contains three Tables (A-C) and four figures (A-D).
S1 Dataset. Additional reports for Serra Leste and full taxa inventories for all caves.
We thank all participants of the Seminário Internacional em Ecologia, Mineração e Desenvolvimento Sustentável (November 18-19th 2015, Rio de Janeiro) for the insightful discussions that helped frame this manuscript, Vale's Environmental Licensing and Speleology Department for its engagement in this endeavor, Dr. Pedro Walfir Souza-Filho for facilitating GIS resources, and three anonymous referees for improving earlier versions of this manuscript.
- Conceptualization: RJ XP JOS.
- Data curation: RJ XP RZ.
- Formal analysis: RJ.
- Funding acquisition: JOS.
- Methodology: RJ XP.
- Project administration: IVB JOS.
- Resources: RJ XP RZ TCG CM.
- Visualization: RJ.
- Writing – original draft: RJ.
- Writing – review & editing: RJ XP RZ TCG VLIF CM GO IVB JOS.
- 1. Phalan B, Onial M, Balmford A, Green RE. Reconciling food production and biodiversity conservation: land sharing and land sparing compared. Science. 2011;333(6047):1289–91. pmid:21885781
- 2. Andriamanalina A, Cardiff SG. Contested Spatial Coincidence of Conservation and Mining Efforts in Madagascar. Madagascar Conserv Dev. 2007;2(1):28–34.
- 3. Gardner T a., Von Hase A, Brownlie S, Ekstrom JMM, Pilgrim JD, Savy CE, et al. Biodiversity Offsets and the Challenge of Achieving No Net Loss. Conserv Biol. 2013;27(6):1254–64. pmid:24033441
- 4. Sochi K, Kiesecker J. Optimizing regulatory requirements to aid in the implementation of compensatory mitigation. J Appl Ecol. 2015;53:317–22.
- 5. Harvey MS. Short-range endemism among the Australian fauna: Some examples from non-marine environments. Invertebr Syst. 2002;16(4):555–70.
- 6. Pipan T, Culver D. C. Subterranean Ecosystems. In: Levin SA, editor. Encyclopedia of Biodiversity. 2nd ed. Waltham, MA: Academic Press.; 2013. p. 44–62.
- 7. Harvey MS, Rix MG, Framenau VW, Hamilton ZR, Johnson MS, Teale RJ, et al. Protecting the innocent: Studying short-range endemic taxa enhances conservation outcomes. Invertebr Syst. 2011;25(1):1–10.
- 8. Auler AS, Piló LB. Caves and Mining in Brazil: The Dilemma of Cave Preservation Within a Mining Context. In: Andreo B, Carrasco F, Durán JJ, Jiménez P, LaMoreaux JW, editors. Hydrogeological and Environmental Investigations in Karst Systems SE—55 [Internet]. Springer Berlin Heidelberg; 2014. p. 487–96. (Environmental Earth Sciences; vol. 1). http://dx.doi.org/10.1007/978-3-642-17435-3_55
- 9. Academia Brasileira de Ciências. Seminário Internacional em Ecologia, Mineração e Desenvolvimento Sustentável. In: (ABC) AB de C, (IBRAM) IB de M, (SBE) SB de E, Sociedade Brasileira para o Progresso da Ciência (SBPC), editors. 18–19 de Novembro [Internet]. Rio de Janeiro; 2015. http://www.abc.org.br/article.php3?id_article=4433
- 10. Trajano E, Bichuette ME. Relevância de cavernas: porque estudos ambientais espeleobiológicos não funcionam. Espeleo-Tema. 2010;21(1):105–12.
- 11. Ferreira RL, Oliveira MPA, Silva MS. Biodiversidade subterrânea em geossistemas ferruginosos. In: Geossistemas ferruginosos do Brasil: Áreas prioritárias para conservação da diversiadde geológica e biológica, patrimônio cultural e serviços ambientais. Belo Horizonte, MG: Instituto Prístino; 2015. p. 195–231.
- 12. Culver DC, Christman MC, Šereg I, Trontelj P, Sket B. The location of terrestrial species-rich caves in a cave-rich area. Subterr Biol. 2004;2:27–32.
- 13. White WB, Culver DC. Encyclopedia of caves. Academic Press; 2012.
- 14. Silva MS, Martins RP, Ferreira RL. Cave lithology determining the structure of the invertebrate communities in the Brazilian Atlantic Rain Forest. Biodivers Conserv. 2011;20(8):1713–29.
- 15. Lunghi E, Manenti R, Ficetola GF. Do cave features affect underground habitat exploitation by non-troglobite species? Acta Oecologica [Internet]. 2014 Feb;55:29–35. Available from: http://www.sciencedirect.com/science/article/pii/S1146609X13001677
- 16. Simões MH, Souza-Silva M, Ferreira RL. Cave physical attributes influencing the structure of terrestrial invertebrate communities in Neotropics. Subterr Biol. 2015;16:103.
- 17. Poveromo JJ. Iron Ores. In: The Making, Shaping, and Treating of Steel: Ironmaking volume. Pittsburg, PA: The AISE Steel Foundation; 1999. p. 547–50.
- 18. R Core T. R: A language and environment for statistical computing [Internet]. Vienna, Austria: R Foundation for Statistical Computing; 2015. http://www.r-project.org
- 19. Fortin M-J, Dale MRMRT. Spatial analysis: a guide for ecologists. Cambridge University Press; 2005.
- 20. Bivand R, Piras G. Comparing implementations of estimation methods for spatial econometrics. In American Statistical Association; 2015.
- 21. Bates D, Maechler M, Bolker B, Walker S. lme4: Linear mixed-effects models using Eigen and S4 [Internet]. 1.1–7. 2014. http://cran.r-project.org/package=lme4
- 22. Pinheiro J, Bates D, DebRoy S, Sarkar D. nlme: Linear and Nonlinear Mixed Effects Models [Internet]. 3.1–117. 2014. http://cran.r-project.org/package=nlme
- 23. MacArthur RH, Wilson EO, MacArthur W. The theory of island biogeography [Internet]. Vol. 1, Princeton University Press Princeton USA Major RE Kendal CE. 1967. p. 203. http://books.google.com/books?hl=en&lr=&id=a10cdkywhVgC&pgis=1
- 24. Culver DC. Analysis of Simple Cave Communities I. Caves as Islands. Evolution (N Y) [Internet]. 1970;24(2):463–74. Available from: http://www.jstor.org/stable/2406819%5Cnfiles/1166/Culver-1970—AnalysisofSimpleCaveCommunitiesI.CavesasIs.pdf
- 25. Brunet AK, Medellín RA. The species–area relationship in bat assemblages of tropical caves. J Mammal. 2001;82(4):1114–22.
- 26. Culver DC, Pipan T. The biology of caves and other subterranean habitats. Oxford University Press, USA; 2009.
- 27. Prous X, Vasconcellos M, Bezerra T. Contribuição ao conhecimento da biologia de cavernas em litologias ferríferas de Carajás, PA. In: Anais do XXXI Congresso Brasileiro de Espeleologia [Ponta Grossa]. 2011. p. 515–20.
- 28. Poulson TL, White WB. The Cave Environment. Science (80-) [Internet]. 1969 Sep 5;165(3897):971–81. Available from: http://www.sciencemag.org/content/165/3897/971.short
- 29. Lomolino M, Riddle B, Whittaker R, Brown J. Biogeography. 4th ed. Sinauer Associates, Inc.; 2010. 764 p.
- 30. Campbell JW, Woods M, Ball HL, Pirkle RS, Carey V, Ray CH. Terrestrial macroinvertebrates captured with a baited ramp-pitfall trap from five limestone caves in North Alabama and Georgia (USA) and their association with soil organic matter. J Nat Hist. 2011;45(43–44):2645–59.
- 31. Taylor SJ, Krejca JK, Denight ML. Foraging Range and Habitat Use of Ceuthophilus secretus (Orthoptera: Rhaphidophoridae), a Key Trogloxene in Central Texas Cave Communities. Am Midl Nat. 2005;154(1):97–114.
- 32. Wynne JJ, Voyles KD. Cave-dwelling arthropods and vertebrates of North Rim Grand Canyon, with notes on ecology and management. West North Am Nat. 2013;74(1):1–17.
- 33. Stone FD, Howarth FG, Hoch H, Asche M. Root communities in lava tubes. Encycl Caves. 2012;658–64.
- 34. Ferreira RL. A vida subterrânea nos campos ferruginosos. O carste. 2005;3(17):106–15.
- 35. Pape RB. Biology and ecology of Bat Cave, Grand Canyon National Park, Arizona. J Cave Karst Stud. 2014;76(1):1–13.
- 36. Iskali G, Zhang Y. Guano subsidy and the invertebrate community in Bracken Cave: the world’s largest colony of bats. J Cave Karst Stud. 2015;77(1):28.
- 37. Ferreira RL, Martins RP. Diversity and distribution of spiders associated with bat guano piles in Morrinho cave (Bahia State, Brazil). Divers Distrib. 1998;4(5/6):235–41.
- 38. Ferreira RL, Martins RP. Trophic structure and natural history of bat guano invertebrate communities, with special reference to Brazilian caves. Trop Zool. 1999;12(2):231–52.
- 39. Souza-Silva M, Bernardi LFDO, Martins RP, Ferreira RL. Transport and consumption of organic detritus in a neotropical limestone cave. Acta Carsologica. 2012;1(140):41.
- 40. Fleming TH, Eby P, Kunz TH, Fenton MB. Ecology of bat migration. In: Bat ecology. Chicago: The University of Chicago Press; 2003. p. 156–208.
- 41. Winter Y, von Helversen O. Bats as pollinators: foraging energetics and floral adaptations. In: Cognitive ecology of pollination. Oxford University Press London; 2001. p. 148–70.
- 42. Fleming TH, Geiselman C, Kress WJ. The evolution of bat pollination: A phylogenetic perspective. Ann Bot. 2009;104(6):1017–43. pmid:19789175
- 43. Hougner C, Colding J, Söderqvist T. Economic valuation of a seed dispersal service in the Stockholm National Urban Park, Sweden. Ecol Econ [Internet]. 2006;59(3):364–74. Available from: http://www.sciencedirect.com/science/article/pii/S0921800905005124
- 44. Jacomassa F a F, Pizo M a. Birds and bats diverge in the qualitative and quantitative components of seed dispersal of a pioneer tree. Acta Oecologica [Internet]. 2010;36(5):493–6. Available from: http://dx.doi.org/10.1016/j.actao.2010.07.001
- 45. Sarmento R, Alves-Costa CP, Ayub A, Mello MAR. Partitioning of seed dispersal services between birds and bats in a fragment of the Brazilian Atlantic Forest. Zool. 2014;31(3):245–55.
- 46. Wunderle JM. The role of animal seed dispersal in accelerating native forest regeneration on degraded tropical lands. For Ecol Manage. 1997;99(1–2):223–35.
- 47. Silveira M, Trevelin L, Port-Carvalho M, Godoi S, Mandetta EN, Cruz-Neto AP. Frugivory by phyllostomid bats (Mammalia: Chiroptera) in a restored area in Southeast Brazil. Acta Oecologica [Internet]. 2011;37(1):31–6. Available from: http://dx.doi.org/10.1016/j.actao.2010.11.003
- 48. Maas B, Clough Y, Tscharntke T. Bats and birds increase crop yield in tropical agroforestry landscapes. Ecol Lett. 2013;16(12):1480–7. pmid:24131776
- 49. Karp DS, Daily GC. Cascading effects of insectivorous birds and bats in tropical coffee plantations. Ecology. 2014;95(4):1065–74. pmid:24933824
- 50. Cleveland CJ, Betke M, Federico P, Frank JD, Hallam TG, Horn J, et al. Economic value of the pest control service provided by Brazilian free-tailed bats in south-central Texas. Front Ecol Environ. 2006;4(5):238–43.
- 51. Maine JJ, Boyles JG. Bats initiate vital agroecological interactions in corn. Proc Natl Acad Sci. 2015;112(40):12438–43. pmid:26371304
- 52. Christman MC, Culver DC, Madden MK, White D. Patterns of endemism of the eastern North American cave fauna. J Biogeogr. 2005;32(8):1441–52.
- 53. Pipan T, Culver DC. Copepod distribution as an indicator of epikarst system connectivity. Hydrogeol J. 2007;15(4):817–22.
- 54. Auler A, Piló L, Parker C, Senko J, Sasowsky I, Barton H. Hypogene cave patterns in iron ore caves: Convergence of forms or processes? In: Klimchouk A, Sasowsky ID, Mylroie J, Engel SA, Engel AS, editors. Hypogene Cave Morphologies. Leesburg, Virginia: Karst Water Institute; 2014. p. 15–9.
- 55. Soares G, Andrade R, Perroni G. Troglomorphic fauna sampling methods in canga formations, Minas Gerais State, Brazil. In: 23rd International Conference on Subterranean Biology. Fayetteville, AK; 2016. p. 96.
- 56. Kriebel D, Tickner J, Epstein P, Lemons J, Levins R, Loechler EL, et al. The precautionary principle in environmental science. Environ Health Perspect [Internet]. 2001;109(9):871–6. Available from: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1240435/pdf/ehp0109-000871.pdf
- 57. Asaad I, Lundquist CJ, Erdmann M V, Costello MJ. Ecological criteria to identify areas for biodiversity conservation. Biol Conserv [Internet]. (online). http://www.sciencedirect.com/science/article/pii/S000632071630533X
- 58. Csuti B, Polasky S, Williams PH, Pressey RL, Camm JD, Kershaw M, et al. A comparison of reserve selection algorithms using data on terrestrial vertebrates in Oregon. Biol Conserv. 1997;80(1):83–97.
- 59. Ledford J, Paquin P, Cokendolpher J, Campbell J, Griswold C. Systematics, conservation and morphology of the spider genus Tayshaneta (Araneae, Leptonetidae) in Central Texas Caves. Zookeys [Internet]. 2012;167(167):1–102. Available from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3272638&tool=pmcentrez&rendertype=abstract%5Cnhttp://zookeys.pensoft.net/articles.php?id=2420
- 60. Andelman SJ, Fagan WF. Umbrellas and flagships: efficient conservation surrogates or expensive mistakes? Proc Natl Acad Sci U S A. 2000;97(11):5954–9. pmid:10811901
- 61. Juan C, Guzik MT, Jaume D, Cooper SJB. Evolution in caves: Darwin’s “wrecks of ancient life” in the molecular era. Vol. 19, Molecular Ecology. 2010. p. 3865–80.