Developing global vulnerabilities and conservation priorities for cave-dwelling bats

Research and conservation interventions are disproportionately focused on taxa perceived as charismatic, while other systems with high levels of endemism, and often under-protected such as caves and subterranean ecosystems remain neglected. Bats are keystone to cave ecosystems making them ideal surrogates to understand diversity patterns and assay conservation priorities. Using a novel framework, we assessed and mapped the global bat cave vulnerabilities and priorities at the biome and site level. Almost half ( N = 678, 48%) of bat species across the world regularly use caves, with 32% endemic to a single country, and 15% currently threatened with extinction. Tropical realms and small-ranged species faced the highest levels of risk. Our analyses consistently showed that most caves with lower threat levels show higher cave biotic potential (i.e., evolutionary distinctiveness and endemism), though this may be due to the loss of species from more disturbed caves. We estimate that 3% to 28% are high-priority caves for conservation in broad-scale (biome level) and fine-scale (site-dependent) analyses respectively. Amongst regions, the highest concentration of conservation priority areas are in the Palearctic and tropical regions (except Afrotropical), which requires more intensive data sampling. Our results further highlight the importance of prioritising bat caves at local scales but show that broader scale analysis is possible if robust cave data is present and effective parameters are included (i.e., appropriate landscape features and threats). Developing priorities for these systems requires more systematic approaches as a standard measure to develop the priorities needed to maintain cave diversity.


Introductory
Global biodiversity is threatened by a variety of drivers, and understanding these is critical to determining effective conservation measures in different systems and scales 1 .
Appropriate evidence-based strategies are essential to optimise the effectiveness of conservation 2,3 . Various strategies have been developed to prioritise taxa and their ecosystems for conservation such as habitat prioritisation and zoning [4][5][6] , but many countries may lack access to such data or resources to effectively implement conservation. Priorities should be matched with future targets and effort allocation to support reaching such targets and ensure that species and ecosystems are well protected [7][8][9][10] . However, many species lack data for effective priorities to be developed, and even fewer receive adequate levels of attention 11,12 . Yet, as much as conservation scientists try to address the most pressing challenges in global biodiversity conservation, many taxa and their associated habitats are overlooked 13 .
In terrestrial ecosystems, megafaunal species (> 15 kg, terrestrial mammals) are often used as flagship or indicator species to provide an umbrella in developing priorities for other taxa and ecosystems, yet they are known to be poor indicators for small-bodied species or the ecosystems they frequently depend on 14,15 . These measures do not ensure effective conservation and often lead to biases in funding allocation, decisions on conservation and management priorities 16,17 thus compromising effective conservation of other taxa with possible ramifications for ecological function 18,19 . Moreover, modern technologies are applied to understand environmental processes on above-ground surface ecosystems (e.g., forest ecosystems) which can easily be mapped with remote sensing. Yet most subterranean ecosystems including caves are challenging to map, and have frequently been overlooked and neglected 20 . This barrier to research risks leaving other ecosystems and the species dependent on them such as bats under-studied and under-protected 21 .
A further barrier to reseach, understanding and even engagement is the special adaptations which not only prevent cave taxa garnering public attention, but the special skills needed to study many cave taxa further adds to the lack of research on cave taxa. Bats are often keystone species in cave ecosystems, and make up the second-largest group of mammals with over 1400 described species distributed across almost all biomes 22,23 . Cave ecosystems host a variety of highly adapted and sensitive organisms, the majority of which are cave obligates and depend on bat guano for nutrients [23][24][25][26] . Some cave systems have relatively high endemism and even a single cave could hold up to 90% undescribed species 27,28 , meaning many thousands of species remain undescribed and potentially at risk.
However, most conservation projects and funds are focused on charismatic megafauna, but neglect ecosystems with higher endemism and without megafauna, such as cave ecosystems.
Bats function as keystone species within cave systems by bringing nutrition from foraging in the surrounding areas into the caves and providing the basic source of nutrition in these largely lightless zones, where the inability for photosynthesis within the system makes this form of nutrition critical. This function makes bats the ideal surrogate taxa to inform priorities and safeguarding these systems 23,29,30 . Yet the reduction of bat populations and habitat alterations is increasingly concerning global conservation biologists and threats to cave-bat populations have been widely discussed 23,31,32 but a standardised index to identify priorities for action to safeguard these systems does not yet exist, hindering larger-scale prioritisation to protect cave systems and their dependant diversity.
Subterranean habitats are systematically threatened by combined direct and surface threats. Bats are conspicuous in subterranean ecosystems but they are vulnerable, as large colony sizes mean that single site impacts can have major consequences on populations especially for locally endemic species 20,32 . The relatively low reproductive rate in bats jeopardizes rapid population recovery from loss and declines 33 . Cave and karst habitats are under-protected, for example, only around 13% of the approximately 800,000 km 2 of tropical Southeast Asia to South China's karst are within protected areas 34 . Unprotected karst is especially susceptible to destruction, with an average loss of around 5.7% of area annually due to mining and other uses [35][36][37] . This loss of cave habitat is coupled with unregulated hunting and tourism, and foraging habitat loss; therefore understanding the impacts of these factors on the population status of bats and biotically important caves are urgently needed 23,30,38,39 . It is estimated that at least 50% of global bat species are dependent on caves 23 but the degree of threat to bat cave communities and prioritisation has never been analysed across large scales 30 . Understanding species distribution cannot ensure the protection of all species but identifying key areas in need of protection enable conservation biologists to develop balanced priorities to inform appropriate protection measures. This is the first extensive study to explore the global diversity patterns and extinction risk of cave-dwelling bats and using this information to create an index to guide effective priority making.
Here we develop a mechanism to integrate different facets of biotic importance and risks across different dimensions and scales using our database to develop global priorities for bat cave protection based on both vulnerability and biotic potential. First, we examine the (i) global distribution and conservation status of cave-dwelling bats, and (ii) patterns of threatening process. Second, we mapped the biotic potential and vulnerability of caves to identify broad-and fine-scale priorities for global bat caves. Our findings pinpoint gaps as a basis for prioritisation schemes, policies relevant to regional and global levels of cave and karst ecosystem conservation and improve Redlist assessments, especially in prioritising karst environments, which represent one of the most challenging ecosystems for redlist assessments.
We compared our cave inventories to that of IUCN-sampled species richness and diversity (Fig. 1). A total of 1930 cave sites were included in the analyses from the current bat cave distribution (Tanalgo et al. submitted in Scientific Data). When compared between IUCN Redlist, our current data for spatial prioritisation included 402 species representing a dataset for 59% of all cave-dwelling species and an equivalent to 29% of global bat species (Fig. 1). We did find a marginally significant relationship between the proportion of species richness and threatened species between the IUCN-based global cave-dwelling bat and DarkCideS dataset (Kendall's  B = 0.60, P = 0.07). The highest completeness of sampled species is in the Neotropics (67.38%) and Indomalayan region (66.08%), where the highest gaps are in Austral-Oceania (40.28%) (Fig. 1). Highest endemism was recorded in Austral-Oceania (58.62%) ( 2 = 227.32, df = 5, P < 0.001). The proportion of threatened species is highest in the Indomalayan (16%) realm ( 2 = 281.18, df = 5, P < 0.01) (Fig. 1). The degree of evolutionary distinctiveness based on bat-cave data significantly differed regionally (Kruskal-Wallis,  2 = 1615.65, df = 5, P < 0.001), and the highest ED was in the Neotropical region (EDmean = 11.92) with the lowest in the Palearctic (EDmean = 5.54). Cave weighted endemism is highest in the Austral-Oceania (CWEmean = 6.32) consistent with the proportion of species endemism observed in the region (Kruskal-Wallis,  2 = 1584.84, df = 5, P < 0.01) (Fig. 2).
Globally based on cave community data, seven out of twelve variables showed a significant relationship with species biotic scores but none of the landscape features and threat variables showed a strong correlation to diversity (Fig. 4a). Tree density, bare ground cover change, and short vegetation cover change showed a positive correlation with species biotic scores. Conversely, distance to river, tall tree loss, nightlight, and population density showed a significant negative correlation (Fig. 4a) though this may in part reflect the challenges of sampling caves in tall forested areas, but also highlights that recently disturbed areas may be threatened. When regionally compared, the mean intensity of landscape features and threat variables differed significantly except for pesticide use (Fig. 4b). Bat caves in the Afrotropical, Austral-Oceania, and Nearctic showed the highest vulnerability to cities and urbanisation. While Nearctic caves showed the nearest distance to roads and the highest mapped mine density. The mean proportion of bare ground cover change is low in arid regions of Afrotropical and Austral-Oceania. Whereas population density and pesticide exposure are exceedingly high in Indomalayan caves (Fig. 4b). Further, socioecological variables (GDP per capita and % forest cover) and cement production as a proxy to assess the vulnerability of geopolitically endemic species showed consistent correlations amongst species diversity attributes (Fig. 4c).

Synthesising Priorities for bat cave conservation
We constructed broad-and fine-scale priorities then compared them across biogeographical realms and climatic regions (i.e., simplified biomes classifications). First, in both scales, there is a higher threshold of cave biotic potential and high vulnerability primarily in the tropics and across most of the world (Fig. 5a-d). Then, we examined the congruences in the relationship between biotic vulnerabilities (BV, status transformed scores) and cave biotic potential thresholds. Primarily, the relationship patterns between cave biological diversity and vulnerability to threats indicate that caves with high biotic potential occur in caves with the lowest threat levels (i.e., cave systems located in relatively pristine ecosystems). In fine-scale analysis, we found a significant relationship in biotic potential and biotic vulnerability score (Kruskal-Wallis test:  2 = 14, df = 3, P = 0.003). Within fine-scale, caves with higher biotic potential are linked to lower vulnerability caves, while there is a lower biotic potential in caves with higher threats and vulnerability (Kruskal-Wallis test:  2 = 6.45, df = 3, P = 0.092). Whilst this pattern is consistent at broad-scales, albeit the relationship was not significant (Kruskal-Wallis test:  2 = 6.45, df = 3, P = 0.092) (Fig. 5ef).
Indices derived from the site-level and biome-level Bat Cave Vulnerability Index (BCVI) were used to construct a regionally comparative prioritisation (Supplementary Table 5). On a broad scale, 95% of the caves show high biotic vulnerability (Status A) but 88% of the caves are in lower biotic potential level (Level 4) in contrast to only 1% with high biotic values (Level 1) (Fig. 6a). The integration of two sub-indices within the broad-scale equates to the uneven and high proportion (83%) of "4A" lower vulnerability index values for overall sampled caves (J' = 0.378) (Fig. 6a). On the other hand, fine-scaled BCVI analyses showed a more even distribution of indices values (J' = 0.917) compared to the broad-scale analyses. At a fine scale, there is an increase of 45% of caves with high biotic potential caves (Level 1), 14% at a mid-high level (Level 2), and 41% of caves had a lower biotic potential (Level 4). Subsequently, an even distribution of biotic vulnerability was also observed with fine-scale BCVI, 45% and 41% of the caves are in high (Status A) to mid-high (Status B) vulnerability, respectively. Moreover, 10% of the caves are in the "1A" high vulnerability index. A high proportion of high vulnerability caves occurs in most of the tropical realm (Fig.   6a).We found a significant difference in cave vulnerability index in both scales (Broad: Our analysis of the sampled caves showed that the priority levels of caves are scaledependent and significantly varied across spatial scales ( 2 = 1281.43, df = 3, P < 0.001) ( Table 4; Data S7). Caves were classified on the basis of the need for different types of intervention or management based on threat and biotic characteristics, most importantly these include "red caves", which whilst diverse are threatened and at need of intervention, and "green caves" which are also diverse but not currently at risk, whilst "yellow caves" are at an intermediate level and may need intervention to prevent species loss and allow recovery. On a broad scale, only 3% of the sampled caves are at "Red Priority", 9% at "Yellow Priority", and 88% at "Blue Priority" levels. The low proportion of high priority caves are significantly concentrated in the Neotropical (45%), Afrotropical (18%) and Indo-Malayan (16%) regions ( 2 = 204.20, df = 10, P < 0.001) (Fig. 6b and Supplementary Table 5). While on a fine scale, we did find a significant increase in the proportion and evenness (J' = 0.395) of high priority caves compared to broad-scale. Of the sampled caves, 28% are "Red Priority" and 4% "Green Priority" caves or sites that contain high biotic potential but low vulnerabilities ( Fig. 6b and Table 1). The patterns of concentration of high priority caves based on realms is highest in the Palearctic (30%) closely followed by Neotropical (29%) and Indo-Malayan (28%) ( 2 = 73.93, df = 15, P < 0.001). When compared by biomes and climatic regions, the 58% and 25% of high priority caves are concentrated in the tropics and temperate regions respectively ( 2 = 56.76, df = 12, P < 0.001) (Fig. 6b). for global bats [40][41][42] . We observed a higher risk of extinction in species with a narrow geographical range distribution (e.g island and nationally endemic species), consistent with other studies 40,43 . Except for the association of endemism to extinction risk varies phylogenetically showing that closely related species have a similar association 43 , island and geopolitical endemism is respectively correlated to sub-orders Yangochiroptera and Yinpterochiroptera.

Understanding threats
Direct threats were directly linked to species extinction risks in all our models. This may be explained by bat traits and behaviour, for example, the individual aggregation behaviour in colonies of most cave-dwelling bat species makes them more susceptible to introduced direct and immediate anthropogenic disturbances such as hunting, harvesting, and unregulated tourism. Large-bodied insectivorous, frugi-nectarivorous are likely susceptible to hunting and harvesting. The 18% of global cave-dwelling bats are threatened by hunting largely in the Afrotropical and Indomalayan regions. This represents a large proportion of 62% of the estimated number of bat species (167 spp.) hunted globally 44 . The high level of hunting and harvesting in the old-world tropics is primarily driven by subsistence and primarily localised particularly in areas with high levels of poverty and driven by the demand for protein sources, food, and traditional medicine [45][46][47] . The most frequently hunted species are those common and hyper-abundant cave species (e.g., Rousettus amplexicaudatus, Eonycteris spelaea, Eidolon helvum), which are usually underprotected because of their widespread and globally non-threatened status 48 . However, as a consequence of the absence of statuary protection for many common species intensifies unscheduled hunting and human use can cause even common and abundant species to become extinct (i.e., locally or regionally) 49 . Separately, unregulated tourism is a threat to 38% of the cave-dwelling species and unregulated activities may alter cave microclimate and affect sensitive species.
Cave-dwelling bats are also at risk from loss and degradation of foraging areas 50,51 , and up to 49% face loss of these habitats by deforestation and agricultural conversion.
Disturbed caves in deforested and agricultural lands drive the loss of specialist bats, whereas less disturbed caves support higher species richness and abundance 50 . Furthermore, increased deforestation and vegetation removal around cave sites increase the exposure of caves to human intrusion and hunting. Additionally, extractive industries of mining and quarrying threaten more than a quarter of cave-dwelling bats, through degredation and destruction of caves and alterations of surface vegetation, which creates barriers that limit species movements and foraging 52 . Although our analyses showed that average mining density is higher in the Nearctic region, this may be caused by a lack of rigorous mapping of mined areas in other regions, and also largely omits quarrying for limestone (as such maps are rarely available) 53 . However, cement export is significantly higher in countries with high cave-bat species richness and threatened species (e.g., in China and Indonesia), in part because of the extensive limestone karsts in these regions.

Conservation and prioritising vulnerable sites
Funding and resources are limited and often focus on a subset of taxa which may not be representative 17,54 , it is imperative to set realistic priorities based on areas with higher risks, vulnerabilities and could protect larger communities 55,56 . The integration of vulnerability to effective conservation planning is often challenging due to vague or inconsistent approaches 57 . Cave ecosystems host a high diversity and endemism yet are rarely included in global priorities 20 . The persistence of high biodiversity in caves is linked to more pristine cave environments with less anthropogenic pressure. Consequently, we observed that large spatial coverage (e.g., biome-dependent) decreases the evenness in priority distribution in a site-level comparison, thus its conservation application to protect caves is scale dependant (e.g., national-level or biome-wide protection).

Conservation decision making depends on the clear delineation between what is
important to develop priorities, as frequently priorities fail to represent what is genuinely important or do not represent communities effectively 14,58 . In particular, species deemed as "charismatic" may monopolise funding whilst failing to provide significant benefits to the wider ecological community 14,15 . Whilst various taxa have been used as indicators for diversity in subterranean habitats 59 , bats represent better surrogates for cave conservation because they not only provide the main source of energy for cave-ecosystems but are also easier to assess and reflect changes from both internal and surfaces 50,51 . Most conservation decision making for habitat protection centres on taxonomic diversity (e.g., counts, abundance, and rarity) 60,61 . Common species can provide ecological services (e.g., nectarivorous cave bats as pollinators) and the reduction of population counts may indicate habitat health due to changes in service provision. Whereas, rare species are often prioritised based on their combined distinct ecological function, high sensitivity and vulnerability to declines 62 .
For cave-ecosystems maintaining the functioning of the subterranean ecosystems requires maintaining both diversity and abundance 30,63,64 , thus a holistic tool that incorporates diversity, rarity and function is needed as a basis for conservation-decision making. There is a slight overlap in priorities between scales such as some high priority caves from the broad-scale analysis are the same in the fine-scale prioritisation, but not all. However, we found broad-scale measures underestimated the priorities of highly vulnerable bat caves (i.e., less even distribution of priorities). Broad-scale prioritisation such as regional or biome levels may facilitate the identification of priority caves. Whereas fine-scale prioritisation (e.g., national-level priorities) enables community-level interactions and responses to be encompassed and also accounts for rare species and the impacts of threats on local populations 30,50 , highlighting the need for good monitoring and assessment data as a basis for priority setting. Furthermore, context-specific threats (e.g., vulnerability to religious activities in Buddhist regions, where an increased proportion of caves may become temples or religious sites) need to be accounted for explicitly for indices to be effective.
For an index to be effective, a clear understanding of diversity patterns and priorities at national levels is an essential first step to implementable policy targets 55,75 . Additionally, few countries have any policy related to the protection of caves and their biota 27 . For example, in the Philippines, the National Cave Conservation Committee aims to identify caves for protection has very broad criteria and focused on archaeological and touristic values rather than ecological components which hamper effective protection or priority setting. National Biodiversity Action Plans (NBSAPS) should include standard provisions for priority identification and monitoring 70 , thus frameworks such as this can provide information that is both consistent between countries and can be usefully applied at national levels.

Caveats and opportunities for Bat cave conservation in the Anthropocene
The uneven distribution of threats and a lack of understanding of their impacts remains a challenge for global bat conservation. Developing any index requires pragmatism in finding indicators that are reliable enough, but for which there is sufficient data available.
For example, the interpretation of extinction risk requires caution as factors may act in synergy and we could not fully account for the intensity of human-induced threats 76 . This is particularly challenging to pinpoint in the context of cave biota, in which even a single disturbance may alter the entire sensitive biota and ecosystems, and as such events cannot be identified vulnerability (e.g accessibility) must be used as an indicator 31,50,77 . Furthermore, the degree of expertise required for bat and cave studies means less data is available compared to other taxonomic groups 74,78 . For instance, in our cave prioritisation, we only accounted the 59% of the global cave-dwelling species, and species coverage varied by region, for example, Indonesia has some of the highest estimated bat cave species richness yet its contribution to the dataset based on surveys and assessments is among the lowest.
Furthermore, accurate systematic and taxonomic studies for bats are vital to appropriate conservation as caves host high endemism and many cave bats (e.g., Rhinolophids) have high numbers of as yet undescribed cryptic species 79 . The lack of distribution data may be explicitly linked to the lack of funding in most biodiverse countries 80 , hindering effective assessment in the countries most in need. Conversely within Europe, the UNEP-EUROBATS provides guidance and regulation which comprehensively includes the protection and monitoring of bats and caves in its territory and member states making parallel and equitable policies for large scale protection 81 . Policies and targets that accurately account for and include monitoring in threatened systems such as caves and karsts are urgently needed and highlight a need for ecosystem-based conservation targets, as species-specific targets risk missing key habitats for neglected taxa 82 .

Synthesis
In conclusion, to protect and conserve a large fraction of species possible it is crucial, we identify highly vulnerable caves with the highest biotic potential and distinctiveness. We develop a comprehensive index to integrate facets of diversity and risk to provide a simple and scalable approach to prioritising caves for protection and delineating between those in need of urgent intervention (high diversity but high threat) and those which whilst not yet threatened require monitoring to ensure they remain protected. Whilst further data is needed, especially for data-poor, species-rich regions, relying on IUCN data alone may not provide balanced National-scale priorities 71,83 . The IUCN Redlist must be utilised alongside other tools and measures of decision making, thus we advocate an integrative conservation approach that synthesises various dimensions of conservation. Here we integrate information from two scales, broad-scale and fine-scale represented by biome-dependent and site-level prioritisation to explore how priorities vary across scales, and how to usefully integrate information to obtain a clearer understanding of sensible priorities.
Our study identified gaps and priorities for bat cave conservation, for example, the lack of data from the Afrotropical region, which reflect what form of intervention may be needed for different sites. To address global gaps for effective conservation and develop more rigorous priorities, collaborative efforts among bat scientists is necessary to enable the development of an effective international agenda and nationally adaptable goals for cave prioritisation and assessment. We propose a framework to identify priorities based on the type of interventions needed to safeguard diversity. This framework calls for standardisation of methods of bat cave assessment and monitoring based on key indicators for diversity and threat, and for this data to be more widely shared to facilitate better conservation and policy.
Furthermore, maintaining cave bat diversity relies on their inclusion into conservation agendas and priorities, and the use of science-based targets and frameworks, synthesising conservation effectiveness (e.g., Conservation Evidence Initiative 84 ) to ensure that approaches such as that detailed here can be effectively applied to enable key sites to be identified, and appropriately protected on a global basis.

Dataset building
We sampled global cave-dwelling bats from the databases of IUCN redlist and the DarkCideS 1.0, a global dataset for cave-dwelling bats, including all species that occur, use, roost, or hibernate in caves and underground habitats for any part of their life histories. All species names were curated and updated using the Mammal Diversity Database 22 and Bats of the World: A taxonomic and geographic database 85 . We included species-specific information including species taxonomy, endemism at geopolitical and biogeographical scales, species range and distributions, conservation status, population trends, ecological traits, and threatening processes. The habitat breadth was determined according to the number of habitats a species occurs 86 . We used the weighted habitat breadth (%) values in the final analyses. Species were then classified based on island endemism and country endemism. We used the data from its congeneric species for those species without trait or ecological data.
We then compared the species turnover between IUCN Redlist and the DarkCideS 1.0 to determine gaps and caveats in the current data knowledge and availability.

Species diversity and distributions of global species
Using IUCN data we calculated geopolitical (e.g., country or continental) species richness based on where the species occurs as (including native, extant, resident, possibly extinct or migrant classifications by the IUCN). Second, we analysed and defined and endemism in two ways: geopolitical endemism (i.e., a country endemic) if the species occurred only in a single sovereign country 87 , and island endemism was classified as island-restricted or predominantly mainland 88 . Species conservation status was assessed according to the IUCN Redlist criteria and were then simplified to obtained binary extinction risk: Nonthreatened (e.g., Least Concern, Near Threatened) and Threatened (e.g., Vulnerable, Endangered, and Critically Endangered). While Data Deficient species were treated as the latter as they may face higher or similar threats with threatened species, hence a lack of data for formal classification 30,40 . We compared species diversity across the biogeographic realm 89,90 (e.g., Indomalaya, Austral-Oceania, Afrotropical, Neotropical, Palearctic, and Nearctic).
Chi-squared test ( 2 ) of association was then used to assess the relationship in species geopolitical endemism, island endemism, conservation status, and population status.
We plotted the logarithm of species richness (i.e., estimated number of species per country) against the logarithm of country land areas (km 2 ) 91 to estimate the broad Species-Area Relationships (SAR) applying the equation S = cA z , where S is the species richness equal to the Area (A) sampled, and z and c are fitted constants 92 . We defined endemism in a simplistic manner that is coherent for national or regional level conservation policymaking.

IUCN-based extinction proportions and risks
Following 93,94 we estimated the species extinction proportions within bat biogeographical realms, endemism (e.g., geopolitical and island), population trends, trophic levels, and families. The IUCN bat data is imperfect and many species lack updated assessments, using the available data we generate the mean extinction risks and upper and lower bound estimates for different groups and species attributes. We calculated the proportion of extinction risk (p̂ extinction) based on the proportion of threatened species as assuming that DD species will have a similar extinction risk as of other threatened categories as they may face similar or higher threats than those non-Data Deficient species 30,94 . We then calculated the lower estimate (p̂ extinction_lower = N o threatened / N o species) with an assumption that DD species are categorised as non-threatened and upper estimates (p̂ extinction_upper = N o threatened + DD / N o species) with the assumption that DD is threatened.
Using a Generalized Linear Model (GLM) we determined the predictors of species binary extinction risk (e.g., threatened vs, non-threatened) for global cave bats and within suborders, Yinpterochiroptera and Yangochiroptera. We used a total of ten (N = 10) explanatory variables divided into geographical variables (geographical range, island endemism, and geopolitical endemism) extracted from, biotic variables (adult body mass (kg, log10) 88 , generation length 95 , forest dependency (yes or no based on the IUCN database), habitat breadth (%), and trophic group. For the trophic groups, we grouped species as herbivores for all species that forage on plant-based resources (e.g., frugivores and nectarivores). Species that forage on animal resources (e.g., insectivorous, sanguivorous, piscivores) were grouped as carnivores. While species that forage on both resources were grouped into omnivores. Our last group of explanatory variables includes the species threat index (SPI) (e.g., direct, indirect, and natural).

Patterns and turnover of threats
We analysed threats in three different levels: species, cave-site, and country-level.
First, we analysed species threats based on the simplified expert-based IUCN redlist. We with a binomial distribution in JAMOVI version 1.2.6, using GAMLj module 97 to determine the influence of species traits on species vulnerability to threats. We used adult body mass (g, log 10), geographical range (log 10) and trophic levels as predictors of risks 98 from to key threatening process that threatens at least 10% of the global cave-dwelling bats.
Second, we mapped the cave surrounding landscapes features to measure cave susceptibility to threats (full details are provided in the supplements). For landscape features, we include canopy height 99 , tree density 100 , distance to river 101 , bare ground cover change 102 , short vegetation cover change 102 , tree canopy change 102 and for vulnerabilities we include distance to urbanisation 102 , distance to roads 103 , mine density 104 , night light 105 , pesticide use 106 , and human density 107,108 . We then assessed the correlation between these variables to species biotic scores (see "Cave Biotic Potential section") and the extent was compared across biogeographical realms.
We explored country ranking in terms of species richness and species-area relationship, threatened and endemic species proportions and its concordance to sociodemographic and environmental indicators as rudimentary indicators of a country's resources to representing its capacity to protect its species and environment. In our analyses, we used percent forest cover 109 as a proxy for habitat intactness and quality per country, with the assumption that caves surrounded by intact forest are more protected and least vulnerable to intrusions 30,50,77 . While we used country cement production (TTM*1000) 110 as a proxy to species susceptibility to direct threats assuming that in areas with higher cement production species are more prone to habitat loss and direct disturbances 30,37 . We applied Kendall's τ B to test for the ranked-concordance in (i) country-level species richness (log 10), (ii) % endemism, (iii) % threatened species, (iv) percent forest cover (%, log 10) (iii) socioeconomic indicators: GDP per capita (log 10), country cement production (TTM). We mapped and visualised estimated species richness, endemism, and threatened species using QGIS version 3.14.

Cave Biotic Potential
We mapped the global bat cave priorities based on its vulnerability at site-level and biome-levels based on geographical data extracted from the DarkCideS 1. The BPcave index score ranges from 0.00 to 1.00, with values near 1.00 indicates higher cave biotic potential and scaled to four levels of priority.

Mapping landscape features, threats, and cave vulnerability
The computed BP is then complimented with the cave Biotic Vulnerability (BV) to derive the cave alphanumeric priority based on vulnerability. We mapped and measured the

Bat Cave Vulnerability Index (BCVI)
For the purpose and limitation of this work, we aimed to assess the global bat cave vulnerability based on our database and mappable landscape features and threats. We performed BCVI prioritisation in contrasting spatial scales. First, biome-dependent representing the broad-scale priorities, which assumes the priorities are compared relative to all caves surveyed based on biogeographical assessments. Secondly, country-dependent: representing the fine-scale prioritisation, based on the survey-relative priorities that represent, with assumptions that determination of priority should be comparable based on the caves surveyed within the country to guide statutory decision making. The alphanumeric index based on BCVI is then converted into four priority levels ( Table 2). We then compared the mean biotic vulnerability threshold (i.e., values range from 1 to 4, the lower the values the higher the vulnerabilities) versus cave biotic potential status, and then the priorities from both scales across biogeographical realms and climatic biomes using a chi-squared test ( 2 ) of association. We used Pielou's index to assess evenness in BCVI and priorities between scales.  Sum of ED and CWE of individual sampled caves (right). The ED and CWE of sampled cave is supplemented in the Data Paper co-published with this manuscript. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.  Relationship and turnover of threats amongst assessed global cave sites. (A) Correlation coe cients (u and 95% CI intervals) between landscape features and threats and species cave biotic scores, (B) biogeographic comparison of mean intensity of vulnerabilities to landscape features and threats, (C) Global concordance of between country socioecological indicators (country GDP, Forest cover, cement export production) and species attributes (up to down: country species richness, endemism, and proportion of threatened species). Landscape and threat exposure of sampled cave is supplemented in the Data Paper co-published with this manuscript.   Table 5 and individual cave priorities are are supplemented in Data Paper co-published with this manuscript. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.