Patterns of biodiversity respond to habitat disturbances and different land-uses. Those patterns possibly vary according to the spatial scale under analysis. Although other studies have shown such responses for different systems, no study has ever demonstrated spatial-scale influences in subterranean terrestrial communities. Therefore, the objective of this paper was to analyze how land use and cave physical structure could influence the terrestrial cave invertebrate species composition. We also determined the influence of different spatial scale on the structure of invertebrate cave composition. We collected environmental data at local scale (e.g. cave size, substrate and environmental stability). For spatial scale we determined land uses at three different landscape scales; we gathered these data into circular areas of different sizes (50, 100 and 250 meters) with centroids in the cave entrances. We finally performed three Distance Based Linear Modeling analyses to test for differences among the predictability of environmental variables when comparing different spatial scales. The best explanatory variable for cave invertebrate similarities was the percentage of covering of the external environment by limestone outcrops. We confirm the scale-dependence hypothesis through the different patterns showed among distinct buffer areas. Models become more precise when larger scales were analyzed to explain cave invertebrate composition. This suggests that larger scales capture important environmental features that explain the cave fauna similarities more precisely. Additionally, we found a strong influence of limestone outcrops at all landscape scale structuring cave communities.

Subterranean, habitat heterogeneity, land-use, limestone outcrop, native vegetation, cattle pasture

Environmental heterogeneity in natural landscapes has been historically replaced by anthropogenic mosaics around the world. As a consequence, several hypothesis describing how landscape characteristics affect biodiversity patterns have been proposed (Tscharntke et al. 2012). Ecological functions and processes are dependent on larger spatial scales than a single habitat patch (Gustafson 1998, Steffan-Dewenter et al. 2002). Biodiversity is often positively correlated with the amount of available habitat (Fahrig 2003), but the effect of land-use changes on biodiversity depends on the landscape context (MacDonald et al. 2000) and on the spatial scale that has been analyzed (McGlinn and Hurlbert 2012, Dumbrell et al. 2008, Zimmermann et al. 2010, Morueta-Holme et al. 2013). Furthermore, ecological communities are structured under processes that act on the landscape, in which both regional and local scales are important factors (Harrison and Cornell 2008).

In the context of landscape influences on biodiversity distribution patterns, caves are good models since they represent simplified and fragile ecosystems (Culver 1982, Culver and Pipan 2009). Since the cave communities are dependent on the allochthonous input of nutrients, alterations in the availability, properties and abundance of these nutrients in the landscape surrounding the caves may affect cave biodiversity. Despite their fragility, caves are under several anthropogenic pressures, and only few studies evaluated how such human activities can affect the invertebrate cave communities, such as inadequate tourism (e.g. Poulson et al. 1995, Moldovan et al. 2003, Pellegrini and Ferreira 2012). Such studies are even scarcer when considering human impact at landscape scale, such as agriculture, urban development, deforestation and mineral resources extraction (Eme et al. 2014, Zagmajster et al. 2014). All these activities lead to aquifer pollution, cave destruction, and biodiversity loss (Beynen et al. 2012). Considering this, the current Brazilian legislation (Brazil. Decree n^o 6.640/2008) imposes that caves should bear a protection area of 250 meters in radius surrounding the cave linear projection on the surface (Portaria IBAMA n^o 887/1990). Although other studies have shown responses for different spatial scales under analysis (e.g. Steffan-Dewenter et al. 2002), few studies have demonstrated such influences in subterranean communities patterns (Eme et al. 2014, Zagmajster et al. 2014).

The main goal of this paper is to explain cave invertebrate composition through environmental variables from within the cave and also from the landscape surrounding the caves at different spatial scales. To that end, we tested the hypothesis that the spatial scale affects the predictability of environmental variables. We also checked for grouping patterns among cave fauna according to the most explanatory variables.

We found 186 invertebrate species, distributed along at least 78 families and 23 orders (Table 1). Among them, the order Diptera was the most representative, presenting 40 species, which belong to at least 17 families. The order Araneae, with 14 families and 33 species was the second richest. We found three troglomorphic species, two belonging to the order Collembola (Entomobryomorpha and Hypogastruroidea) and one Isopoda (Platyarthridae, Trichorhina sp.). Entomobryomorpha was found in two caves, while Hypogastruroidea was restricted to a single cave. Trichorhina sp. was more broadly distributed, being found in four caves.

We found no correlation between geographical distance and caves invertebrate similarity (Correlation R _{MANTEL TEST} = -0.3366; p = 0.9111).

Lapa das Pacas Cave had the highest Environmental Stability Index (ESI=3.53). Ninho de Pérolas Cave presented the smallest value, ESI = -0.56 (Table 2). The habitat varied highly among caves. Only two caves showed bat guano, Gruta do Sumidouro and Lapa da Várzea. Water bodies were also found in only two caves, Gruta do Sumidouro and Lapa das Pacas. Gruta do Sumidouro Cave presented 12 different types of habitats and the Shannon diversity index for this parameter was 2.25, the highest value among the caves analyzed. The lowest Shannon diversity value was found at the Lapa da Várzea Cave (1.30) with only 6 different types of habitats (Table 3).

At landscape scale we found four main types of land cover: limestone outcrop, cattle pasture, native vegetation and others. At all buffers the two principal land covers were pasture and native vegetation. The only buffer with water was the 250 m buffer of the Gruta do Sumidouro Cave (Table 4).

Limestone outcrop was the most important predictor variable of community composition (Jaccard index - considering species identity in the community) in all buffer scales, although other variables varied with landscape scale. The best model for the 50 m Buffer presented an adjusted R² value of 0.40, and used only two variables (limestone outcrop and others). The best model solution for Buffer 100 m presented an adjusted R² value of 0.45, and revealed three variables, limestone outcrop, cattle pasture and native vegetation. Finally, the best model for Buffer 250 m showed an adjusted R² value of 0.73 and used four variables (limestone outcrop, water, environmental stability index and substrate Shannon`s diversity index) (Table 5). Therefore, large-scale models have higher R² values.

There are few studies on spatial patterns of cave communities although some studies evaluate differences of spatial scale sampling on species patterns (e.g. Zagmajster et al. 2008, Eme et al. 2014). Few studies evaluated spatial-scale dependence on community-land uses relationships for cave invertebrates (e.g. Bento 2011). These studies are important to reliably identify essential features and patterns for conservation and management actions. Our findings confirm the scale-dependence hypothesis on explaining similarities among cave communities. Models get more precise at larger scales, it is possible to incorporate new explanatory variables that may be absent at smaller scales. The combination of different scales variables explain better cave community composition.

Karst areas have different historical land uses and human impacts vary according to landscape characteristics (Frumkin 1999). While karst depressions are more easily cultivated, rocky karst slopes and carbonate outcrops are less suitable for such uses (Frumkin 1999). At PESU we could clearly observe this pattern, since impacted areas, with cattle pasture, are mainly those located in depressed areas. Additionally, the best-preserved areas were those located on limestone outcrops or in their surroundings and they were the best predictor of cave community structure for all landscape scales. Such areas also host denser vegetation ensuring better conditions for cave invertebrate communities, most identified species in this study were troglophiles, also present in the surface habitats.

The second factor that explained the cave similarity in the 50 m buffer was “others”, represented by cities, human constructions, roads and bare land, as results of urbanization. According to McIntyre and Hobbs (1999), urbanization is one of the most destructive human activities generating habitat change and loss of ecological function. As urbanization may cause a strong reduction in invertebrate diversity (Buczkowski and Richmond 2012), this effect could influence cave communities in two ways: i) we suggest that caves could be a refuge for invertebrate species, especially for those caves near such areas or ii) urbanization could reduce both, epigean and hypogean invertebrate communities. Here, smaller scales indicated an intensified urbanization impact near cave entrances (Figure 3), thus suggesting that caves may offer shelter to invertebrate species, offering optimal conditions for many invertebrate species, especially for edaphic species that establish a continuum of life from the surface soil to the deep subterranean environment (Gers 1998, Ortuño et al. 2013).

Figure 3. 

Detailed figure of Gruta Helictites showing different land uses within the 50, 100 and 250 m buffers.

The 100 m buffer indicated, in addition to limestone outcrops, cattle pasture and native vegetation as important variables. The landscape cover determines food availability inside caves, which possibly affects cave communities. Although one would expect to have richer communities in caves surrounded by forests (as a rich source of organic matter that can be brought inside caves), guano also constitutes an important resource for many cave invertebrates. Some species are highly dependent on guano, and cave communities associated to this resource can be relatively complex (Ferreira and Martins 1999, Ferreira et al. 2007). The conversion of forests to pastures may favor some species (Gillieson and Thurgate 1999), aside from also reducing foraging habitats for several bat species. Habitat loss and fragmentation seems to favor Desmodus rotundus, a hematophagous bat species that remains in areas transformed into rural landscapes (Aguiar et al. 2010). The change in the organic resource quality can result in a remarkable invertebrate substitution (Souza-Silva et al. 2011), which, in turn, could result in several changes in patterns and processes inherent to the cave fauna. Such changes could eventually enhance the cave community instability, which can become more vulnerable to environmental disturbances.

Considering the 250 m buffers, the model incorporated three different explanatory variables aside from the limestone outcrops: water bodies, environmental stability index and habitat heterogeneity. The importance of the allochthonous nutrient input through water transport is well known (e.g. Hawes 1939, Culver 1982, Romero 2009). Water acts as a molding agent and a vehicle in which nutrients, gases, minerals and even microorganisms are transported underground (Culver and Pipan 2009). The only cave that presented epigean water in the 250 m buffer was the Gruta do Sumidouro. That water is a lake connected to the cave by a sink and its invertebrate fauna is unique because flood pulses continually disturb and remove the accumulated food resource (Souza-Silva et al. 2011). In addition, floods may help maintain a regular food supply to caves, thus operating as distribution agents (Hawes 1939), contributing to the faunal singularity.

The higher community similarity among caves with similar values of ESI was expected. Stable associations by community in some cave sectors, which exhibit optimal climatic conditions, were reported for some invertebrate species (Di Russo et al. 1997). In this context, Bento et al. (in press) conducted a work in 24 cavities in the Brazilian Caatinga (semi-arid landscape) and found that more stable caves showed less variation in the invertebrate community composition when comparing their communities in both seasons (rainy and dry) than less stable caves. In this paper cave stability was calculated by the same index used in this study. Considering these, the more stable the cave environment the more similar the faunal elements at PESU, favoring some species over others (Tobin et al. 2013); especially species with high specialization for cave life, or edaphic spaces.

The habitat diversity hypothesis proposes that species diversity in a landscape will increase as the greater structural complexity increases, because of the higher resource abundance and the potential addition to the number of partitionable niche dimensions (MacArthur and MacArthur 1961). Environmental heterogeneity is correlated to patterns of groundwater crustacean richness (Eme et al. 2014), although it is not correlated to distribution patterns of this group (Zagmajster et al. 2014). Furthermore, cracks and stones, which increase environmental heterogeneity, may provide shelter for small invertebrates in caves (Carchini et al.1982). We found that habitat heterogeneity is an important factor explaining community composition.

Although the landscape scale explains better species composition, the local scale model suggests an influence of the habitat heterogeneity and stability on cave community. The variables ESI and habitat heterogeneity were important only in Buffer 250 m, the largest evaluated scale. In smaller buffers, such local variables were not important probably because other landscape variables had already explained the community variation. In that case, including ESI and habitat heterogeneity in smaller buffers would not increase the model explanation. It has long been known that cave ecosystems are highly vulnerable to external events, even those occurring at some distance from the cave (Gillieson and Thurgate 1999). This indicates the importance of studies on larger areas, which could avoid erroneous conclusions on the real influence of each epigean environmental variable on the cave communities. It is worthy to mention that in situations of geographically closer caves, as Ninho de Pérolas Cave and Gruta do Grilão Cave, the results found could be at least in part due to the 250 meters buffer superposition. However, the distance between caves was not related to caves invertebrate similarity.

Landscape use can even make terrestrial troglophile populations more isolated, severely reducing their dispersal possibilities, by conversion from forest to pasture (Eberhard et al. 1991). Hence, results of community influences by landscape structure (predominance of natural vegetation or disturbed spaces) clearly indicate the need to conserve the adjacent landscapes of caves in karstic zones. Furthermore, attention should be given not only for intact landscapes, but also to continuous forest corridors between caves, that promote fauna dispersal between habitat patches, that should be also protected. In this study, the caves surrounding landscapes show impacts of the human use, and the small remaining forest patches were important for cave invertebrate species maintenance.

Considering the current Brazilian legislation (Brazil. Decree n^o 6.640/2008), there is an obligation to protect the area corresponding to the cave linear projection on the surface and also a radius of 250 meters around this projection (Portaria IBAMA n^o 887/1990). Unfortunately, there are no studies showing an eventual efficiency of such radius to preserve cave communities. Our study could be the first step to improve Brazilian legislation, since it provides a new methodology to evaluate three aspects: cave invertebrate communities, cave physical traits and surface land use.

We sincerely thank Rogério Tavares, who is manager of “Parque Estadual do Sumidouro” and the Instituto Estadual de Florestas de Minas Gerais (IEF) for logistic support. We also thank Dr. Marcelo Passamani, Msc. Marcus Paulo de Oliveira, Msc. Maricélio Medeiros, Msc. Victor Hugo Oliveira for support and valuable suggestions in the field and specially Dra. Carla Ribas, Dr. Paulo Pompeu, Dr. Júlio Louzada and Rafaela Bastos Pereira for assistance in writing this paper. This research paper was partly produced during the discipline PEC 527 – Scientific Publication and also PEC 813/814 - Field Curse, from the Post-Graduate School in Applied Ecology, at Federal University of Lavras. The authors were financed by CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior). Funding was provided by the Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG) and Conselho Nacional de Pesquisa to R.L.F. (CNPq grantnr. 304682/2014–4). Finally we would like to thank Dr. Claudio Di Russo, Dr. Leonardo Latella and Dra. Oana Moldovan, for their suggestions, which certainly improved the work. Authors declare that all experiments comply with the current Brazilian laws, in which the work was performed.