Mountains host significantly more data deficient and threatened bat species than lowlands

To calculate and map the species richness of bats on mountains, we downloaded all available distribution polygons of bat species from the IUCN Red List of Threatened Species database (IUCN 2022). As of 16th January 2023, the database contained 1331 bat species of which distribution polygons were available for 1315 species. Over 1300 of these species were assessed after 2014 (Fig. S1). Our data, therefore, represent almost 90% of the current known global bat diversity (Simmons and Cirranello 2022). We rasterised all bat distribution polygons at a fine-scale resolution of 1/6° × 1/6° and calculated species richness in each grid as a sum of overlapping polygons. We chose this resolution as many mountains are localised areas of high relief and do not span vast degrees of latitude. This analysis was performed using the ‘fasterize()’ function of the ‘raster’ package (Hijmans 2022) in R. We defined mountains following Körner et al. (2017) and the polygons provided by them were accessed via: https://​ilias.​unibe.​ch/​goto_​ilias3_​unibe_​cat_​1000515.​html.

We also used the bat distribution polygons and the mountain polygons to calculate the proportion of each bat species’ distribution range that falls within mountains. To do this, we first calculated the total area of the distribution range for each bat species. We then overlaid the mountain polygons over the distribution polygons of bats and calculated the areas of intersection between each species and each mountain. Many species’ distribution ranges overlapped with multiple mountain polygons, thereby generating multiple areas of intersections. Therefore, the areas of intersection of each species were summed to calculate the total area of its distribution that falls within mountains. The intersections were marked using the ‘st_intersection()’ function and all areas were calculated using the ‘st_area()’ function of the ‘sf’ R package (Pebesma 2018). We observed that the distribution polygons of 29 species were either improperly marked or projected. Additionally, their elevational range limits reported on the IUCN website suggested that they occur only in mountains but the distribution polygons did not intersect mountain polygons. We treated them as non-mountain dwelling species in further analyses. We acknowledge that the converse is also possible with IUCN data i.e. that distribution polygons may encompass mountains when the species may, in fact, be found in the lowland surrounding a mountain.

Species that have at least 50% of their distribution range in mountains may be defined as species that predominantly occur on mountains. For a stricter cutoff, we defined a species as ‘mountain dwelling’ when ≥ 75% of its distribution range was on mountains. Another study (Rahbek et al. 2019) used similar cutoffs to categorise the mountain affinity of different vertebrate taxa. To test how sensitive our analysis and findings were with respect to deviations from the 75% cutoff value, we conducted a sensitivity analysis using cutoff values of 50, 60, 70, and 80%.

We defined a ‘highland dwelling species’ as one that occurs exclusively in upper montane, subalpine and alpine habitats. These habitat zones are often zones of high endemism (Fjeldså 1993; Noroozi et al. 2018; Sonne et al. 2022; Mendelsohn et al. 2023) but vary in elevation in different mountains. For example, the highlands of Fiji reach up to 1300 m above sea level (a.s.l), and have an upper montane region which would not be considered the same as the upper montane region of the Himalaya. Therefore, we refrained from using an elevation cutoff to define highland species. To classify species as ‘highland dwelling’, we looked for phrases/words such as “highland forest dependent”, “endemic to highland areas”, “montane or cloud forest”, or “coniferous forest dwelling” (for tropical species) in the geographic range description in the IUCN species pages. We acknowledge that this approach depends on textual accuracy. Additionally, the improper distribution projections of 29 species mentioned above meant that some of those species were not classified as mountain dwelling (based on spatial overlap between the species distribution and mountains) but were classified as highland dwelling (based on textual descriptions). This is a caveat of our approach but we believe that going by textual descriptions of mountain habitats would perform better at identifying highland species at a global scale than a uniform global elevational cutoff.

Following the definition of IUCN, we considered a species as ‘threatened’ when it belonged to Vulnerable (VU), Endangered (EN) or Critically Endangered (CR) Red List categories (https://​www.​iucnredlist.​org/​documents/​RedListGuideline​s.​pdf). Species categorised as Least Concern (LC) and Near-Threatened (NT) were considered as non-threatened. Using univariate generalised linear models (GLM), we assessed if the probability of being data deficient or threatened is higher for mountain dwelling vs. other species; and whether the probability of being data deficient or threatened is higher for highland dwelling species vs. other species. The GLMs included a binary response variable (whether a species was data deficient/threatened or not”) and a binary explanatory variable (whether a species was mountain dwelling/highland dwelling or not).

We tested if body mass (as a proxy of body size), diet, and biogeographic realm are associated with a bat being mountain dwelling. For this, we performed a phylogenetic generalised linear model with binomial distribution based on a pruned phylogenetic supertree (Upham et al. 2019). Whether a bat species is mountain dwelling or not was considered a binary response variable and body mass (continuous variable), diet, and biogeographic realm (categorical variables) were used as explanatory variables. We performed the regression using the ‘phyloglm’ function of ‘phylolm’ package (Tung Ho and Ané 2014).

Body mass and diet were obtained from COalesced Mammal dataBase of INtrinsic and Extrinsic traits (COMBINE) (Soria et al. 2021). COMBINE also presents data on forearm length which is a standard measure of body size in bats. However, body mass had fewer missing values and was correlated with forearm (Pearson’s r = 0.9, p < 0.001), so we chose body mass as our proxy for body size. Body mass is also correlated to wing loading (Norberg and Rayner 1987) which is an important trait determining dispersal ability and colonisation potential (Norberg and Rayner 1987; Furey and Racey 2016; Chakravarty et al. 2018). Additionally, larger bats are believed to be better adapted to cool mountain environments (de Carvalho et al. 2019). Wingspan would have been a better predictor for dispersal ability and colonisation but those data are absent from COMBINE. The diet data presented in COMBINE come from PHYLACINE 1.2 database (Faurby et al. 2018). PHYLACINE 1.2 presents diet as a proportion of plant matter, vertebrate prey, and invertebrate prey consumed by a mammal species. We assigned the ‘main diet’ for each bat species (plant, vertebrate or invertebrate) using the category with the highest proportion. Diet is an important predictor of a species’ niche which includes its geographic range (Rodríguez-Castañeda et al. 2010; Slatyer et al. 2013), therefore diet may be associated with a species being mountain dwelling or not. After removing extinct species and matching the taxonomies in IUCN 2022 and COMBINE databases (details in Table S1), we retrieved diet data for 1141 species (i.e. 190 species without data) and for all but four genera. At such broad categorisation, the main diet was observed to be phylogenetically conserved within genera when visualised by plotting bar plots of diet per genus (Fig. S2). Therefore, we replaced missing data with the main diet of congeners. We assigned the monophyletic species Dryadonycteris capixaba to plant-feeding following Nogueira et al. (2012) whereas Setirostris eleryi and Submyotodon latirostris were assigned to insect-feeding as they split from the insectivorous genera Mormoopterus and Myotis respectively (Reardon et al. 2014; Ruedi et al. 2021). Only the recently-described Eudiscoderma thongareeae could not be assigned to any specific diet. The missing species and the diets that they were assigned to are listed in Table S2. We removed Hipposideros gentilis and Nyctimene wrightae as they were not recognised in the COMBINE dataset. We only had three sanguivorous species (0.02% of the total species) and seven carnivorous species (0.05% of the total species) and therefore removed these from subsequent analyses (as keeping them would increase the number of degrees of freedom in the respective model). We assigned species to biogeographic realms following Udvardy (1975). A hundred and eight species (8.5%) spanned multiple biogeographic realms and were assigned to the realm in which majority of their distribution polygon lay based on visual assessment. After cleaning up the data, the final model included 1165 species that were represented on the phylogeny and had no missing trait values.

The most species-rich hotspots for bat diversity in mountains were found to be in parts of the northern Andes and the Guiana Highlands. These estimates are congruent with previous projections for mammal or bat diversities (Jetz et al. 2012; Jenkins et al. 2013; Alves et al. 2018). In fact, comparing the species richness of bats and other major mammalian orders suggests that bats have a disproportionate effect in increasing the overall mammalian diversity in northern South America (Jetz et al. 2012; Jenkins et al. 2013). In the Palaearctic realm, the Caucasus, Zagros, and Elburz Mountains have 10–15 more mammalian species than the surrounding lowlands (Jetz et al. 2012, accessed from https://​mol.​org/​patterns/​richnessrarity?​taxa=​mammals). Our species richness estimates for bats suggest that this increase is also disproportionately influenced by the region’s bat diversity, especially in the Caucasus. The same effect is seen, although to a lesser extent, in mammalian species richness in the Himalaya and the lowlands to its south, and Tibetan Plateau to its north (Jetz et al. 2012, accessed from https://​mol.​org/​patterns/​richnessrarity?​taxa=​mammals). In the western Himalaya, roughly 40–50% of the mammalian species are bats (Jetz et al. 2012, accessed from https://​mol.​org/​patterns/​richnessrarity?​taxa=​mammals; Chakravarty et al. 2020). However, it must be noted that the high species richness in mountains in comparison to surrounding lowlands could be an artefact of mapping inaccuracies causing lowlands and mountains to fall within the same grid.

Apart from highlighting areas of high diversity, our data also underscore abrupt discontinuities in species richness patterns. Two of these are worth mentioning here: the decline in species diversity as one goes from west to east in (a) the Himalaya, and (b) the New Guinea Highlands. For most well-studied taxa, species richness increases from west to east in the Himalaya (Srinivasan et al. 2014; Tamma and Ramakrishnan 2015; Tamma et al. 2016). Eastern Himalaya and Northeast India are also among the most species-rich biodiversity hotspots in the world (Myers et al. 2000). In the case of New Guinea, the decline in bat diversity coincides precisely with the political boundaries between Indonesia and Papua New Guinea. The same pattern is seen even when visualising species richness of the entire island (and not just mountains as in our analysis) (Jenkins et al. 2013). Both the above examples likely point towards strong sampling biases i.e., the Eastern Himalaya of India and the Indonesian New Guinea are poorly sampled for bats. The strong separation at the political boundary of Indonesia and Papua New Guinea may also reflect differences in research interest or investment in sampling bat species in the respective countries.

An important result of our study is that mountain dwelling species are more likely to be data deficient than species that are predominantly found in the lowlands. Data deficient species are normally excluded from conservation prioritisation and funding schemes because of their uncertain status (Bland et al. 2017). Yet, data deficiency is far more common in elusive taxa like bats in comparison to other mammals (Frick et al. 2020). We now demonstrate that a greater proportion of these species are found on mountains. Recent predictions unequivocally demonstrate that data deficient species are more likely to be threatened (Bland et al. 2015; Borgelt et al. 2022) which makes our findings even more worthy of attention. Moreover, 28 of the 148 mountain dwelling species have been identified as high-priority island endemic species for research and conservation, including two species (Neopteryx frosti, and Hesperoptenus gaskelli) that rank in the top-10 (Conenna et al. 2017).

There are eight main reasons why species are assessed as data deficient: uncertain provenance, type series, less than five records, records from before 1970, uncertain population status or distribution, uncertain threats, new species (discovered in the last 10 years), and taxonomic uncertainty (Bland et al. 2017). Each of these justifications applies to the data deficient mountain and highland dwelling species (next subsection) in our dataset. Fifty seven percent of all bat species—including those that are classified as ‘least concern’—have unknown population trends (compared to 39% mammals and 8% of birds) (Frick et al. 2020). Additionally, over 270 bat species have been described since 2005 (Frick et al. 2020), a statistic that continues to grow year on year (Saikia et al. 2022; Garbino et al. 2024; Fonseca et al. 2024; Patterson et al. 2024). Of the 1482 currently recognised bat species (Simmons and Cirranello 2022), over 150 have not even been evaluated by the IUCN. The number of data deficient species, although very high, is still an underestimate.

Mountain environments pose several challenges in sampling wildlife such as lack of accessibility, inclement and uncertain weather conditions, and short breeding seasons. Consequently, even observable vertebrates like birds are much more poorly known in montane and alpine habitats (Scridel et al. 2018). The lack of sufficient records and uncertain population trends are widely recognised as hampering accurate IUCN assessments, yet ironically, in the current research landscape, funding is not easily available for taxonomic inventorying and long-term monitoring (Anderson 2017). These problems make it all the more important to gather data on rare and elusive taxa. As the threat of land use change and climate change intensifies, data deficient species could go extinct without notice (Howard and Bickford 2014; Borgelt et al. 2022).

We found 46 bat species across biogeographic realms that occur close to mountain tops, with the Oriental realm hosting the highest diversity. Alarmingly, we also found that these highland dwelling species have a higher probability of being threatened that those that are found in lowlands or across elevations. Irrespective of geography and taxon, species dwelling in high elevations are under severe threat. For instance, long-term demographic data linked with local climate data have identified a consistent annual population decline in the Haleakalā silversword—a plant endemic to a single volcano summit in Hawaii (Krushelnycky et al. 2013). Another example using simulation models projected range contractions for range-restricted highland dwelling aquatic invertebrates and vertebrates under future climate change scenarios in North America and the Himalaya (Muhlfeld et al. 2011; Sharma et al. 2021b). The impacts of climate change on alpine regions are also a major factor driving population declines of many taxa, including mountain breeding birds in Europe (Lehikoinen et al. 2014, 2019).

Many of the highland species in our dataset are categorised as threatened on account of their small distribution areas, as population trends for bat species are largely unknown (Frick et al. 2020). Most of these species are known from single locations or a few disjunct locations. The impact of disjunct distributions across valleys has already been shown to be drastic in a long-eared bat species Plecotus balensis. Specifically, Plecotus balensis occurs on sky islands in a few summits of the Ethiopian Highlands. Local populations show marked genetic differentiation across valleys and the species’ range has been progressively shrinking since the last glaciation period. The problem is likely to be accentuated by the effects of climate and land-use change (Razgour et al. 2021). Three other bat species (Alionycteris paucidentata, Mirimiri acrodonta, and Pteralopex pulchra) are restricted to mountain tops within isolated small-sized islands. Both Mirimiri acrodonta and Pteralopex pulchra are already critically endangered. Additionally, five highland dwelling species (Anoura fistulata, Nyctophilus microdon, Pipistrellus collinus, Sturnira aratathomasi, and Sturnira bidens) are strictly associated with cloud forests (Karger et al. 2021). Although these species are data deficient or least concern, their restricted elevational range warrants additional surveys and periodic monitoring. It must be noted that these species are found in regions that have not been extensively surveyed and their actual geographical distributions may be larger than currently known.

In terms of the biogeographic relevance of our results, we found that four out of the six mountain dwelling species of the genus Plecotus are highland dwelling (Fig. 3; Online Appendix Table S1). It appears that Palaearctic species of this genus show relatively broader elevational distributions but tropical and subtropical species occur in narrower elevational ranges. A recent record of Plecotus homochrous comes from 2200 m above sea level in Hoang Lien National Park in northern Vietnam (Fukui et al. 2020). This record was published following the IUCN update of this species in 2019 (Srinivasulu and Srinivasulu 2019). This is the first record of this erstwhile Himalayan endemic from Southeast Asia. Interestingly, even in Southeast Asia, it was caught in a high elevation site. Assuming that low elevation sites are better sampled, this observation suggests that P. homochrous has a wide geographical distribution but a restricted elevational distribution. As the genus Plecotus predominantly has a Palaearctic distribution, the tropical species of the genus are likely to be cold-adapted. Species occurring in subtropical and tropical mountains may therefore occupy higher elevations, as is also seen in some Palaearctic birds, mammals, and arthropods in the Himalaya (Martens 1984; Päckert et al. 2012). These observations, however, should be treated as preliminary. IUCN distribution maps are not ideal for macroecological research as they adopt a conservative approach in depicting the complete geographical range of species and are replete with sampling biases (Herkt et al. 2017). A global phylogeny of the genus and research on phylogeography will be helpful in examining the reasons behind the mountain and highland affinity.

In total, 1829 species of mammals have > 75% of their distribution range in mountains (Rahbek et al. 2019), so mountain dwelling bats (148, based on our results) only make up 8% of this diversity. Our analysis revealed a significant association between biogeographic realm with the mountain dwelling nature of bat species. Specifically, our analysis suggests that species belonging to the Neotropical, Australasian, and Oriental realms (in ascending order) are more likely to be mountain dwelling.

The association between mountain dwelling and biogeographic realm is complex. The probability of a bat being mountain dwelling was highest in the Oriental realm. In the Oriental realm, the genus Murina has a disproportionately high number of mountain dwelling species (16 out of 32 species). In fact, new species are constantly discovered in this genus from mountainous regions (for example, Ruedi et al. 2012; Son et al. 2015; Soisook et al. 2017) perhaps further exaggerating the data deficiency of mountain dwelling bat species (see above). Murina may be an interesting genus for a phylogeographic study targeted at understanding its affinity to mountains. However, in this realm the high likelihood of being mountain dwelling is possibly because a major part of the realm, especially in the biodiverse Southeast Asia, is covered in mountainous terrain (Fig. 1). The converse is true for the Palaearctic and Afrotropical realms which—in proportion to its area—are excessively devoid of mountainous terrain. Neotropical mountains, such as the Andes and Guiana Highlands are among the most biodiverse mountain ranges including for bats (see above). In this realm, species of the genera Sturnira and Anoura especially have a disproportionately high proportion of mountain dwelling species. It has been phylogenetically shown that Anoura has a montane origin (Calderón-Acevedo et al. 2022). The mountain dwelling species of the Australasian realm are spread across different genera and predominantly occur on Sulawesi and New Guinea. However, as explained above, this analysis must be considered a preliminary exploration which hopes to generate questions and motivates future research.

Our study highlights that bat species found in mountains are more data deficient and threatened than lowland species. We therefore call for targeted surveys of bats on mountains (and more funding for them), especially in the Global South which is generally understudied and where most of the data deficient or threatened mountain and highland dwelling bat species occur. Recent bat species discoveries on Mount Nimba in West Africa boldly underscore the need for such surveys (Simmons et al. 2021; Grunwald et al. 2023). Long-term monitoring schemes must be initiated to improve our knowledge of the elevational distributions and population trends of mountain and highland dwelling bat species. It is also advisable to initiate a dedicated working group on mountain bat ecology within the IUCN Bat Specialist Group or Global Union of Bat Diversity Networks (GBatNet). Lastly, we call for better reporting of elevational distribution in IUCN assessments and in the Map of Life database (https://​mol.​org/​). Elevation data were missing for 533 species and when they were reported it was not uncommon for the lowest elevation to not be reported (or to be misreported as ‘0’) even when the geographical distribution description mentioned phrases like “lowlands” or “only reported from 2600 m”. Lastly, we hope that our study would draw attention to mountain biodiversity in general (and mountain dwelling bats in particular) in the same vein as island biodiversity given the absence of sufficient data and the looming threats posed by climate change.