The CAB was developed to represent rich, species-level data about Afrotropical bees found in literature. The basic unit of information is a taxon citation, meaning that the CAB is a page-level index of scientific literature on Afrotropical bees, most of which is the output of 260 years of taxonomic research. The CAB is updated regularly and will be useful for future taxonomic research, studies in pollination ecology in the Afrotropical Region or in any research that depends on bee taxonomy.

An important purpose of data published on the GBIF Data Portal from the CAB is to validate bee names and integrate specimen-records from distributed databases (e.g. to relate junior synonyms to senior synonyms). Integrating online data is a major challenge in biodiversity informatics (Jetz et al. 2012).

Specific use-cases of the online CAB, which make it a basic starting point for bee research in any context, include: finding the valid name corresponding to any junior synonym or understanding the nomenclatorial history of any species; generating a list of countries in which a bee species has been recorded or a list of bee species from a given country; generating a list of pages in research articles on which a bee name is cited; finding the type depositories and details of the type designation of bee type specimens; finding the host-plants, nest-plants, parasites or hosts of bee species for which this information was found in literature; finding the common names of Afrotropical bee genera, which were recently added to the CAB.

We queried the CAB to analyse and summarise the number of bee species recorded from each country in the Afrotropical Region, mainly to test whether knowledge of the country-scale distribution of bees is lacking. Data from the CAB were supplemented with specimen-records downloaded from the GBIF Data Portal.

We summarised the main periods of research in Afrotropical bee taxonomy and the taxonomic contributions of the main contributors, to try to explain the knowledge gaps we found in the observed country-scale distribution of bees.

We also summarised knowledge of bee type depositories.

In biodiversity informatics our objectives were:

1. To promote awareness of the CAB in the scientific community and explain its development and potential uses;

2. To promote the general need to mobilise (publish or share) high-quality, standardised species-level data, including rich taxonomic data and rich ecological data such as host relationships, which can potentially be used in ecological studies (e.g. pollination studies).

A significant finding in developing and using the CAB (Figure 1) is that we were able to represent meaningful units of taxonomic research, specifically a citation of a taxon on a page and the specific taxonomic meaning of this citation (e.g. whether it was of nomenclatorial importance and what kind of nomenclatorial act it signified). It was also significant that we were able to customise a Specify database for the purpose of creating a richly annotated checklist of species for publication, both as standardised online data as well as a textual report for conventional publication. The advantage of the textual report is that it can be formatted to show the nomenclatorial history of the species, i.e. to update the checklist of Eardley and Urban 2010 at the click of a button.

By representing nomenclatorial acts (e.g. an original description or a new combination) and taxonomic acts (e.g. a synonymy or a change in taxonomic rank) – as records in the Specify taxoncitation table – we were able to execute a query to summarise all the taxonomic work that has been done since the previous publication of the CAB almost a decade ago (Table 1). A list of species descriptions published since 2010 is given in Appendix 1 and a list of species synonymised or removed from synonymy since 2010 is given in Appendix 2. Taxonomists should find this summarising capability of such a database design useful in their work, especially in compiling a regional, richly annotated checklist of species.

Of the 1.03 million GBIF specimen-records of bees tagged with country names that we downloaded, only 39 176 (3.8%) were from countries in the Afrotropical Region (Table 2). Specimen-records downloaded from the GBIF Data Portal therefore made a small contribution to the dataset that we used in the analysis presented below.

The analysis of country-scale distribution presented below is for valid species-level names only (i.e. not including junior synonyms or subspecies). We included species described up to and including 2018.

Records of bees other than Apis mellifera have been reported from 45 of the 47 African countries or territories constituting the Afrotropical Region as defined by Crosskey and White (1977) (i.e. not from Mayotte nor Guinea-Bissau). The CAB reported 2 783 valid species from the Afrotropical Region. The following introduced species were included in the analysis: Hylaeus perhumilis (South Africa), Megachile semivestita (Zanzibar) and Megachile multidens (Democratic Republic of Congo).

The countries of origin of the following 7 valid species remain unknown and they were therefore not included in the country counts: Thyreus calcaratus, Xylocopa ditypa, Halictus frontalis, Lipotriches quartinae, Megachile mastrucatella, Megachile mediana and Heriades glomerans.

i) Species-per-country

The 2 776 species with country records were expanded to 7 196 species-per-country combinations. The number of species-per-country records (a potential indication of species richness) ranged from 4 (São Tomé and Príncipe) to 1258 (South Africa, 65 of which were contributed by the GBIF dataset) (Table 3).The GBIF specimen-records added 382 new species-per-country records (combinations of 242 species and 33 countries – see Appendix 3), representing 5% of the data overall. The GBIF country records that agreed with those from the CAB numbered 1 312 or 19% of the total species-per-country records, which again underscores the importance of the CAB country-scale distribution data.

Thirty-four of the 45 countries (76%) had fewer than 200 species (or 7% of Afrotropical species) per country (Table 3). The most species-rich country (South Africa, with 1258 species or 45% of Afrotropical species) had almost twice as many species as the country in second place (Democratic Republic of the Congo, with 640 species or 23% of Afrotropical species) (Table 3). In terms of the percentage of species-per-country records, South Africa had more records than each of the country groups that included 3, 5 and 19 countries, and almost as many as the group containing 11 countries. These observations support the conclusion that knowledge of the country-scale distribution of bees is lacking.

There are other online sources of taxonomic data or species-level information about Afrotropical bees (Atlas Hymenoptera 2019, Discover Life 2019). Discover Life reports 1 161 species of bees in South Africa, 610 species in the Democratic Republic of Congo and 400 species in Zimbabwe. These numbers are close to our numbers in Table 3 (not including the 65 species records counted for South Africa that were found on the GBIF Data Portal rather than in published literature).

ii) Countries-per-species

Frequencies of the numbers of countries in which bee species have been recorded had a skewed distribution, with 62% of species occurring in only one country and a further 14% occurring in only two countries (Figure 2 and Table 4). Rather than indicating endemism, we interpreted this as support for the conclusion that country-scale knowledge of the distribution of bees is lacking.

Either the observed country-scale distribution is a true reflection or an artifact of different sampling effort or research intensity between different countries. Bee species richness is known to differ greatly between regions with different topography, climate or vegetation (Michener 1979, White 1983). We found that many of the countries with very low bee species numbers (e.g. Ivory Coast) had diverse climatic regions and vegetation types. Further, neighbouring countries of similar sizes at similar latitudes, which have similar vegetation and climatic conditions (White 1983), had very different numbers of species. For example, Ghana (area 238 535 km²) had 101 species and Ivory Coast (322 463 km²) had 51 species.

i) Afrotropical bee taxonomists and regional taxonomic work

The analysis of activity by taxonomists included all described taxa (i.e. counting both senior and junior synonyms as well as subspecies described before 1961).

In the Afrotropical Region there have been only four resident bee taxonomists. Three of them lived in South Africa (H. Brauns, V. Whitehead and C. Eardley) and the fourth (A. Pauly) lived and worked in Madagascar and Gabon and currently resides in Belgium. Other collectors did not study their own material but made it available to taxonomists. H. Brauns, for example, did not describe many species but he did collect for H. Friese who described numerous species from South Africa. Similarly, J. Ogilvie and A. Mackie collected mostly in South Africa for T.D.A. Cockerell. The fact that South Africa has four large bee collections (Table 6) is an indication that collecting effort has been greater in South Africa than in other African countries. There have also been a number of revisionary studies that included only the southern African bee fauna, as tabulated in Eardley et al. (2010). The bee fauna of Madagascar was the focus of several collectors in the past, as indicated by de Saussure (1892) and Pauly et al. (2001). The latter is a revisionary study that makes Madagascar arguably the best studied country in the Afrotropical Region.

In other countries with high numbers of species records, i.e. DRC, Zimbabwe, Namibia, Kenya and Tanzania, there were correspondingly high collecting efforts. In the DRC, collecting was apparently driven through the Musée royal de l’Afrique centrale, Brussels (MRAC). In Zimbabwe, P.A. Sheppard and R. Stevenson, among others, collected extensively. Staff at the National Museum of Namibia, Windhoek (NMNW), undertook numerous collecting expeditions. During the earlier part of the 20^th Century, a number of collectors collected extensively in Tanzania and Kenya. Species recorded from the above-mentioned countries, as well as South Africa, account for 49% of the known species-per-country records in the Afrotropical Region. This shows that countries known to have focused collection efforts have higher recorded numbers of bee species than those with less intensive or no known collecting efforts.

Even with the intensive effort to collect and conduct research on the taxonomy of South African bees, a recent study suggested that a number of bee species may remain undiscovered in South Africa. Substantiating this claim, Melin and Colville (2019) noted that two-thirds of South Africa has been under-sampled for bees and about 98% of the country has received little survey effort. These authors expressed concern over the decline in capacity for taxonomic research on insects, both globally and locally, and emphasised the urgent need for exploration to document this ecologically and economically important group.

ii) Authorship of taxonomic descriptions

Table 5 shows the number of species described by each author who described 10 or more species. Cockerell worked mostly on material from South Africa and the DRC, publishing 1772 taxonomic descriptions of bees (38%). Friese wrote the second highest number of taxonomic descriptions (655 species or 14%), and he received much of his material from Brauns’ South African collection. He also received much material from Tanzania. Nine authors, including Cockerell and Friese, each described over 100 species.

iii) Taxonomic descriptions over time

Whereas taxonomic research is an ongoing process, the description of new species may be used as an indicator of taxonomic activity over time.

Between 1758 and 2018 (260 years) 4654 taxonomic descriptions of Afrotropical bees were published. Three eras can be discerned (Figure 3). During the first era, up to the late 1800s, taxonomic publications largely comprised faunal works containing descriptions of many new species, and few keys and species comparisons were made. The era that followed (from the late 1800s up to 1965) included taxonomic descriptions, species comparisons and keys, but few revisions. In this era the period of maximum productivity was during 1903–1947 when 3 236 species (70%) were described, peaking in 1937 with the description of 319 new species (Figure 3). In only 18 of these years were 50 or more species described per year. This was the period when Cockerell and Friese were most productive. Between them they described 2 407 species and the first major revision (Friese 1909) was produced. The era of revisionary studies began in 1965 with Pasteels’ revision of Megachile (Pasteels 1965) and it continues to the present. A large number of genera have been revised or partly revised since 1965. Some of these studies include part of the region or certain subgenera within a genus (Eardley et al. 2010). This was a period of consolidation when many species names were synonymised and new distribution records were added to the knowledge about existing species.

Bee type material depositories

Taxonomists have to study type material in order to determine the correct name of a species. The locations of type depositories are therefore important from a logistical point of view. We found the type depositories of 4261 bee type specimens or type series (including subspecies) (Table 6 and Figure 4). An additional 393 types or type series have either been lost or their type depositories remain unknown or uncertain, making a total of 4654 types or type series.

Cockerell, the most prolific author of new bee species-level taxa, deposited most of his type material in the Natural History Museum, London (NHML), and this museum has the largest holding of Afrotropical bee types (1607 types, Table 6 and Figure 4). Most of Friese’s material is today housed in the Museum für Naturkunde der Humboldt-Universität, Berlin, Germany (ZMHB) (the remainder is distributed among a number of museums, Rasmussen and Ascher 2008). ZMHB has the second highest number of type specimens (919 types).

Eighty-one per cent of types are housed in Europe, 11% in Africa (9% in South Africa) and 8% in North and Central America.

Standardisation and publication of high-quality biodiversity data

Continuous updating and publication of the CAB will allow the data to be shared with researchers and other users of biodiversity information. The CAB is different from other online sources of information about Afrotropical bees because it is published on the GBIF Data Portal. This also implies that the data are both standardised (Biodiversity Information Standards) and integrated with other distributed data, allowing the user to find more data more easily. Standardisation of data is particularly important (Wieczorek et al. 2012, Walls et al. 2014). This refers to the practice of renaming database fields according to community-debated and agreed lists of standard terms (e.g. Taxon:scientificName) organised into broad classes, e.g. Occurrence, Event or Taxon. These classes and their terms function to organise the concepts and knowledge of biodiversity in a way that is compatible with the latest web technology, specifically that of Linked Open Data (Berners-Lee et al. 2001). This is a way to use the web as ‘a web of linked data’ rather than ‘a web of linked, differently formatted documents’. Central to this technology is the use of Uniform Resource Identifiers (URI), akin to the web URL, to name or identify things or resources (represented by data) on the web. Since URIs identify data values or concepts uniquely (i.e. without ambiguity), the URIs can be thought of as a dictionary of terms (Figure 1) to describe and structure, or organise, the concepts and knowledge used in Biodiversity Science. This technology will enable machines to do much of the drudge-work needed to find, integrate and prepare data for analysis. The technology may even allow machines to interpret data on behalf of experts. It is important to publish high-quality data if such endeavours are to be successful.

In the latest version of the CAB, all the genera of Afrotropical bees have common names. If a genus already had a common name this name was used, but genera that are endemic to the Afrotropical Region were given new names (keeping in mind existing genus names). Common names are important when communicating with citizen scientists or learners.

We found evidence strongly suggesting that knowledge of the country-scale distribution of Afrotropical bees is lacking. It is unlikely that similar, neighbouring countries at the same latitude would differ by a factor of two in their numbers of bee species, or that 76% of species would occur in only two countries. Probably, country differences in collecting effort or taxonomic research are responsible for the observed differences in species numbers between countries. Increased taxonomic capacity is therefore essential for a better understanding of the species richness and diversity patterns reflected by Afrotropical bees. The most important group of pollinators are worthy of far greater investment in mobilisation, to the GBIF Data Portal, and preservation, of African data.

It is difficult to produce useful taxonomic revisions, including reliable identification keys of Afrotropical bees, without studying type material. The logistics involved in studying types in different museums is influenced by the size of the museums’ holdings and the distances between them. Imaging of types and their labels will help taxonomists studying African bees, but usually it is necessary to examine type specimens. As 82% of types are housed in Europe, travel to this region is essential for revisionary taxonomic research. South Africa (9% of types) and the USA (8% of types) have significant type depositories and also need to be visited by bee taxonomists.

The taxonomic revisions of today are based on comparative morphology. Although morphology is indispensable in understanding and documenting species richness, future taxonomists will have to unravel complexities in species identity to which morphology cannot allude, such as separating sibling species and determining the limits between intra-specific and inter-specific variation. The obvious tools to do this are in molecular biology.

A well-designed, standardised biodiversity data management system is essential for scientific research. This will be especially important for the inclusion of additional data, such as associations with other species and biogeography. The biodiversity information system and content of the CAB will continue to be developed, updated and kept abreast of developing biodiversity metadata standards. For example, for validation purposes, the system needs to include literature citations of records of ecological associations, to allow their veracity to be evaluated.

The information system currently relies on a prototype system component, developed using Microsoft Access, to condense the data from the Specify database. This component needs to be developed into a reliable application with a convenient user-interface. Further, the data in the Specify database need to be migrated to a schema that will more precisely reflect the concepts and practices of taxonomy. Ultimately, the reason to digitise any data is to preserve its structure, integrity and safety, so that it may continue to be used.