African bat database: curated data of occurrences, distributions and conservation metrics for sub-Saharan bats

Abstract

Accurate knowledge of species distributions is foundational for effective conservation efforts. Bats are a diverse group of mammals, with important roles in ecosystem functioning. However, our understanding of bats and their ecological importance is hindered by poorly defined ranges, mostly as a result of under-recording. This issue is exacerbated in Africa by the ongoing rapid discovery of new species, both de novo and splits of existing species, and by inaccessibility to museum specimens that are mostly hosted outside of the continent. Here we present the African bat database – a curated set of 17,285 unique locality records of all 266 species of bats from sub-Saharan Africa, vouched for by specimens and/or genetic sequencing, and aligned with current taxonomy. Based on these records, we also present Maxent-based distribution models and calculate the IUCN Red List metrics for Extent of Occurrence and Area of Occupancy. This database and online visualization tool provide an important open-source resource and is expected to significantly advance studies in ecology, and aid in bat conservation.

Background & Summary

Accurate taxonomic assessments and a thorough understanding of species distributions provide the foundation for studying our environment across time and space¹. Species occurrence data are often used to test hypotheses in evolution and ecology, with important applications in assessing the vulnerability of species and ecosystems in changing environments, e.g., through species distribution modelling and IUCN Red List assessments^2,3. However, these efforts are hindered by the fact that many mammal species remain unknown, under-sampled and under-reported, especially in tropical regions such as Africa^4,5. Furthermore, many researchers have highlighted the underrepresentation of sub-Saharan Africa, in particular in global biodiversity research^6,7.

Approximately a quarter of mammalian biodiversity is comprised of bats, with 1,482 species currently recognized globally, of which over a fifth occur in Africa^8,9. Bats provide essential ecosystem functions such as pollination, seed dispersal and insect pest suppression^10–12. Despite their importance, bats remain poorly documented, especially in Africa^13,14. This is partly because voucher specimens are not easily accessible for evaluation by African biologists, as most specimens of African bats are housed in European and North American museums¹⁵. An additional challenge is the need for continual updating of species names in museum collections or databases, arising from the rapid description of new bat species and revision of systematic relationships, driven by advances in genetic and acoustic technologies, and increased sampling^16–21 (Fig. 1). This complicates our understanding of which species are represented in past records, and highlights the shortcomings of existing global datasets of mammalian distributions²² for local and regional studies.

Fig. 1.

Rapid increase of described bat species in sub-Saharan Africa: (A) cumulative increase in the number of new species over time (solid line), and the number of new species per decade (bar plot); (B) the number of species recognized per family, categorized into those recognized before 2005, de novo species recognized since 2005, and those split from existing species since 2005, based on the Mammal Diversity Database⁹.

The lack of accurate distribution maps for many bat species complicates their conservation²³, as knowing where a species occurs is essential for informing where it may be best protected. Robust occurrence records are necessary for calculating crucial conservation metrics, such as the extent of occurrence (EOO) and the area of occupancy (AOO) of a species²⁴. These are key parameters in assessments undertaken by the International Union for Conservation of Nature (IUCN) Red List of Threatened Species to evaluate the conservation status and risk of extinction of a species. Distribution records are also needed for accurate species distribution modelling, increasingly used to present the ranges of animals, and applied to their conservation²⁵. Importantly, these applications depend on accurate records, as poor-quality input data can result in misleading metrics and distribution models, which in turn can significantly impact conservation planning, environmental impact assessments, and subsequent mitigation strategies²⁶.

The proportion of species facing extinction continues to grow, with over a quarter of mammals currently threatened and a third of bats defined as threatened or data deficient by the IUCN²⁷. As the number of species facing elevated extinction risk rises each year²⁸, it becomes increasingly important to accurately assess each species’ conservation status. This underscores the need for comprehensive distributional knowledge of each species. Without this, we cannot compare current distributions with those of the past and/or predicted future, to detect declines and changes that provide warnings of how different drivers are altering the structure and functioning of biodiversity over time^29–31.

Here, we present a comprehensive database of bat occurrence records for species in sub-Saharan Africa, with each record verified by voucher specimens and/or genetic data³². This database builds on the 6,042 records of southern and central African bats in Monadjem et al.¹³, to which we have added over 11,200 additional records. We present unique locality records and models of predicted suitable habitat for each species of bat known from this region. We then calculate two distributional metrics that can be used in future Red List assessments for African bats as they become outdated and for species still requiring initial assessments. Our aim is to facilitate and expedite this process not only with the newly calculated metrics but also by presenting a standardised, replicable method for calculating these metrics in the future. Furthermore, this database complements the recently published African bat traits (AfroBaT) dataset³³. The data are released as open source and presented via an R Shiny app tool that allows easy visualization of the distribution records of any selected species, all made available for use by the wider research community.

Methods

Taxonomic and geographic coverage

Our dataset includes occurrence records of 266 bat species from 53 genera and 12 families (order Chiroptera, class Mammalia) from sub-Saharan Africa. We define the region as all the countries of Africa partially or completely south of the Sahara, excluding oceanic islands such as Madagascar, the Comoros, and Seychelles³⁴, but including Bioko, Zanzibar and Mafia islands, as these lie close to the main continent and have shared bat communities with the mainland^35,36. Although our database only includes species that occur in sub-Saharan Africa, we include records for these species from across the continent, including North Africa (north of the Sahara). We include the year of collection for 10,960 records (63%), with most records collected in the decades since 1970 (Figure S1). We included records up until 28 April 2024. We follow the taxonomy of Simmons & Cirranello⁸, and use it along with the Mammal Diversity database⁹, to quantify the number of species in sub-Saharan Africa over time, and the number of species now recognized by family (Fig. 1). The African bat database includes 16 taxa for which a specific species identity cannot be claimed. These species are included in the database with species epithets beginning with “cf.” followed by the name of the taxon closest in morphology to it, but from which it is distinct. The inclusion of these unnamed species allows for the recognition of morphologically similar taxa that require further investigation for definitive identification and aims to enhance the comprehensiveness of the dataset while acknowledging taxonomic uncertainties.

Data categories

All records are provided as individual-level data, with taxonomic information including family, genus and species. Each record is accompanied by a location (with latitude and longitude coordinates in decimal degrees), name of the country and location, and data source or reference. Other columns in the database include: i) date of collection (41% of records); and ii) museum or collection number (66% of the records have a museum assigned to them). We indicate the specimens that we have inspected ourselves (30%), and we also indicate whether the specimen is a holotype (247 specimens) for currently recognized species only. We attempted to apply the most recent nomenclature to each record, which we were able to do for most specimens. However, in a few cases this was not possible, and we referred to them as cf. species (16 taxa). These fell into two groups. On the one hand, these include taxa with deeply divergent lineages whose taxonomy has not yet been resolved, such as Hipposideros cf. ruber^37,38 and Miniopterus cf. inflatus³⁹. On the other hand, these include specimens that we were unable to inspect and for which no genetics existed. Furthermore, these latter specimens were far beyond the known range of the species. We could have removed these records, but we felt that keeping this data might be useful for certain future analyses.

Data sources

The African bat database was meticulously compiled by considering only records with confirmed species identification that could be interrogated against current taxonomic knowledge. All specimens included in this database are vouched for by one of the following: a museum specimen, a genetically sequenced tissue sample, or a photograph. A huge effort was undertaken to catalogue museum specimens, which were examined by the authors (over 5,100 specimens), alongside other published literature records of museum specimens, as well as unpublished data of the authors (300 records). Records of bats that were captured and released were included only when tissue samples were taken and genetically sequenced to confirm species. Photographs were used very scarcely to confirm ID and only in cases where the identification was unquestionable (e.g., for unmistakable species like Myotis welwitschii, Taphozous mauritianus or Glauconycteris variegata).

Species distribution models

The theory and practice of species distribution models (SDMs) have advanced rapidly in the past two decades, with numerous approaches developed in recent years^25,40. However, Maxent has remained consistently popular and robust for species distribution modelling and can be easily implemented in R software^41,42. Furthermore, Maxent models have performed well even when compared with ensemble modeling⁴³, and have been recommended as the method of choice for certain situations, particularly when the data are presence-only⁴⁴. As Maxent has been used previously to develop SDMs for African bats^45,46, we chose to keep the methodology consistent for this database.

We modelled the predicted suitable habitat space of each bat species that had seven or more locality records within Africa under present climatic conditions using Maxent⁴¹. Models were run at a resolution of 2.5 arc min (approximately 5 km) using BIOCLIM variables (Table S1) and elevation from the WorldClim database⁴⁷. To deal with multicollinearity between variables, we used the Variance Inflation Factor (VIF) to select a set of uncorrelated environmental variables³⁴. We calculated the VIF for the 20 variables and followed a backward model selection process to determine the minimum adequate model removing the variables with the highest VIF one at a time, until we were left with 10 variables that had VIF < 10: elevation, BIO02, BIO03, BIO08, BIO09, BIO13, BIO14, BIO15, BIO18 and BIO19 (see Table S1 for definitions).

We ran Maxent models in R version 4.1.2⁴⁸ using the package dismo⁴², following the procedure outlined in Monadjem et al.⁴⁹. We used hinged and categorical variables that smooth variable responses and generally improve model performance^50,51. We randomly divided bat species occurrence records from southern Africa into training (75%) and test (25%) datasets. The selection of the geographical background has important implications for the results of species distribution models^52,53; a suitable background reflects the geographical space available to the species by dispersal⁵⁴. Therefore, for each species, we randomly sampled 10,000 background points from 100 km circular buffers around all occurrence points for that species. We used the value of 100 km because this is the distance that Nycteris thebaica (a particularly sedentary, clutter-foraging bat species) is able to cover during dispersal⁵⁵, and therefore, this buffered range represents the minimum area available to any of the species we included in our analyses⁵¹. We tested each model with the area under the receiver operating characteristic curve (AUC) statistic, which ranges from 0 to 1 with higher values signifying a better fit⁵¹. Values equal to or less than 0.5 indicate models no better than random, while values greater than 0.75 represent good model fit⁵⁶. We used the same 12 environmental variables (Table S1) and Maxent parameters for all species⁴⁵. We converted the predicted model outputs from Maxent (i.e., probabilities of suitability) into “presence-absence” maps using species-specific thresholds that maximized the sum of sensitivity and specificity, which is appropriate for presence-only data⁵⁷.

Calculating EOO and AOO

Extent of occurrence (EOO) is defined as “the area contained within the continuous imaginary boundary which can be drawn to encompass all the known, inferred or projected sites of present occurrence of a taxon, excluding cases of vagrancy”⁵⁸, and is calculated as the minimum convex polygon around all the known localities of a species^24,59. We employed the same procedure to calculate EOO of African bat species, excluding any species with less than 3 locality records. For each species, we calculated a minimum convex polygon using the function mcp() in the package changeRangeR⁶⁰.

Area of occupancy (AOO), in contrast to EOO, “represents the area of suitable habitat currently occupied by the taxon”⁵⁸, and can be calculated by measuring the 2 × 2 km cells known or predicted to be occupied by the species⁵⁹. To calculate AOO, we clipped the predicted species distribution models (i.e., outputs from the Maxent analysis described above) with the EOO so as to avoid including areas outside the known range of the species⁶¹. Finally, we also calculated minimum and maximum elevation from the predicted Maxent model (clipped to AOO), based on the BIOCLIM digital elevation model (see above). For species without SDMs (i.e., species with less than seven occurrence records, see above), we calculated AOO by multiplying the number of localities by 4 (2 × 2) km².

Data Records

The African bat database and SDMs are available on figshare (https://figshare.com/articles/dataset/African_Bat_Database/26363308)³². The distribution of each bat species, based on our database, can be visualized through the R Shiny App (https://adam-kane.shinyapps.io/african_bat_dist/). The metadata associated with the African bat database is as follows. Family: family of bat record; Genus: genus of bat record; Species: species of bat record; Museum_number: museum accession number of bat record (see Table S2 for definition of the museum acronyms); Date: date on which the bat was collected or otherwise recorded; Year: year in which the bat was recorded; Country: country in which the bat was recorded; Location: name of the location or locality that the bat was recorded; Latitude: latitude in decimal degrees of the record; Longitude: longitude in decimal degrees of the record; Reference: source of the record (see Appendix 1 for the references); Holotype: whether the record is a type specimen; Checked: whether the bat specimen was examined by the authors.

Data coverage

The database contains 17,285 unique locality records of all of the 266 species of bats currently recognized from sub-Saharan Africa (Fig. 2). The number of species recognized in this region has increased by 60% in the last 100 years, and 17% since 2000 (Fig. 1A). In recent years, the increase in species numbers was primarily due to new species being discovered (both de novo descriptions and epithet splits) in the families Vespertilionidae, Rhinolophidae, Pteropodidae and Miniopteridae (Fig. 1B). Species distribution models were built for the 209 species that had seven or more locality records. Species richness based on these modelled distributions is shown in Fig. 3. EOO was calculated for 242 species recorded from sub-Saharan Africa, with three or more occurrence records (Table S3), and ranged from 27 km² in Rhinolophus hilli to 35,126,201 km² in Nycteris thebaica. AOO was calculated for all 266 species (Table S3) and ranged from 4 km² in several species known from just the type locality to 5,344,996 km² in Taphozous mauritianus. Minimum and maximum elevation of all records ranged from sea level up to 4,207 m above sea level (Table S3), with 33 species having been recorded above 3,000 m above sea level. The species with the largest elevational range was Afronycteris nana, which was recorded at sea level and up to 4,207 m above sea level.

Fig. 2.

Map showing the localities at which the 17,285 occurrences of 266 species in the African bat database were obtained from.

Fig. 3.

Species richness of bat species in sub-Saharan Africa based on the 209 SDMs created in this study.

Limitations

We acknowledge that there are still significant data gaps, especially for under-recorded species and regions that require more sampling efforts; which applies to most of Africa²³. This is clearly seen in the 24 species for which there were insufficient occurrence records (<3 records) to calculate EOO (Table S3). These included species only known from the type locality, e.g., Mops gallagheri, Glauconycteris kenyacola, G. machadoi, Myotis nimbaensis, Pipistrellus permixtus, and Pseudoromicia kityoi. African regions that are clearly under-sampled for bats include the Congo Basin (mostly within the Democratic Republic of Congo), the Lower Guinea Forest (especially Equatorial Guinea, Gabon and the Republic of Congo), the Guinean and Sudanian savannas of Central African Republic, Chad and South Sudan, and the Horn of Africa (Ethiopia, Eritrea and Somalia). However, even within relatively well surveyed regions such as South Africa and Eswatini¹³, new species are still being discovered and described¹⁹, highlighting the urgent need for more sampling across the continent.

Additionally, new species continue to be described (Fig. 1), and some taxa remain unresolved. Perhaps the best example here is the Hipposideros cf. ruber complex, which has at least five distinct lineages³⁸ that may represent separate species, but it still awaits taxonomic revision. The lack of specimen collection dates for a large portion of the database is also undesirable, however, this is somewhat ameliorated by the fact that almost two-thirds of the specimens have the year of collection. Where needed, missing dates will have to be obtained from the museums where the specimens have been deposited.

Accurate species identification is increasingly reliant on genetic data, especially to distinguish cryptic species¹⁹, and these data are generally available online (e.g., on Genbank). We acknowledge that incorrect identifications or outdated taxonomy in the specimens represented in these databases can result in inaccuracies. While we do not provide genetic data in the current version of the African bat database, this would strengthen future versions by improving traceability across records and genetic reference material.

Technical Validation

All data were entered manually and were cross-checked by multiple authors. A series of checks were undertaken to examine the validity of the data, and the accuracy of the location metadata. We examined the distributions of each species by plotting all the occurrence points and then carefully identifying possible outliers, or incorrect records plotting over water. For records that were marginally plotted over water (within 2 km of the shoreline), we slightly adjusted the latitude or longitude if these were not based on Global Positioning Systems (i.e., records based on quarter-degree squares). Records plotting obviously outside the core of a species range were also checked to ensure that they were not mapping errors. We encourage users to notify the corresponding author of any errors or additional records to ensure the dataset remains accurate and up to date. All updates will be subject to the same stringent technical validation process outlined previously.

Usage Notes

The African bat database³² is the first accessible dataset of occurrence records, distribution models and conservation metrics for all bat species in sub-Saharan Africa where all the records have been carefully examined and corrected for current taxonomy. This database has been a long time in the making and will be crucial in resolving the distributions of poorly recorded bat taxa in sub-Saharan Africa.

Existing databases, such as iNaturalist, GBIF, and other citizen science platforms, have severely limited accuracy in bat records because many bats cannot be identified based solely on appearance. Most published resources on African bat distributions present range maps^62,63 rather than point localities, and have out-of-date records and taxonomy.

To the best of our knowledge, the African Chiroptera Report¹⁵ is the only published resource that contains point localities for all African bats. However, because this report is a document rather than a database, the underlying data is not readily available for analyses, and inaccuracies frequently persist over time. Distributional mistakes and outliers (for example, in the distribution of Stenonycteris lanosus, which is restricted to the highlands of East Africa⁶³, but yet has records in the report in central and western Democratic Republic of Congo) have an impact on any future analysis and may have a negative effect on conservation planning. Our versioned digital dataset aims to overcome these issues by offering regularly updated, curated data with accurate point localities, and providing a reliable, accessible resource for research and conservation planning.

We hope this comprehensive digital dataset will integrate previously disconnected data resources, and bring African bats into focus for novel and integrative applications that have historically been limited to other regions and taxonomic groups⁶⁴. It may be useful in ecological and evolutionary research, for species distribution modeling, macroecology, biogeography, and community ecology studies^65–67. Ideally, it will help inform biodiversity monitoring and climate change adaptation policy, by predicting the responses of species and ecosystems to increasing threats and changing environments, including habitat destruction and climate change. For taxonomy and systematics, our dataset supports species identification and phylogenetic studies, highlights unresolved taxa, and provides a tool for continuously updating records. In conservation planning and management, we expect that our database will aid both regional and global IUCN Red List assessments by providing systematic and replicable calculations of distribution metrics based on records in the database, which in turn may help inform and facilitate protected area designation as well as environmental impact assessments. Moreover, the online R Shiny App visualization tool allows for interactive exploration of bat distributions, which we hope will be used for educational and outreach efforts, and to support citizen science initiatives.

The database is intended to be a dynamic, living dataset that will be updated regularly, at least once a year, or as new data becomes available, and errors are identified and corrected. Future updates will welcome new or previously unavailable data, will incorporate taxonomic classification changes, and will include the distribution of newly resolved and described species. Updates of the dataset will be made directly on Github (https://github.com/kanead/Bat_database) and will be reflected in the associated R Shiny App. All updates will follow the same technical validation described in this manuscript.

Author contributions

A.M. conceived the idea. A.M. created and populated the database with contributions from all coauthors. A.M., C.M., P.J.T. and A.K. performed the analysis and wrote the first draft. All coauthors read and revised drafts of the manuscript.

Code availability

Two R scripts are available on Figshare (African Bat Database)³². One generates the species distribution models based on bioclim data (Historical climate data — WorldClim 1 documentation). The other generates area of occurrence (AOO) and extent of occurrence (EOO) information.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41597-024-04170-7.