Roosts are essential for the survival of most animals. Due to homothermic requirements, mammals are particularly dependent on roost quality and availability. Bats select their roosts in a species-specific way, likely related to species´ different physiological and adaptive needs. Unlike species whose individuals roost solitarily, roost selection is critical for bats forming large colonies due to the requirements for maintaining thousands of individuals in a single shelter. This is the case of Pteronotus (Mormoopidae), whose colonies reach hundreds of thousands of bats. Using captures, bioacoustics, and automated censuses, we evaluated how cave size, ceiling characteristics, environmental stability, temperature, and humidity influence the formation of exceptionally large colonies, species richness and composition in caves in north-eastern Brazil. We expected that colonies would be positively related to cave size and stability and internal cave selection would be species-specific, but larger and more environmentally stable caves would have higher richness. Pteronotus colonies were positively related to cave size, stability, and ceiling characteristics, and their presence strongly influenced cave temperature variation. Species richness was positively correlated to a cave stability index. Species other than Pteronotus preferred different climatic and ceiling characteristics. We detected an indirect influence of the large colonies of Pteronotus on the species richness and occupation inside caves. On the other hand, such caves favour species coexistence, as they offer a range of microenvironments, reducing niche overlap in their interior. P. gymnonotus and P. personatus are both key- and umbrella-species for cave ecosystems, stressing the need for specific conservation strategies in Brazil.

1.1  Study area

We sampled eleven caves distributed in the states of Sergipe (2 caves), Pernambuco (3), Rio Grande do Norte (3), and Ceará (3), in the Northeast Region of Brazil (Figure 1). The caves differed in formation, having limestone or sandstone origin (Table 1).

The caves of Sergipe are located in the ecotone between the Atlantic Forest and Caatinga biomes, where the Tropical Megathermal climate predominates, with dry season during summer and temperatures varying between 23 and 27°C. Average annual rainfall in Sergipe ranges from 1,000 and 1,400 mm, usually concentrated from April to August (Alvares et al., 2013; Aragão et al., 2013). The other caves (n = 9) are located in the Brazilian semi-arid region, which is part of the Caatinga biome characterized as a seasonally dry tropical forest (Silva et al., 2017). The region has a hot semi-arid climate, with temperatures ranging between 25 and 30°C and average annual rainfall varying spatiotemporally between 600 and 1,200 mm usually concentrated from March to July (Andrade et al., 2017). Caves in Sergipe, Rio Grande do Norte, and Ceará were sampled in July 2019, whereas caves in Pernambuco were sampled in February 2021. Although samplings were carried out in different periods, the precipitation and temperature conditions at the time of both sampling were similar (INMET, 2019; 2021).

1.2  Abiotic data collection

Cave size ranged from 44 to 707 m in horizontal projection (Table 1). We obtained data for cave size from the Brazilian National Registry of Speleological Information - CANIE (CECAV 2019) and estimated the size of unregistered caves through topographic mapping (Cavalcanti, 1996).

The assessment of the availability of microhabitats for bats inside caves was carried out by characterizing the ceiling of each cave, considering the structures that species use. Thus, we counted the number of crevices, domes, and holes (Table 1). Crevices were considered cracks found in rocks. Domes and holes are concavities present in the ceiling and were distinguished by their dimensions: domes were defined as being larger in diameter than depth, whereas holes are more deep than wider.

We assessed the environmental stability of caves using the Environmental Stability Index – ESI (e.g., Ferreira, 2004; Bento et al., 2016; Pellegrini et al., 2016; Barros et al., 2020). The index (ESI) reflects cave stability in relation to the influences of the external environment, considering the cave size (HP), the number of entrances (NE), the average distance between entrances (DE - if there is more than one), and the entrance width (EW), according to the formulas:

ESI = ln (HP/EW), for caves bearing a single entrance; and

ESI = ln [HP x (HP/∑EW) / (NE*DE)], for caves bearing more than one entrance.

Environmental stability means that the influence of the daily and annual climatic variations from the external environments are minimized. Thus, larger caves with a single entrance tend to have a higher ESI, and probably will have less influence from the external variation of the climatic conditions, than smaller caves and caves bearing multiple entrances (Ferreira, 2004). The ESI of the sampled caves ranged from 2.18 to 6.27 (Table 1).

We calculated the average and the range of variation of temperature and humidity inside caves to assess microclimatic conditions. We took ten equidistant measurements inside each cave, considering its total horizontal projection (from the entrance to the deepest chamber), following Barros et al. (2020). The average temperature inside the different caves ranged from 27.1 to 36.3°C, with a variation range between 1.9 and 9.8°C. The relative humidity recorded ranged from 55.8 to 94.4%, with a variation range between 8 and 29% (Table 1).

1.3  Biotic data collection

We captured bats inside the cave using hand nets during the day to get an estimate of species richness at each cave. Captured bats were identified to species in the field using identification keys and then released (Gardner, 2008; Diaz et al., 2016). Bat capture was authorized by the SISBIO license 68992-1 and carried out in accordance with the standards of the American Society of Mammalogists (Sikes et al., 2019), with the permit # 114/2019 of the Ethics Committee for the Animal Use of the Federal University of Pernambuco – CEUA/ UFPE. To confirm identification of species difficult to capture (e.g., insectivores), we carried out acoustic recordings with ultrasound detectors (model SM4 – Wildlife Acoustics, Massachusetts, USA) positioned at the cave entrance simultaneously with infrared filming. Recorders were configured to capture the frequency range up to 384 kHz and 16-bit depth of audio. The echolocation signals recorded were analysed using the Raven Pro 1.5 software (Cornel Lab of Ornithology, Ithaca, USA) and identified at the lowest taxonomic level possible, following Arias-Aguilar et al. (2018).

We used the non-invasive technique of population census based on thermal infrared imaging to estimate bat population in each cave (e.g., Otálora-Ardila et al., 2019). A thermal camera (FLIR E60) was installed at the cave entrance for three hours during emergence, then, using a specially developed algorithm for counting (Rodrigues et al., 2016), bats identified in the images were counted.

1.4  Data analysis

First, we tested data for normality. We used Spearman’s correlation analysis to assess the relationship between species abundance and other variables (richness, stability, horizontal projection, temperature, and humidity). We opted to use correlation because, for some variables (e.g., temperature and humidity), abundance can be an independent variable, and for others (e.g., cave size and stability), abundance acts as the dependent variable.

To evaluate the effect of all cave characteristics on richness, we performed linear regressions with the variables: horizontal projection (HP), environmental stability index (ESI), and temperature and humidity range of variation. We opted to use linear regressions because of the number of caves sampled and the low number of variables.

We performed a multivariate analysis of abundance based on generalized linear models (Wang et al., 2012) to assess the relationship between cave characteristics and species composition, and cave characteristics and individual species. Models were created combining the following variables: horizontal projection, environmental stability index, average and range of temperature and humidity, holes, domes, and crevices (see Supporting Information). This analysis was performed in the mvabund package, using the manyglm function (Wang et al., 2012), in R version 4.0.3 (R Core Team, 2020). Although the analysis evaluates abundance, it is possible to use presence/absence data through the manyglm function. As the abundance values obtained from the thermal camera counts are for all bats in the cave without differentiating species, we used presence and absence data in this analysis, with the “binomial” distribution family specified in each model. We carried out diagnostic tests using the graphics of the residual distribution to ensure that the assumption inferred in each model was correct (Figure S1–S6). We used the anova.manyglm function to assess the significance of predictor variables on species composition. We applied univariate tests using the unadjusted p-value to determine the individual response by each species. For this analysis, we considered only species that occurred in at least three caves, as the use of rare species could generate biased results.

We used presence and absence data from the bat species to analyze, using mvabund package, the relationship with the caves characteristics. The data is organized in a spreadsheet containing in each column the values for the caves features, followed by the data of presence/absence of each species. The file “mvabund” was used as an input in R on the script described below.