Optimizing speleological monitoring efforts: insights from long-term data for tropical iron caves

Introduction

Quantifying long-term changes in abundance of cave-dwelling organisms and identifying indicator species, reflecting the health status of subterranean ecosystems, are among the fundamental research goals of modern subterranean conservation biology (Mammola et al., 2020)⁠. For instance, the lack of knowledge about the factors underpinning abundance patterns in space and time are among the main impediments to the effective protection of cave fauna (Cardoso et al., 2011)⁠. Long-term studies in caves are scarce (Di Russo et al., 1997; Salvidio et al., 2019)⁠, and most previous efforts assessing community-level responses have evaluated population dynamics (Bichuette & Trajano, 2003; Ferreira et al., 2005; Lunghi, 2018)⁠, ecological niches (Mammola & Isaia, 2016; Mammola, Piano & Isaia, 2016)⁠, or temporal and spatial variation (Tobin, Hutchins & Schwartz, 2013; Ferreira et al., 2015; Owen et al., 2016; Paixão, Ferreira & Paixão, 2017; Mammola & Isaia, 2018; Ferreira & Pellegrini, 2019; Pellegrini, Faria & Ferreira, 2020)⁠. Few studies have evaluated the influence of anthropogenic disturbance on cave biodiversity (Bernardi, Souza-Silva & Ferreira, 2010; Pellegrini & Lopes Ferreira, 2012; Faille, Bourdeau & Deharveng, 2015; Cajaiba, Cabral & Santos, 2016; Pellegrini et al., 2016; Jaffé et al., 2018).

Due to the unique characteristics of subterranean environments, an important fraction of cave fauna exhibits adaptations for life in these extreme environments (Pipan & Culver, 2013)⁠. Some of these species are obligate subterranean dwellers and often comprise narrow-range endemic and threatened species (Harvey, 2002)⁠, so stringent legislation has been put in place in some countries to protect them (Harvey et al., 2011; Culver & Pipan, 2014)⁠. In Brazil, companies executing projects that could potentially impact cave ecosystems are required by law to assess the extent of impacts and implement control, monitoring and/or compensation measures (CONAMA, 1986; Brasil, 2008; MMA/ICMBio, 2019)⁠. After environmental licenses are granted, some caves are included in long-term monitoring programs, ultimately seeking to detect possible disturbances on subterranean fauna. These studies generate comprehensive biological databases containing valuable information for numerous caves sampled over long periods of time (Jaffé et al., 2016, 2018; Trevelin et al., 2019)⁠. However, although many recommendations have been made to monitor cave biodiversity (Eberhard, 2001; National Park Service, 2015; Culver & Sket, 2016)⁠, methods still lack standardization, and no quantitative assessment has yet tried to optimize the amount and type of information required to efficiently identify disturbances in cave ecosystems. This is nevertheless essential to design systematic, repeatable, and intensive surveys of cave-dwelling organisms, allowing the formulation of evidence-based management decisions (Wynne et al., 2018, 2019)⁠.

In Brazil, most cave monitoring programs have focused on assessing temporal changes in relative abundance in a set of selected species (ATIVO AMBIENTAL, 2019; BRANDT, 2019)⁠. However, the temporal frequency of field surveys, the impact of seasonal fluctuations in population size, and the sample sizes needed to detect temporal changes in population abundance, are yet to be systematically assessed. Moreover, the selection of species surveyed in these monitoring programs is not based on their usefulness as disturbance bio-indicators. Here we aim to fill these gaps, taking advantage of a comprehensive cave monitoring dataset containing abundance measurements for target taxa surveyed across iron caves along four years.

Material & methods

Study area

The study was performed in the Serra dos Carajás region, southeast of the state of Pará, in the Brazilian Amazon. This region is within the limits of the Floresta Nacional de Carajás, a protected area of 400,000 ha allowing sustainable use. The caves analyzed in this study are located in two highlands known as Serra Norte and Serra Sul (Fig. 1). These two regions harbor banded ironstone formations known as cangas, unique campo rupestre ecosystems resembling mountain savannas (Zappi et al., 2019)⁠, and one of the world’s largest deposits of iron ore (Poveromo, 1999)⁠.

Results

Overall community composition was weakly influenced by seasonality, cave size, environmental conditions, and the composition and configuration of landscapes surrounding caves, as more than 87% of variance in community composition remained unexplained by these factors (Table 2).

Seasonality did not influence the ability to detect species abundance trends over time, since the interaction effect between sampling date and season was not significant in any taxa nor cave (Fig. 2). Increasing the number of samples resulted in more similar model coefficients between full and subset models, and root mean squared errors usually stabilized after three surveys (Fig. 3). However, in most species the dry season datasets allowed a more accurate detection of temporal abundance trends, as revealed by lower root mean squared errors (Fig. 3).

Whereas relative abundance was associated to at least one disturbance metric in 22 species (Fig. 4), temporal trends in abundance were found associated with disturbance metrics in only five species (Fig. 5). Overall, two taxa displayed consistent responses across effects, which makes them potential indicator species for cave monitoring programs: The troglobiont Charinus ferreus, which appeared negatively affected by disturbance, and a species belonging the Theraphosidae family, which seem to be favored by disturbance (Table 3).

Discussion

By analyzing abundance measurements for 33 target taxa surveyed across 95 caves along four years, we found that overall community composition was weakly influenced by seasonality, cave size, environmental conditions, and the composition and configuration of landscapes surrounding caves. Furthermore, our results show that seasonality did not influence the ability to detect abundance trends over time. However, in most species, abundance estimates assessed during the dry season resulted in a more accurate detection of temporal abundance trends, and at least three surveys were required to identify global temporal abundance trends. Finally, we identified a subset of species that could potentially serve as short-term disturbance indicators, some showing consistent responses in different analyses.

Subterranean communities have been shown to be affected by seasonality, environmental conditions, cave characteristics, and the structure of surrounding landscapes (Simões, Souza-Silva & Ferreira, 2015; Pellegrini et al., 2016; De Bento et al., 2016; Mammola & Isaia, 2018; Salvidio et al., 2019; Pellegrini, Faria & Ferreira, 2020; Rabelo, Souza-Silva & Ferreira, 2020). However, our results reveal that overall community composition was only weakly influenced by these factors, as our model explained merely 13% of total variation in community composition (Table 2). In contrast, previous work have found that cave morphology, microclimate, cave depth, and sampling date explain up to 50% of the variation in community structure in limestone and marble caves (Tobin, Hutchins & Schwartz, 2013; Lunghi, Manenti & Ficetola, 2014)⁠. Our results thus suggest that other factors, not considered in our analyses, play an important role structuring subterranean communities of iron caves. Inter-specific interactions, for instance, are known to have a profound influences on community structure (Ferreira & Martins, 1999; Mammola, Piano & Isaia, 2016)⁠. Alternatively, biological samples collected in iron caves may not capture the dynamics of the entire subterranean habitat, comprised by a network of fissures and voids and traditionally referred to as Milieu Souterrain Superficiel (MSS) (Culver & Pipan, 2014; Mammola et al., 2016; Mammola, 2018)⁠. For instance, most of the surveyed caves were larger than 5 × 5 m (Fig. S1), so they did not represent suitable sampling sites for the MSS (Mammola et al., 2016)⁠.

Even though seasonality affected overall community composition, it did not influence the ability to detect species abundance trends over time. External climatic conditions are increasingly attenuated at higher cave depths (Tobin, Hutchins & Schwartz, 2013)⁠, so species occurring in the inner portions of caves appear to have life cycles decoupled from external seasons, whereas species inhabiting the outermost portions of caves seem to be more strongly affected by seasonality (Di Russo et al., 1997; Gunn, Hardwick & Wood, 2000; Bichuette & Trajano, 2003; Ferreira et al., 2015; Mammola, Piano & Isaia, 2016; Lunghi, 2018)⁠⁠. Recognizing the impact of seasonality on species detection, the current Brazilian legislation stipulates that cave biodiversity surveys need to comprise at least two sampling events, one during the dry and one during the rainy season (MMA, 2017)⁠. It is worth emphasizing that these sampling requirements targeted a more accurate estimation of species richness, but not the continuous monitoring of focus species in time. Two sampling events are likely insufficient to obtain reliable species richness estimates for highly diverse caves (Auler & Piló, 2015; Wynne et al., 2018)⁠, so some authors have argued for the estimation of optimal sample sizes based on species accumulation curves (Trajano & Bichuette, 2010; Trajano, 2013)⁠. Our results provide the first evidence-based recommendations to optimize sampling efforts of monitoring programs seeking to assess target species abundance through time. Specifically, our findings suggest that monitoring efforts aiming to detect changes in abundance through time do not need to sample during two different seasons each year (Fig. 2). Sampling efforts of such monitoring programs could thus be optimized by performing more focused surveys and by surveying a larger number of caves during the same period each year. Importantly, restricting sampling to a single season could substantially attenuate the negative impact of cave visitation by researchers on subterranean communities (Pellegrini & Ferreira, 2016; Pellegrini & Lopes Ferreira, 2012; Bernardi, Souza-Silva & Ferreira, 2010).

Although the composition and spatial distribution of subterranean communities can remain constant over periods of several years (Salvidio et al., 2019)⁠, our results suggest that sampling during at least three years is necessary to detect temporal changes in abundance patterns in most of our focus species (Fig. 3). We note that our dataset only spans a period of four years (although some caves were sampled multiple times during the same season/year), so it cannot capture longer temporal changes in abundance. We also caution that these results cannot be generalized to all subterranean fauna, as different life histories and generation times will ultimately determine how fast these organisms respond to disturbances (Ferreira, 2005; Mammola et al., 2016; Culver & Pipan, 2019)⁠. Sampling in different seasons did not influence the ability to detect general abundance trends over time, but the dry season datasets allowed a more accurate detection of temporal abundance trends in most species. These results suggest higher detection probabilities in the dry season for the subset of species where RMSE curves show a steeper decrease during the dry season (Fig. 3). Interestingly, this was the case for the troglobitic amblypygid Charinus ferreus, a species that is difficult to detect like other troglobionts (Wynne et al., 2018; Lunghi, 2018)⁠. Our results thus suggest that monitoring programs focusing on terrestrial subterranean fauna from our study region could concentrate sampling activity in the dry season, where most species seem to be easier to detect. Likewise, our findings highlight the importance of implementing long-term monitoring efforts spanning at least 3 years.

The concept of indicator species in ecosystem management relies on the idea of identifying taxa responsive to environmental change, that could inform policies, protocols, and best practices (Carignan & Villard, 2002)⁠. Such environmental indicators (McGeoch, 1998)⁠ seek to provide cost and time effective guidelines to address pressing conservation issues, such as those faced by large-scale mining projects (Sonter, Ali & Watson, 2018)⁠. Assessing the response of subterranean fauna to anthropogenic disturbance nevertheless requires access to long-term cave monitoring datasets, which are remarkably rare for tropical caves (McGeoch, 1998; Carignan & Villard, 2002; Mammola et al., 2020)⁠. Here we identified 20 taxa where overall abundance responded to cave disturbance, and five where temporal abundance trends where associated with disturbance. Only two taxa displayed consistent responses across effects, which makes them candidate indicator species for cave monitoring programs: Charinus ferreus and a species belonging the Theraphosidae family (Table 2). Both are arachnids, a group that was recently identified as biodiversity indicator for iron caves (Trevelin et al., 2019)⁠. Being a top predator restricted to cave ecosystems, the first species is a well-known troglobitic Amblypygi (De Lao Giupponi & De Miranda, 2016)⁠. Its strong and consistent response to disturbances (Figs. 4 and 5) suggest the species is associated with pristine and undisturbed ecosystems, which makes it an ideal disturbance indicator. Theraphosidae spiders, on the other hand, are sedentary sit-and-wait predators from the epigea, rarely occupying subterranean environments for reproduction or shelter (Fonseca-Ferreira, De Zampaulo & Guadanucci, 2017)⁠. Our results suggest that they apparently benefit from disturbance to opportunistically colonize caves, or alternatively, that disturbances in the surrounding external habitats are forcing them to look for shelter inside the caves. The species nevertheless awaits formal taxonomic description, which currently limits its usefulness as an indicator species.

Effect sizes of disturbance on overall abundance and temporal abundance trends where generally small, suggesting that some effects could have remained undetected because they would require sampling over longer time periods (Di Stefano, 2001; Legg & Nagy, 2006)⁠. For instance, the ability to detect trends in tropical bat population abundance was shown to be dependent on the duration of the monitoring efforts, and only long programs (>20 years) showed sufficient statistical power to reliably detect abundance trends (Meyer et al., 2010)⁠. This could explain why some of our focus species did not exhibit coherent responses across analyses, like the troglobionts Pyrgodesmidae sp. and Escadabiidae sp., or opportunistic colonizers like the anuran Leptodactylus pentadactylus or Pristimantis fenestratus. Although empirical evidence from long-term cave monitoring efforts focusing on invertebrates is scarce (Faille, Bourdeau & Deharveng, 2015; Cajaiba, Cabral & Santos, 2016; Owen et al., 2016)⁠, our results suggest that longer monitoring efforts are needed to detect disturbance responses in most cave-dwelling species.

Conclusions

Our study reveals the importance of long-term cave monitoring programs for detecting possible disturbances in subterranean ecosystems, and for using the generated information to optimize future monitoring efforts. Results show that iron cave monitoring programs implemented in our study region could focus sampling efforts in the dry season, where detectability of target species is higher, while assuring data collection for at least three years. Charinus ferreus was identified as the most promising short-term disturbance indicator species.

Supplemental Information