Invertebrates diversity in mountain Neotropical quartzite caves: which factors can influence the composition, richness, and distribution of the cave communities?

Twenty caves located in a high altitudinal quartzite area in Brazil were examined for invertebrate richness and composition and in terms of environmental factors that determine community structure. We evaluate how distance, altitude, cave extension, environmental stability, number and size of cave entrances and stream presence can act on species composition and richness. The caves presented a high richness of troglophilic (463 spp.) and troglobitic species (6 spp.) in relation to other siliciclastic caves around the world. The average richness was 39.55 species per cave (sd = 21.87), the quantitative similarity among caves was 41% and turnover was βrepl. = 0.769. Araneae (20% of the sampled species), Diptera (18%) and Coleoptera (14%) were the dominant orders regarding species richness. Only twenty percent of the caves were placed out of the confidence interval of the average taxonomic distinctness (∆+); however, the ∆+ decreased with the increase of environmental stability. Cave extension and stream presence were the main factors determining the variation of species composition among caves. Cave extension also influenced species richness variations. Furthermore, the total richness and richness of troglobitic species increased with cave extension. The threats to these habitats further revealed that the fauna is at risk due to tourism, trampling and natural soil erosion that can promote microhabitat alterations. Therefore, quartzite caves also require special attention regarding conservation actions in order to keep their natural biological dynamics. Subterranean Biology 33: 23–43 (2020) doi: 10.3897/subtbiol.33.46444 http://subtbiol.pensoft.net Copyright Marconi Souza Silva et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. RESEARCH ARTICLE Subterranean Biology Published by The International Society for Subterranean Biology A peer-reviewed open-access journal


Introduction
Studies related to ecology and conservation of subterranean fauna have been receiving increasing attention in recent years (Mammola 2018, Mammola and Isaia 2018, Pellegrini et al. 2018, Moldovan et al. 2018. However, most of the knowledge regarding subterranean fauna has been historically focused on limestone caves so that other kinds of rocks, such as siliciclastic, have received less attention (Sharratt et al. 2000, Souza-Silva et al. 2011. Despite quartzite caves being less frequent than caves in other lithologies, studies with fauna from these habitats have been conducted in some regions of the world, as in Cape Peninsula in South Africa (Sharratt et al. 2000), in the Northwestern United States (Peck and Peck 1982), in the Amazon Tepuis in South America (Galán andHerrera 2006, Aubrecht et al. 2012), in the Brazilian Atlantic Rain Forest (Zeppelini et al. 2003, Gallão andBichuette 2015) and in Brazilian Savannah (Guadanucci et al. 2015). Some of those studies have shown that quartzite and sandstone caves are important habitats for troglophilic and troglobitic fauna in Neotropics, thus deserving attention for conservation (Gallão andBichuette 2015, Souza-Silva et al. 2015).
In Brazil, the cave fauna started to be systematically studied in the 1980s (Dessen et al. 1980), but only a few studies assessed ecological aspects of cave community structure. Furthermore, most of the knowledge on cave ecology in Brazil comes from works conducted in limestone caves (Bento et al. 2016, Pellegrini et al. 2016, Simões et al. 2015, Rabelo et al. 2018.
However, a study performed on 33 temperate sandstone caves of the Cape Peninsula in South Africa (approximately 20-90 m long), found only 85 species (Sharratt et al. 2000). They did not present the number of species per caves, which also precludes further comparisons.
However, the previous studies did not present information on how composition, richness, and distribution cave communities are related to some cave attributes (distance between caves, cave size, number and size of entrances and altitudinal position among others). Such features were proven to be determinant for local and regional species richness and dissimilarity, mainly in the tropics (Souza-Silva et al. 2011, Simões et al. 2015, Jaffé et al. 2016. This study aimed to evaluate how some local environmental factors (distance, altitude, cave extension, environmental stability, number and size of entrances and stream presence) can act on invertebrate composition and richness. In addition, we described the human uses and alterations inside and surrounding the caves.

Study area
The study was conducted in the Ibitipoca mountain quartzite province, south of Minas Gerais state, Brazil. The caves in this region were formed and modeled by hierarchically organized drainages influenced by differences between the local water table and the regional base level. The Ibitipoca Mountain belongs to the Andrelândia geological group mainly composed of quartzite rocks of Mesoproterozoic lithostratigraphic age (Auler and Sauro 2019).
The mountain is located within a protected area, with 1,488 ha of extension and altitude ranging from 1,200 to 1,784 m asl. This reserve was created in 1973 and protects epigean fauna and flora and quartzite caves and their fauna (Nobre et al. 2013). The dominant vegetation is grasslands on hilltops and rainforest in the valleys ( Figure  1). The climate is tropical (Cwb of Köppen) with mild summers, with dry (May to August) and rainy (August to January) well defined seasons (Alvares et al. 2013). The external temperatures range from 2 to 20 °C and internal caves temperatures range from 12 to 20 °C (Souza-Silva et al. 2013).

Environmental features of quartzite caves
Most of the caves in this study were mapped by Silva (2004) using a standardized mapping methodology with a British Cave Research Association (BCRA) -4C survey grade. Additional information related to the cave extension (for those caves that were not mapped), number and area of entrances were determined during the fieldwork (as in Bento et al. 2016). The altitude above sea level and geographic position of the caves were obtained with a Global Positioning System (GPS) in decimal degrees (Table 1). Moreover, the width of entrances and linear development of the caves were measured using a laser. The greatest horizontal length of an entrance was considered as width, while the greatest vertical length was considered as height. The cave linear development represented the linear development sampled in each cave since some caves were not sampled throughout their total length (Souza-Silva et al. 2011).

Faunal survey
Species richness and composition of the invertebrate communities were assessed in 20 caves (Table 1, Figure 1). Only one visit to each cave was conducted. Sampling was carried out only once by visual search within the accessible parts of each cave, prioritizing places with organic matter (such as plant debris, carcasses, and guano) and humid soils, cracks, speleothems, water bodies and spaces under rocks. Hand collections were made with the aid of tweezers, brushes and entomological nets (Bento et al. 2016, Wynne et al. 2019. Invertebrates were collected from water bodies with the aid of forceps. In the Table 1. Biotic and abiotic characteristics of the 20 quartzite caves in southeastern Brazil. Cave with stream (S) or dry caves (D), total richness (TS); relative richness (RS), abundance (A), altitude (ALT in meters), geographic coordinates (Lat and Long in decimal degrees), Sampled extension of the caves (SE in meters), Number of morphotypes with troglomorphic traits (ST), entrance number (NE) extension of the entrances (EE) and results of the environmental stability index (IEA). *Caves open to tourist visitation. Numbers above cave names are the morphotypes with troglomorphic traits; 1 -Blattodea, 2 -Brasilomma enigmatica, 3 -Hypogastruridae, 4 -Projapygidae, 5 -Palpigradi, 6 -Pselaphidae. laboratory, invertebrates were separated into morphotypes (Oliver and Beattie 1996). The sampling time spent in each cave was dependent on the extension of the cave. The determination of potentially obligate cave species was conducted by identifying the specimens with troglomorphic traits (Christiansen 2012). However, for some specific groups (e.g., Palpigradi, Diplura), other traits were considered, such as the increased number of sensory receptors (lateral organs) (Souza and Ferreira 2012) and an increased number of antennomeres and cercal articles (Sendra et al. 2012). Furthermore, experts on Diplura, Blattodea, and Collembola were consulted to evaluate the degree of troglomorphisms.

Descriptions of human uses and impacts
Human modifications were determined in relation to uses and impacts. Tourist and religious activities were considered uses while real impacts were trampling, illumination and construction resulting from these activities . Impacts were determined for each cave as a function of the presence or absence of visual modifications inside the cave.

Data analyses
The components of beta diversity were calculated using the BAT package developed by Cardoso et al. (2015) in the software R to assess βtotal (overall beta diversity), βrepl (the replacement component) and βrich (the richness differences component) (R Development Core Team 2008).
The average taxonomic distinctness (∆+) analysis was conducted with the software Primer 7 using 297 from the 463 morphotypes because they were identified up to family level (Anderson et al. 2008). We used Phyla (weight 100), Class (weight 80), Order (weight 60), Family (weight 40) and morphotypes (weight 20) as variables for a matrix of morphotypes distribution among caves (Anderson et al. 2008). Previous analysis comparing the group of sampled morphotypes and the above mentioned 297 morphotypes were carried out using a Spearman Rank Order Correlations to test the representativeness of this sub-sample (of species identified until the family level), which revealed regression Rs = 0.93 and p ≤ 0.05.
The Environmental stability of each cave was determined using the Environmental Stability Index (IEA) proposed by Ferreira (2004) (Pellegrini et al. 2016, Bento et al. 2016, which considers the degree of isolation between the cave and epigean environments through a mathematical ratio, calculated as follows. For caves having just one entrance: IEA = ln(AT/EE) and for caves having more than one entrance: IEA= ln((AT/∑EE)/((NE)*(x -DEE)/AT))), in which: AT (total size of each cave), EE (sum of the highest and longest measurements of cave entrances using perpendicular lines), NE (number of entrances), DEE (average distance between entrances, taken from one reference entrance).
A non-parametric multivariate analysis (DistLM -Distance-based Linear Model) was used to evaluate the influences of the distance between caves (dist), the extension of the sampled cave (SE), environmental stability (IEA), number (NE) and size of entrances (EE) and altitude (Alt) over invertebrate composition, total richness and average taxonomic distinctness with AICc as selection criteria and Forward as selection procedure (Anderson et al. 2008). The similarity measure based on Bray-Curtis Index was used for fauna composition and Euclidean distance was used for total richness and average taxonomic distinctness in DistLM analysis (Anderson et al. 2008). Jaccard similarity based on presence/absence, specifically as a measure of ß diversity and tested against predictors variables (dist, SE, IEA, NE, EE, Alt) using DistLM analysis. Once the definition of ß-diversity is based on variation in the identities of species (Whittaker 1972), the similarity measures with abundance do not provide a ß diversity measure (Anderson et al. 2008).
The distance-based redundancy analysis (dbRDA) was performed to determine the strength and direction (-or +) of the predictor variables relationship selected by the DistLM routine. A metric multidimensional scaling (MDS) using bootstrap-average analysis was performed to determine the level of variation in species composition within sampled caves with and without stream and to produce two 95% bootstrap regions (Clarke et al. 2014).
Similarity analysis(ANOSIM) one-way layout with pairwise analysis was used to select faunal group formation based on Bray-Curtis similarity using caves with and without streams, frequency classes of cave sampled extension (SE), environmental stability (IEA), number (NE) and extension of entrances (EE) and altitude (Alt) as selected factors. The Similarity Percentages analysis (SIMPER) was used to determining species responsible for sample groupingsrs using Bray-Curtis dissimilarities (Clarke 1993). These analyses were performed in the Plymouth routine in Multivariate ecological research -Primer 7 (http://www.primer-e.com). We also used the Spearman correlation test (R s ) (Zar 1984) to evaluate the relationship between cave physical attributes (SE, IEA, NE, EE, and Alt) and total richness and average taxonomic distinctness. Significant differences of average richness, diversity and taxonomic distinctness among caves with and without stream were evaluated using Mann-Whitney U Test (Sprent and Smeeton 2000).

Environmental factors of the caves
Caves are located at altitudes between 1270 and 1669 m asl (sd = 150 m). Sampled cave extension varied from 20 to 650 m in length (sd = 202 m), presenting one to six entrances. Surface of entrances varied from 1 to 80 m 2 (sd = 26 m). Environmental stability varied from 1.4 to 10.93 (Table 1).

Faunal composition, abundance, and richness
We found a total of 12,123 individuals distributed in 463 morphotypes and at least 117 families. The composition considering higher taxa is presented in Figure 2A and Table 2.
The Distance-based linear models (DistLM) revealed in marginal tests that the sampled extension (SE) of the caves (R 2 = 0.08, AICc = 168.7, Pseudo-F = 1.5071; p = 0.01) was the only predictor determining the similarity of cave communities, both for Bray-Curtis and Jaccard similarities. The two axes of the distance-based redundancy analysis (dbRDA) graphic model captured nearly 52.6% of the variability in the fitted  model and 16.4% of the total variation in the data cloud. The first overlay showed that the first dbRDA axis is strongly related to cave sampled extension (SE) (Figure 4). The Distance-based linear models (DistLM) revealed that the sampled cave extension (SE) was also the only predictor influencing variations over species richness (R 2 = 0.335, AICc = 117.45, Pseudo-F = 9.1066, p = 0.009) of the cave communities. Furthermore, only total richness (R S = 0.58, p ≤ 0.05) and richness of obligate cave species (R S = 0.58, p ≤ 0.05) increased with cave extension. Finally, Figure 5 shows the two groups formed in ANOSIM between caves with and without a stream (R = 0.262, p = 0.02) and the bootstrap data variation within the 95% confidence interval. Araneae (Mesabolivar sp., Ochyroceratidae spp., Theridiidae sp.), Opiliones (Mitogoniella indistincta Mello-Leitão, 1936), Diplopoda (Pseudonannolene sp.), Ensifera (Endecous sp.), Psocoptera (Ptiloneuridae sp.), Collembola sp., Diptera (Tipulidade sp.), Isopoda sp., Hemiptera (Emesinae sp.) andTrichoptera sp., were the taxa that most contributed for disimilarity between the two types of caves.

Trophic resources for the cave invertebrates
Organic resources were composed of plant debris deposited close to vertical or horizontal entrances, as well as sparse roots, root stalagmites, termite galleries, guano of carnivorous bats (Chrotopteus auritus Peters, 1856), hematophagous bats (Desmodus rotundus (É. Geoffroy, 1810)) and swifts (Streptoprocne biscutata (Sclater, 1865)), a very abundant bird in these caves. In some caves, bacterial and fungal biofilms were The two axes explained nearly 55% of the variability in the fitted model and nearly 17% of the total variation in the data cloud. The first overlay shows how the first dbRDA axis is strongly related to cave sampled extension. seen on different substrates, which were identified as an alternative organic resource for invertebrates.
The bat guano piles were always small and scarce, but were colonized by Diptera larvae (1500 specimens in total), Collembola (300 specimens in total), Staphylinidae (250 specimens) and Leiodidae (300 specimens), while guano deposits of swifts harbored Acari (50 specimens), Diptera larvae (300 specimens, crickets and spiders. The bat species Desmodus rotundus was the most frequent in caves on the park periphery, quite close to cattles. A high diversity of invertebrates, such as Ensifera, Acari, Coleoptera and Diptera larvae, Annelida (Haplotaxida) were found in the guano of these bat. In Casas cave, Collembola, Acari, and Blattodea were observed associated with termite galleries or abandoned nests, the only macroscopic organic matter observed inside this cave. Top predator taxa were Opiliones (about 1000 counted Mitogoniela sp.), Reduviidae (about 200 counted Zelurus sp.), Pholcidae (about 500 counted Mesabolivar sp.) among others.

Description of human uses and impacts
All impacts observed in the caves were the consequences of tourism. Impacts like graffiti on the cave walls, trails and trampled soil were the most common, which were observed in almost all the studied caves. In some caves, the entrances located near the touristic trails garbage (organic and plastics) was found. In the Cruz cave, a wooden ladder was installed to facilitate the access of visitors. Currently, only three caves in the Park are open to tourist visitation: Pião, Coelhos and Monjolinhos caves. Other caves, like Bichana I and II, showed no signs of visitation, although they are located near the access road. Although few impacts were found inside the Park boundaries, the surrounding forests were removed for pastures and monocultures (such as Eucalyptus).

Discussion
Besides the influence of potential epigean colonizers, the higher average richness found in the present work (39.55 spp., SD = 21.87 spp.) may have been influenced by the size of the studied caves (average: 180 m in length), which can promote a greater heterogeneity of habitats and support richer fauna. Furthermore, the dissimilarity and turnover may also have been determined by a higher variation in the heterogeneity of caves in terms of number and size of entrances, microhabitats and trophic conditions, which can promote diverse and heterogeneous communities (Howarth 1993, Souza-Silva et al. 2011, Lunghi et al. 2014, Simões et al. 2015. The environmental structural complexity by habitat variation and resource amount may increase the niche availability and thus allow more species to coexist (Tews et al. 2004, Schneider et al. 2011, Stein et al. 2014, Busse et al. 2018.
Many troglophilic species do not occur randomly in caves, being preferentially found in humid, deep and dark areas (Lunghi et al. 2014). Furthermore, some studies in limestone Brazilian caves have suggested that factors such as cave extension, number, size and position of entrances, besides their trophic conditions and human impacts, may influence the invertebrate species richness and replacement (Ferreira and Horta 2001, Souza-Silva et al. 2011, Simões et al. 2015, Rabelo et al. 2018. It is known that cave extension represents an important environmental component determining species richness, since it can eventually provide more habitat heterogeneity (Culver and Pipan 2009, Lunghi et al. 2014, Manenti et al. 2015 and availability of organic resources (specially guano), since larger caves may support larger bats colonies (Brunet andMedellín 2001, Souza-Silva et al. 2011).
The influence of big cave entrances is well known since cave entrances act as ecotones between epigean and hypogean sites, sheltering rich communities, with representatives of both epigean and hypogean fauna (Prous et al. 2015).
The presence of streams increase humidity and bring organic matter inside caves, and can also transport epigean species inside caves (Waters 1981, Souza Silva et al. 2011a. All these characteristics provided by the presence of streams can lead to distinct community compositions in wet and dry quartzite caves, as found by Simões et al. (2015) for some limestone caves in Brazil.

Obligate cave species diversity
In Brazil, with exception of the iron ore cavities Ferreira et al. 2018), non-carbonate caves usually have a low richness of obligate cave species (one to three species) in comparison to limestone (Souza-Silva et al. 2011). However, for the Ibitipoca quartzite landscape, a higher number of obligate cave species was found. Two factors may be determinant for this richness. It is known that variations in environmental conditions associated with a complex geological history are leading causes for the colonization of the subterranean environment in some areas of the world, e. g. western Balkans (Deharveng et al. 2012, Mammola 2017, which can be also the cause of cave colonization in Ibitipoca caves. However the 37 species with troglomorphic traits sampled by Gallão and Bichuette (2015) Bichuette, Pinna & Trajano, 2008). Sharratt et al. (2000) reviewed the fauna of 31 sandstone caves placed between 450-750 m asl on the Cape Peninsula, South Africa, and found 85 invertebrate morphotypes (13 of them troglobites). Peck and Peck (1982) observed 15 troglophiles and 2 obligate cave species in Devils Den cave, USA. Although these studies used different sampling methodologies, these caves showed a lower average richness when compared to the quartzite caves from this study. The smaller size, lower number of entrances might play a role in the richness in siliciclastic caves. To these features, the latitude of caves can be added possibly influencing the richness, since tropical regions have greater species richness (and diversity) compared to temperate regions (Gaston and Blackburn 2000) and thus, the pool of potential cave colonizers is richer.

Threats to the cave fauna
The use of caves for tourism is not uncommon and can be extremely important for the local economy worldwide (Bočić et al. 2006, Polak andPipan 2011). This practice can be conducted with direct alterations to the cave structure, with the installation of stairs, railings and lighting equipment or without such modifications, characterizing an option for sporting nature (Polak and Pipan 2011). In fact, such modifications can cause intense instantaneous or cumulative pressures to the fauna (Gillienson 2011), directly influencing the species richness and abundance. In some cases, the touristic impact may focus on alterations in entrance zones, affecting both the flow of animals to the environment (e.g. bats) and the availability of organic resources (Polak and Pipan 2011). However, it is important to emphasize that depending on the cave structure (like the size of the chambers), the impacts can be different for the cave biodiversity (Faille et al. 2015).
Ibitipoca is a State Park and represents one of the most visited areas in the state of Minas Gerais. Nevertheless, its caves are not arranged for tourists. The management plan of the Park included caves, but cave species were not considered for planning conservation strategies (Trajano et al. 2007).
Areas deforested for cattle ranching were observed in the surroundings of Ibitipoca Mountain and may have directly influenced the abundance of Desmodus rotundus (hematophagous bat) in peripheral caves. Therefore, further attention to the preservation of forests surrounding the Park is also required for the conservation of cave invertebrates because obligate cave species depends on the bat and swift guano.

Final considerations
The present study revealed a rich and diverse invertebrate community sheltered in Ibitipoca quartzite caves, influenced mainly by the extension of the caves and presence of streams. This relationship has been recurrent in several other studies conducted with caves associated with different lithologies in Brazil and can be considered a keystone element for the maintenance of cave biodiversity.
Carlos, José Geraldo, Elias and José Luiz for provide accommodation and help with sample permission. RLF is grateful to the National Council of Technological and Scientific Development (CNPq) for the research grant No. 308334/2018-3. The authors are also grateful to the editors and reviewers for theirs valuable suggestions. Ross Martin Thomas reviewed the English. Lucas Mendes Rabelo contributed to this work as a Master degree student