Extending beyond individual caves: a graph theory approach broadening conservation priorities in Amazon iron ore caves

Introduction

The Amazon region is renowned for its extraordinary biological diversity (Oliveira et al., 2017), sheltering the largest tropical rainforest in the world (Cardoso et al., 2017). In addition to its biological importance, this region harbors significant mineral reserves, including open-pit mines for iron, nickel, copper, and gold extraction, which generate 10% of Brazil’s exports (Piló, Auler & Martins, 2015; Agência Nacional de Mineração–ANM, 2018). Notably, the Amazonian ferruginous geosystems contain over 1,000 recorded caves, despite their typically small size (approx. 30 m long) (Piló, Auler & Martins, 2015). These caves host highly diverse invertebrate communities, including many cave-restricted species (troglobites) (Jaffé et al., 2016, 2018; Ferreira, Oliveira & Silva, 2018). The relative shallowness of these caves, coupled with an intricate network of interconnected void spaces within the rock matrix, enables nutrient exchange and migration of species between surface and subterranean habitats (Ferreira, Oliveira & Silva, 2018). Despite their ecological significance, the preservation of these habitats is often challenged by mineral exploitation. The preservation of a single cave in this region can represent the withholding of around US $38 billion in commodity, given the ferruginous matrix in which they are embedded (Auler & Piló, 2015).

Conflicts between cave conservation and land use are not restricted to the Amazon region. Karst areas are widely distributed around the world and have been affected by urban and agricultural expansion in almost all the continents. In addition, the increasing demand for minerals, many of which are associated with caves, poses a threat to these habitats (Ferreira, Oliveira & Silva, 2015; LaMoreaux, Powell & LeGrand, 1997; Ferreira et al., 2022). Therefore, many countries established specific laws to mediate such conflicts, such as USA (National Historic Preservation Act (NHPA), 1966; Archaeological Resources Protection Act-ARPA, 1979; Native American Graves Protection & Repatriation Act-NAGPRA, 1990), China (Zhifang, 1991; People’s Republic of China, 1989), Australia (Environmental Protection Authority-EPA, 2016, 2007), Slovenia (The Republic of Slovenia, 2003), Slovakia (The Republic of Slovakia, 2002) and Brazil (Federal Republic of Brazil, 2008; Auler & Piló, 2015). However, for all of these countries, current conservation decisions regarding caves often consider only their intrinsic attributes, while ignoring the importance of adjacent landscapes and interactions among different caves and other karst compartments (Christman et al., 2016; Jaffé et al., 2018; Mammola et al., 2020). This impasse raises questions about the most effective conservation strategies: whether to preserve individual caves or groupings of interconnected caves.

The discussion on nature reserve design began in the 1970s, focusing on whether a single large or several small areas should be used for species conservation purposes (referred to as SLOSS) (Simberloff & Abele, 1976; Ovaskainen, 2002; Groenveld, 2005; Tjørve, 2010). This debate has guided conservation proposals in various ecosystems, including tropical forests (Laurance & Bierregaard, 1997), marine environments (McNeill & Fairweather, 1993), and even urban areas (Valente, Pasimeni & Petrosillo, 2020), as well as for different groups of species, such as butterflies (Baz & Garcia-Boyero, 1996) and birds (Lindenmayer et al., 2015). However, this approach has not yet been applied to subterranean fauna. As a result, important caves are often isolated in fragmented landscapes, limiting the dispersal of species, including troglobitic species (Jaffé et al., 2018; Hoch & Ferreira, 2012). Therefore, how can we define conservation targets considering a scale larger than the cave itself? One possibility is to use biological parameters of cave communities associated with landscape and habitat information, such as similarity matrices, to support an assessment among caves rather than only for each cave individually.

In this study, we aimed to identify the predictors that influence the similarity among cave communities, proposing the use of this analysis as a proxy to preserve the subterranean fauna associated with Amazon ferruginous geosystems. Specifically, we sought to answer the following questions: (a) Does the similarity between cave communities differ when considering the entire community compared to only troglobitic species? (b) To what extent can we detect such similarities? (c) Which factors are more influential in shaping these patterns: geographical proximity, habitat similarity, or landscape traits? (d) How can we leverage similarity to identify connectivity between caves, and in this context, (e) what are the most suitable conservation strategies? As hypotheses, we expect to find differences in the similarity between caves depending on the species considered in the analysis (troglobitic vs non-troglobitic), due to differences in the degree of endemism for each of these groups. We expect that the proximity between caves will outweigh habitat and landscape traits in determining similarity, considering the increased likelihood of species migration between nearby communities. We anticipate identifying clusters of nearby caves with similar troglobitic species due to their subterranean interconnectivity. Finally, we predict that strategies focused on the conservation of cave clusters will be prioritized over individual preservation efforts.

Materials and Methods

Study site

We carried out this study in 69 iron-ore caves located in Serra do Tarzan (6°19′32″S, 50°06′36″W), eastern region of the Amazon biome, Pará state, Brazil (Fig. 1). Serra do Tarzan is situated within the Campos Ferruginosos National Park, which is part of a network of protected areas known as the Carajas Mosaic, surrounding the mineral extraction province (Carajás Mineral Province). We selected this study area because it represents the closest unit to the pristine condition of the Amazon ferruginous geosystem. Geologically, this region is characterized by banded iron formation (BIF), represented by jaspilites exhibiting intercalations of iron oxide and silica. The hills are covered with a thick ferruginous lateritic layer, known as canga, where caves develop, preferably at the edges of plateau fractures (Piló, Auler & Martins, 2015). The climate is characterized by an average temperature ranging from 23 °C to 25 °C and an annual rainfall of approximately 2,400 mm (Piló, Auler & Martins, 2015). However, the distribution of rainfall is irregular, with the rainy season occurring between October and April (with a range of 160 to 340 mm per month), and a dry period spanning from May to September (with a range of 10 to 90 mm per month) (Sahoo et al., 2016).

Results

Spatial autocorrelation for cave communities

The species composition and the presence of phylogenetically related taxa exhibited notable similarities among cave communities located within a radius of 700 m from a given cave (see Figs. 2A and 2B). When considering the cave-restricted fauna, we observed a significant influence on species composition for caves located up to 1,243 m apart (see Fig. 2C). Furthermore, the presence of troglobitic species in a particular community promoted the occurrence of phylogenetically related taxa in neighboring caves situated within a distance of up to 745 m (see Fig. 2D).

Thresholds for cave community similarity

We identified specific ranges of values that contribute to the similarity between pairs of caves (see Fig. 3), taking into account the significant predictors identified in the models (see Table 1). For the overall fauna composition, caves exhibited greater similarity when their relative humidity values were within a range of ±26% RH, escarpment heights were similar (±11 m), water features had a similar seasonal pattern, caves were surrounded by forests within a range of ±10 ha, and the average epigean temperature was approximately ±0.1 °C (see Fig. 3A). In terms of phylogenetic distance, we observed greater similarity between communities in caves that were geographically closer (±618 m), had a similar area (±191 m²), and exhibited a minimum humidity level within a range of ±18% RH (see Fig. 3B).

Regarding the troglobitic species composition, we observed similarities between caves that are located until 5 km apart, inserted in escarpments with slight differences in height (±3.5 m) and that present relative air humidity with similar values (MaxH ±4% RH) (Fig. 3C). On the other hand, caves hosting communities with phylogenetically related troglobites displayed similarities in area (±185 m²), relative humidity (MaxH ±6% RH) and guano accumulation (±0.25 m²). Furthermore, these caves shared similar water features, and the predominant sediment granulometry differed by at most one level (e.g., if one cave had a predominance of sand, the others would have a predominance of granules or silt) (Fig. 3D). Additionally, regarding the surrounding landscape, phylogenetically similar communities were observed in caves located at similar altimetric levels (±70 m), cliffs (±8.5 m), and experiencing similar epigean average temperatures (±0.6 °C) (Fig. 3D).

Cave connections and conservation priorities

We found that the pairwise connectivity (CPC) was higher when considering the species composition compared to the phylogenetic distance (see Fig. 4). Additionally, the graph generated from spatial autocorrelation exhibited higher CPC values, in contrast to what was observed for the phylogenetic distance graph (see Figs. 4A and 4B).

All the studied caves are connected when considering both composition and phylogeny traits (connections of Weight 2) (Fig. S4). We found 27 caves with at least one connection that encompasses all the similarity metrics used (Weight 4) (Fig. S4). Based on the most parsimonious graph, we observed that 21% of all potential connections among caves in Serra do Tarzan were present (see interactive 3D graph in Fig. S5). The cluster analysis revealed the presence of six groups of caves that are densely interconnected (Modularity = 0.472) (Fig. 5). Regarding the intermediation centrality, eight caves acted as bridges between different regions in the study area: ST_0038, ST_0039B, ST_0014, ST_0027, ST_0018, ST_0063, ST_0039A and ST_ 0012 (see Table S6). Finally, we indicate 16 caves as priorities for conservation since they present a singularity value higher than one (Fig. 6 and Table S6). These caves are highly influenced by 23 other distinct caves, which we recommend as secondary conservation priorities (Fig. 6).

Discussion

Our findings provide support for the hypothesis that the similarity among caves varies depending on whether all species or only troglobitic ones are considered. Interestingly, the spatial range of similarity among caves based on troglobitic species was larger than expected. Surprisingly, we found that geographical distance does not play the most crucial role in determining patterns of similarity among caves. Instead, landscape and habitat characteristics emerged as more influential factors, as evidenced by their significant predictors across all similarity measures examined. Our predictions of similar troglobitic communities within clusters of nearby caves were confirmed, underscoring the existence of subterranean connectivity between them. In terms of conservation, our hypothesis was partially supported: it is important to prioritize cave clusters for conservation, but these clusters should consist of both unique caves and those capable of influencing biological communities. We believe that this approach strikes a better balance between cluster conservation and individualized recommendations.

We identified variations in similarity metrics (species composition and phylogenetic distances) when considering the entire cave community vs only troglobitic species. Caves serve as ecotones bridging surface and subterranean ecosystems (Moseley, 2009), providing refuge for species from different landscape compartments (de Fraga et al., 2023). As a result, pairs of caves can exhibit either similarities or distinctiveness depending on the specific set of species considered. Interestingly, we found that the taxonomic groups with the highest diversity in the overall cave community were not well represented among troglobites. Instead, it was the troglobitic species that contributed most significantly to the similarities observed among caves. This implies that cave communities diverge as we include species associated with other compartments, such as surface ecosystems, many of which are locally restricted and unrelated to troglobites.

We also observed a narrower spatial range for similarity measures when considering the entire community species. The ferruginous geosystems encompass a diverse range of landscapes, ranging from savannahs to forests (Mota et al., 2015). Therefore, non-troglobitic populations found within the caves are distributed based on the local conditions and resources available in both the cave and surface ecosystems. Consequently, the similarity between distant cave communities is influenced by euryecic species that can adapt to various conditions or by the level of homogeneity in the external landscape components. In the latter case, the troglobitic components of the communities achieve a broader range as the subterranean habitats they inhabit attenuate variations in external environmental conditions. Likewise, the narrower spatial range of phylogenetic similarity, in comparison to species composition, reflects the distinct tolerance capabilities of each taxonomic group in adapting to environmental conditions, thereby influencing their dispersion patterns according to specific characteristics and limitations.

Jaffé et al. (2018) emphasized the role of geographic distance as the primary factor influencing community similarity in Amazon iron ore caves. However, we underscore the significance of landscape and habitat characteristics, which can also influence similarity measures and may overlap with the importance of geographic distance. The divergent results obtained in our study compared to the aforementioned study appear to be associated with the predictor characteristics employed. While Jaffé et al. (2018) considered physical and trophic characteristics of caves, their landscape data was obtained at a larger scale. In contrast, our study utilized a wide range of fine-scale predictors, providing a higher resolution of conditions and resources both within the caves and in the surrounding landscapes. Moreover, our research was conducted in a pristine environment, whereas the previous study examined caves located near areas of mineral exploitation and livestock (Jaffé et al., 2018). It is worth noting that the authors of the previous study suggested that anthropogenic land use did not necessarily act as a barrier to subterranean connectivity (Jaffé et al., 2018). However, such changes might have fragmented the landscape, limiting species dispersal to groups of closely located caves, thus amplifying the importance of geographic distance in determining community similarity.

We observed that landscape characteristics had a predominant influence on species composition, whereas habitat attributes emerged as determinants of phylogenetic similarity. This suggests that the landscape plays a role in selecting the species occurring across the entire study area, while only groups already pre-adapted to subterranean conditions thrive within the caves. Howarth & Hoch (2012) reported a higher likelihood of subterranean colonization for taxa whose preferred habitats correspond to those found in caves.

Regarding troglobitic species, there are physical (e.g., area, granulometry), climatic (e.g., water features, humidity), and trophic (e.g., guano) constraints that influence the phylogenetic composition similarity between cave pairs, but there are no spatial limitations. This indicates that for many taxa, the ferruginous matrix itself comprises the preferred habitat, and these species are found in caves only when specific favorable conditions are present. Studies conducted in ferruginous areas of Australia have demonstrated a high diversity of taxonomic groups in shallow subterranean habitats (Eberhard et al., 2009; Guzik et al., 2011; Halse & Pearson, 2014; Halse, 2018). In Brazil, the majority of the known fauna in ferruginous ecosystems has been found in caves, as most studies have focused on these habitats (Ferreira, Oliveira & Silva, 2018). Therefore, expanding research to consider different compartments of the ferruginous landscape may reveal new patterns of phylogenetic similarity between different regions within such systems.

Troglobitic species exhibit lower tolerance to variations in habitat and landscape traits compared to non-troglobitic species. This reaffirms the notion of the extensive distribution of troglobites within the ferruginous matrix (Ferreira, Oliveira & Silva, 2018; Jaffé et al., 2016, 2018). Many troglobitic species are exclusive to habitats with specific environmental conditions. Moreover, it highlights the vulnerability of troglobitic populations to even slight changes in certain environmental variables, which can lead to their local extinction. We emphasize once again the role of anthropogenic activities, such as mining and livestock farming, in selecting which troglobitic species persist in the landscape. The elimination of certain taxa enhances the similarity among cave communities. Consequently, the high similarity observed among pairs of caves in disturbed landscapes (as indicated by Jaffé et al., 2018) may be attributed to the homogenization of the remaining fauna, favoring the more resilient troglobitic species and artificially increasing connectivity among cave communities.

We observed distinct values for the similarity measures, wherein geographic distance emerged as a significant predictor when considering spatial autocorrelation or segmented regression. Through the application of the Moran’s I test, we were able to assess the degree to which the similarity measure in a particular cave affects the occurrence of similar values in neighboring habitats. The breakpoint obtained from the segmented regression provided insight into the distance at which the relationship with the similarity between cave pairs becomes statistically insignificant. Hence, we propose the combined use of these two approaches as complementary measures for landscape zoning. The intersection area resulting from the application of both distance metrics to a given cave indicates a higher level of influence among cave communities. Non-congruent regions that still encompass one of the distance metrics represent areas with relatively lower potential to influence the inter-cave fauna interactions.

Although several studies have discussed the connectivity among ferruginous caves (Ferreira, 2005; Souza-Silva, Martins & Ferreira, 2011; Jaffé et al., 2016, 2018; Ferreira, Oliveira & Silva, 2018), this study presents the first empirical model that elucidates such connectivity. Our observations reveal that this connectivity is primarily determined by the specific identity of individual taxa rather than the broader taxonomic group to which they belong. The graphs obtained for the different similarity measures and approaches indicate that widely distributed species exhibit phylogenetic relatedness, whereas spatially restricted species belong to distinct taxonomic groups. Consequently, the greater phylogenetic diversity is confined to nearby caves that possess unique combinations of habitat characteristics, which are also influenced by the limited dispersal capability of certain species.

Despite the observed connectivity among all caves in the study area, we identified specific clusters of caves that exhibited a higher degree of interconnection. This finding raises important considerations regarding current environmental legislation, particularly in countries like Brazil (Auler & Piló, 2015), where caves are evaluated and managed individually. By solely focusing on habitat attributes during conservation prioritization, there is a risk of promoting landscape fragmentation, leading to the gradual isolation of cave communities. Under such circumstances, arthropod communities are likely to experience a reduction in diversity (α, β, ψ) and altered spatial-temporal dynamics, ultimately impacting ecosystem functioning (Bestion et al., 2019). To ensure the preservation of interactions among cave communities and the sustainable utilization of speleological heritage in the face of mineral and/or agricultural exploitation, it is crucial to adopt a holistic approach that considers the connectivity and interdependence of cave systems.

Even within clusters that exhibit high connectivity based on similarity, certain caves stand out for harboring an exclusive fauna and/or displaying specific habitat characteristics that are unique to the study area. Our observations indicate that these exceptional caves possess habitat and landscape features that promote subterranean diversity, such as large dimensions, high humidity, and location within forested areas. Importantly, these caves are often geographically distant from one another. This highlights the fact that individual “supercaves” have the potential to accommodate troglobitic species in a given region of the study area at comparable or even higher levels of diversity compared to groups composed of multiple cavities. These findings suggest that a mixed conservation approach is necessary for the sustainable management of ferruginous geosystems, incorporating both individual prioritization of “supercaves” and collective prioritization of other caves that exert significant influence over singular caves. This concept aligns with the idea proposed by Ovaskainen (2002), which suggests that a large number of small patches may be optimal for long-term species persistence, as species ranges increase with the number of patches. Therefore, the most effective strategy for preserving subterranean communities in iron ore landscapes would involve the creation of a network of scattered reserves, as observed in other ecosystems (e.g., Baz & Garcia-Boyero, 1996; Mo et al., 2019; Liu, Li & Lv, 2019).

We also observed that caves with limited representation in terms of habitat and landscape characteristics, as well as a low number of troglobitic species, can serve as connectors between groups of caves. However, the question arises as to whether these caves should be considered for conservation within a framework of sustainable use. Contrary to what may be expected, this situation does not align with the concept of “stepping stones” (Moilanen, 2011), as the species are dispersed throughout the entire ferruginous matrix. Instead, these caves exert simultaneous influence on communities from different groups. Therefore, it is crucial to consider their conservation value when there are unique caves among the communities they influence.

Conclusion

In conclusion, it is crucial to develop protocols for the protection of subterranean biodiversity that encompass not only the caves themselves but also the surrounding landscape, with a particular focus on troglobitic species. Our graph-based approach enables environmental management at both group and individual cave levels. By applying this approach, we have elucidated the connectivity patterns, identified distinct groups of caves, highlighted the importance of preserving singular or unique caves, and identified caves that exert influence on these unique habitats. This methodological advancement is essential for the formulation of effective conservation policies aimed at safeguarding speleological heritage in areas facing ongoing economic pressures (Ferreira et al., 2022).

Supplemental Information