Received: 28 April 2023 Accepted: 29 December 2023

DOI: 10.1111/icad.12715

ORIGINAL ARTICLE

Fine-scale bee species distribution models: Hotspots of

richness and endemism in South Africa with species-area
comparisons

Annalie Melin 1,2 | Colin M. Beale 3,4,5 | John C. Manning 1,6 |

7
Jonathan F. Colville

1
Compton Herbarium, South African National Abstract
Biodiversity Institute, Claremont, South Africa
2
African Centre for Coastal Palaeoscience, 1. While global patterns of bee diversity have been modelled, our understanding of
Nelson Mandela University, Port Elizabeth, fine-scale regional patterns is more limited, particularly for under-sampled regions
South Africa
3
such as Africa. South Africa is among the exceptions on the African continent; its
Department of Biology, University of York,
York, UK bee fauna (ca. 1253 species) has been well collected and documented, including
4
Leverhulme Centre for Anthropocene mass digitising of its natural history collections. It is a region with high floral diver-
Biodiversity, Department of Biology,
University of York, York, UK
sity, high habitat heterogeneity and variable rainfall seasonality.
5
York Environmental Sustainability Institute, 2. Here, we combine a South African bee species distributional database (877 bee
University of York, York, UK species) with a geospatial modelling approach to determine fine-scale
6
Research Centre for Plant Growth and
(11  11 km grid cell resolution) hotspots of bee species richness, endemism and
Development, School of Life Sciences,
University of KwaZulu-Natal, Scottsville, range-restricted species.
South Africa
3. Our analyses, based on the probabilities of occurrence surfaces for each species
7
Statistics in Ecology, Environment and
Conservation, Department of Statistical across 108,803 two-minute grid cells, reveal three bee hotspots of richness: Winter
Sciences, University of Cape Town, rainfall, Aseasonal rainfall and Early-to-late summer rainfall. These hotspots contain
Cape Town, South Africa
large numbers of endemic and geographically restricted taxa. Hotspots with particu-
Correspondence larly high bee diversity include the Fynbos, Succulent Karoo and Desert biomes; the
Annalie Melin, Compton Herbarium,
latter showing 6–20 times more species per unit area than other biomes. Our
South African National Biodiversity Institute,
Claremont, South Africa. results conform with global species-area patterns: areas of higher-than-expected
Email: [email protected]
bee density are largely concentrated in Mediterranean and arid habitats.
Funding information 4. This study further enhances our knowledge in identifying regional and global hot-
Natural Science Collections Facility; National spots of richness and endemism for a keystone group of insects and enabling these
Research Foundation, Grant/Award Number:
UID127738; Mapula Trust to be accounted for when setting conservation priorities.

Editor/Associate Editor: Laurence Packer KEYWORDS

Apoidea, endemic species, natural history collection data, range-restricted species, species density,
weighted endemism

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited.
© 2024 The Authors. Insect Conservation and Diversity published by John Wiley & Sons Ltd on behalf of Royal Entomological Society.

474 wileyonlinelibrary.com/journal/icad Insect Conserv Divers. 2024;17:474–487.

 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
HOTSPOTS OF SOUTH AFRICAN BEE DIVERSITY 475

I N T R O D U CT I O N richness pattern. Projecting Western Hemisphere bee models onto

Eastern Hemisphere regions (e.g., Orr et al., 2021) may be problematic
Bees are keystone pollinators for both wild flowering plants (Neff & due to inherent biotic differences, distinct species compositions and
Simpson, 1993; Kearns & Inouye, 1997; Ollerton et al., 2011) and agri- different evolutionary histories (De Palma et al., 2016; Kreft &
cultural crops (Garibaldi et al., 2013; Klein et al., 2007). Despite their Jetz, 2007; Michener, 1979; Ricklefs, 2004). In addition, the diverse
ecological and economic importance, a major drawback in developing environmental variability between hemispheres, encompassing climate
effective conservation priorities is our limited understanding of disparities and variations in habitat structures (Kreft & Jetz, 2007;
the patterns of bee diversity at finer scales, including local hotspots Wiens, 2011), poses a challenge when extrapolating models based on
of richness and range-restricted taxa. To remedy this, there has been geographically and taxonomically restricted data to other regions and
an ongoing effort to mine data from natural history collections taxa (De Palma et al., 2016; Elith & Leathwick, 2009). These limita-
(Bartomeus et al., 2018; Bystriakova et al., 2018; Orr et al., 2021). tions become less significant in data-rich regions, such as South Africa
Such data have also been used as an effective tool in assessing long- (Melin & Colville, 2019), where it would be more parsimonious to use
term changes in bee diversity and species declines due to global the source data to model potential species distributions and quantify
change, and in identifying priority areas for their conservation patterns of diversity.
(Bartomeus et al., 2013, 2018; Critchlow et al., 2022; Kougioumoutzis The main aim of this paper is to explore patterns of bee species
et al., 2022; Soroye et al., 2020; Zattara & Aizen, 2021). richness and endemism within South Africa, at a fine-scale resolution
Before the mass digitization of natural history collections, bee (11  11 km grid cell), and using a uniquely comprehensive digitised
biologist Charles Michener, using his accumulated expertise and inter- bee dataset that included representatives from all six bee families that
pretation of the literature, first described global patterns of bee spe- occur in the region. South Africa is a region of high species richness
cies richness as highest in warm temperate, xeric regions of the world (ca. 1253 species (Eardley & Coetzer, 2023)) and the Mediterranean
(Michener, 1979). This pattern contrasts with the diversity of many and arid habitats in the southwest of the country are particularly rich
other taxonomic groups, which are concentrated in the tropics follow- in bees (Bystriakova et al., 2018; Kuhlmann, 2009; Michener, 1979,
ing a latitudinal gradient of diversity (Gaston, 2000). A recent global 2007; Orr et al., 2021). The South African bee fauna is characterised
analysis modelled bee species richness using digitised natural history by high levels of endemism (Bystriakova et al., 2018; Kuhlmann, 2009)
collection data (Orr et al., 2021). It confirmed Michener’s descriptive and is regarded as evolutionary significant since it includes some of
patterns, finding bee species richness to be highest at mid-latitudes, the most early diverging bee lineages (Danforth et al., 2006; Litman
especially in warm temperate and xeric regions, showing a bimodal et al., 2011), as well as genera showing specialised pollinator adapta-
latitudinal richness gradient (Orr et al., 2021). tions (Steiner & Whitehead, 1990; Whitehead & Steiner, 1993).
In addition to this study, several macroecological analyses of bee Uniquely within the African continent, the bee fauna of the region
richness have been done across global or biogeographic regions has received a substantial amount of focused collection and taxo-
(Bystriakova et al., 2018; Kaloveloni et al., 2018; Leclercq et al., 2023; nomic effort over the past 250 years, including mass digitization of its
Michez et al., 2009; Patiny & Michez, 2007; Williams, 1998). Surpris- natural history collections (Melin & Colville, 2019). In this study, we
ingly, few studies investigating fine-scale (approximately <30  30 km use publicly available natural history collection data and employ a
grid cell) patterns of bee richness and endemism have been under- geospatial modelling approach (Colville et al., 2020; Mecenero
taken at the country level (i.e., regional scale) (but see Kuhlmann et al., 2015). This allows us to produce the most detailed assessment
(2009); Kougioumoutzis et al. (2022); Patiny and Michez (2007)). Anal- of South African bee diversity and distribution to date, with the aim of
ysis at this geographic scale requires data of sufficient quantity and delimiting local hotspots of bee species richness and endemism. We
precision using the widest taxonomic coverage. Although regions of also used richness-weighted indices to identify patterns of richness
interest could potentially be extracted from currently available global for restricted and widespread species (Crisp et al., 2001). Further-
species richness maps (e.g., Bystriakova et al., 2018; Orr et al., 2021), more, we determined the extent to which there is spatial congruence
there are several limitations to using these datasets to assess local among patterns of species richness, national endemic richness and
patterns within regions. These include using coarse-resolution data range-restricted species richness, as well as between national endemic
(e.g., ranging from 110  110 to 550  550 km grid cells), using a tax- richness and range-restricted endemic richness.
onomic subset of the data (e.g., family Colletidae) and projecting mod- The high bee diversity of Mediterranean and arid climate regions
elled Western Hemisphere bee data on to Eastern Hemisphere has attracted much attention over the past few years (Kaloveloni
regions. Coarse-resolution data are likely to obscure fine-scale rich- et al., 2018; Leclercq et al., 2023; Marshall et al., 2023; Meiners
ness patterns, particularly in regions with high habitat heterogeneity et al., 2019; Orr et al., 2021). We place South Africa’s bee diversity
(Cowling et al., 2017; Cowling & Lombard, 2002). The subfamily Colle- within this global pattern and compare how richness scales with area
tinae has been put forward as a taxon to represent richness patterns (Gotelli & Colwell, 2001) within South Africa—taking into account dif-
for all bees (Bystriakova et al., 2018); however, this could introduce ferent biomes and rainfall seasons—and with studies from other arid
spatial biases. For example, the richness pattern retrieved for and Mediterranean regions worldwide. Our approach involves esti-
South Africa by Orr et al. (2021), which used a more comprehensive mating species density of bees using a species-area relationship, build-
dataset of bees, strongly contrasts with Bystriakova et al.’s (2018) ing on recent studies that have employed this methodology (Meiners
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
476 MELIN ET AL.

et al., 2019; Minckley & Radke, 2021). Comparing bee density across we retained the occurrence records in our database for spatial
these different geographic areas can provide insights into understand- continuity.
ing global patterns of bee diversity (Minckley & Radke, 2021). Several biases are inherent in the presence-only or ‘atlas-type’
Detailed, modern biogeographic studies on South Africa’s insect data used here (Bartomeus et al., 2018; Graham et al., 2004). For
diversity are lacking (Colville et al., 2014; Melin & Colville, 2019). Con- example, collector effort might be biased towards areas that are
sequently, insects are underrepresented in conservation planning, pri- accessed more easily (e.g., towns, cities and road verges) as opposed
marily due to the large numbers of species involved, and the outdated to areas that are difficult to access. Also, collectors may target species
taxonomy for many groups, compounded by a decline in taxonomic of special interest over common taxa (Bartomeus et al., 2018; Graham
expertise (Chown & McGeoch, 1995; Hamer, 2012; Scholtz & et al., 2004; Mecenero et al., 2015; Melin & Colville, 2019). Another
Chown, 1995). The results of this study should further enhance our limitation of natural history collection data is that it provides informa-
knowledge in identifying important areas of insect richness and ende- tion on where species are present, not where they are absent. How-
mism in South Africa at a fine geographical scale and enabling these ever, biases can be reduced by combining such data with appropriate
hotspots to be accounted for when setting conservation priorities. statistical analyses (Araújo et al., 2019; Barbet-Massin et al., 2012;
Given the idiosyncratic patterns of bee richness (e.g., highest in Medi- Colville et al., 2020; Pearce & Boyce, 2006). To account for these
terranean, xeric and temperate habitats) (Bystriakova et al., 2018; biases, we employed a geospatial modelling technique (Mecenero
Michener, 2007; Orr et al., 2021), finer-scale data may also help guide et al. (2015) and recently used by Colville et al. (2020)) to interpolate
future studies, which aim to understand the macroecological factors the distribution records for each bee species and to calculate a contin-
driving bee diversity. uous probability of occurrence surface for each species at a 2-min grid
cell scale (11  11 km). Grid cells of this scale are considered an
appropriate spatial scale for capturing the highly heterogeneous habi-
MATERIALS AND METHODS tats of South African landscapes (e.g., Colville et al. (2020)), such as
steep upland/lowland gradients and floristic and edaphic transitions.
Determining hotspots of bee richness from species Using interpolated species distributions offered advantages over using
distribution models raw presence-only data, as our measure of richness and endemism
were not overly biased by gaps in the data (i.e., false absences).
We obtained species locality records from digitised natural history We followed the same modelling approach (‘Spatial Model 1’)
collections, mostly (86%) from South African institutions: Albany and code described in detail by Mecenero et al. (2015) and that had
Museum, Iziko Museums of South Africa, the National Collections of built on earlier models by Hengl et al. (2009). In summary, for each
Insects and Ditsong Museums of South Africa. The remainder of the bee species, we built a model at 2-min resolution combining point pat-
data is derived from institutions outside of South Africa; publicly avail- tern analysis methods with environmental niche information, to
able on GBIF.org (GBIF.org, 2023a, 2023b, 2023c, 2023d, 2023e, account for ecological similarity, inferred observer effort and geo-
2023f). The South African datasets constitute over 250 years of col- graphical distance. This process involved two stages. The first stage
lecting bees including ad hoc and targeted collecting by museums and involved selecting a sample of non-focal species records to act as
entomologists. Melin and Colville (2019) and Coetzer and Eardley pseudo-absences (reflecting the pattern of observation in the dataset),
(2019) provide detailed historic information including a summary of and the second stage involved interpolating distributions based on
institutional and collector contributions. Together this original dataset presence and pseudo-absence records. The first stage consisted of
is comprised of 49,498-point locality records, which underwent a several separate steps: (1) mapping all records of the focal species and
range of quality checks to improve its reliability and accuracy. For generating a kernel density estimate for records of this species,
common geospatial issues found in natural history collections (2) identifying all records of all other bee species >100 m from records
(Robertson et al., 2016) such as point-localities outside of their range of the focal species and generating similar kernel density estimates,
(including points out at sea) for South Africa, we verified their accu- (3) computation of the difference in density estimates between focal
racy and either corrected or removed them if georeferencing was not and non-focal species (an approximate index of the probability of
feasible. Similarly, disjunct distributions were identified as any data encountering the focal species), (4) computation of an environmental
points >200 km from its nearest neighbour. These were flagged as envelope within a principal component analysis of rainfall (mean
potential outliers and manually checked and corrected. annual rainfall and rainfall season) (Schulze, 2007) and temperature
We restricted the data to include only valid species by removing variables (mean winter and mean summer temperature)
records without a species name (e.g., spp. and sp.). To maintain taxo- (Schulze, 2007), (5) computing the environmental distance between all
nomic identity, we followed the published regional checklist for 2-min raster cells and the centroid of the environmental envelope
South African bees (Eardley & Coetzer, 2023) and concerning synon- occupied by the focal species and (6) sampling records of the
ymy and spelling we followed Eardley & Urban (2010) and Coetzer non-focal species using the environmental distance and geographic
and Eardley (2019). Our cleaned dataset consisted of 26,474-point probability of encountering the focal species to bias selection towards
locality records across 877 bee species (ca. 70% of South African spe- locations where absence was most likely. Once pseudo-absence
cies), of which 234 species had less than five-point locality records. records were selected, the second stage of analysis used regression
For South African bee species that also occur in Lesotho and Eswatini, kriging of the presence/absence points onto the 2-min raster surface,
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
HOTSPOTS OF SOUTH AFRICAN BEE DIVERSITY 477

using the rainfall and temperature covariates. For 234 bee species considered important drivers of biogeographic patterns of plant and
recorded from <5 locations, we were unable to accurately interpolate insect diversity (Colville et al., 2014; Procheş & Cowling, 2007).
distribution and we therefore generated a raster map with presence We calculated the species density for South African biomes and
(1) and assumed absence (0) directly from the raw data. One of us rainfall seasons by dividing the number of species by the respective area
(A.M.) verified the modelled species distributions for groups they have of the biome (n = 7; Figure 3a) or rainfall season (n = 6; Figure 3b). To
detailed knowledge on by checking the maps against known distribu- compare density among biomes or rainfall seasons, we calculated the
tions. Mecenero et al. (2015) undertook the same approach and found ratio by dividing density of one biome/rainfall season by another.
that modelled species distributions matched expert opinion, stressing
the general accuracy of their modelling approach.
Our final modelled bee dataset consisted of probabilities of Bee density: Global comparisons
occurrence surfaces for each bee species across 108,083 2-min grid
cells. Species richness and endemism were calculated for each 2-min To compare South Africa’s bee biome density with areas sharing simi-
grid cell as the summed probability surfaces for all our bee species. lar habitat and climate conditions, we collated data for arid and Medi-
terranean sub-regions from nine published studies and one
unpublished dataset (see Data S1). To also make comparisons at a
Determining hotspots of endemism and range- global scale, we took advantage of recently collated (Orr et al., 2021)
restricted species regional-level data for 170 countries with areas over 5000 km2
(Data S1). To ascertain how richness scales with area, we used a
To determine endemic species richness, we compiled a national list species-area relationship as a double logarithmic equation
of endemic bee species (646 species) from Eardley and Coetzer (Rosenzweig, 1995, and as used recently in bee density studies by
(2023). All national endemic species are not known to occur in Meiners et al., 2019; Minckley & Radke, 2021) based on the species
Lesotho or Eswatini. However, our grid cells fell across political richness and area data presented in Data S1.
boundaries, and we therefore allowed our models to predict spe- To assess whether studies conducted in arid and Mediterranean
cies distributions across these artificial boundaries. Of the areas (n = 17, sub-region and biome data in Data S1) recorded more
646 listed endemic bee species, our modelled species distributional or fewer species than expected, we followed Meiners et al.’s (2019)
dataset contained 341 (53%). As with overall species richness, we and Minckley and Radke’s (2021) approaches and measured the dis-
summed the probability of occurrence for each endemic species in tance from each study’s species-area point to the overall log–log
a grid cell to obtain a measure of endemic bee species richness. regression line. With these measurements, we generated a bar plot to
We determined hotspots of range-restricted species richness for rank and compare how much the observed number of bee species
all species and for national endemics using two indices (weighted deviates, either above or below, relative to the predicted number of
endemism [WE] and corrected WE [CWE]) proposed by Crisp et al. bee species.
(2001) and modified to accept our probability of occurrence for each All analyses were carried out using the program R version 4.1.2
species. CWE accounts for the correlative role of species richness in (R Core Team, 2021) and packages: ‘ggplot2’ (Wickham, 2009) was
determining endemic and range-restricted richness patterns. To esti- used to plot the species-area relationship for different countries,
mate WE, we weighted the probability of occurrence for each species biomes and sub-regions, and ‘sp’ (Bivand et al., 2013) ‘spatstat’
by the inverse of their range (1/number of cells occupied by a species, (Baddeley et al., 2015), ‘rgdal’ (Bivand et al., 2018) and ‘gstat’ (Graler
summed across all species in a cell), and for CWE we correct WE for et al., 2016) were used to perform species distribution models and
overall species richness in a cell (WE/cell richness). spatial overlays. Species richness maps, as in Figures 1–3, were ren-
dered in QGIS version 2.18.14 (QGIS Development Team, 2023).

Congruence of species richness with endemic richness

and range-restricted species richness RE SU LT S

We used Pearson’s correlations to compare the spatial similarity of Species richness concentrates into three distinct hotspots
overall species richness with national endemic richness and range- (Figures 1–3): Winter rainfall bee hotspot the arid and mesic winter
restricted species richness. Similarly, national endemic richness was rainfall zone along the west coast and incorporating parts of the Fyn-
correlated with range-restricted endemic richness. bos and Succulent Karoo biomes; Aseasonal rainfall bee hotspot the
mesic aseasonal rainfall zone extending along the south coast incorpo-
rating parts of the Fynbos, Succulent Karoo and Albany Thicket
Bee density: Biomes and rainfall seasonality biomes and encroaching inland towards the drier interior of the very
late summer rainfall zone; and Early-to-late summer rainfall bee hot-
We calculated bee richness for South Africa’s biomes (Mucina & spot a narrow band along the east coast (subtropical) extending inland
Rutherford, 2006) and for rainfall seasonality (Schulze, 2007) by per- into the high-altitude montane region of the eastern side of the Dra-
forming spatial overlays. Biomes and rainfall seasonality have been kensberg Mountains, and comprising the north-eastern extent of the
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
478 MELIN ET AL.

F I G U R E 1 South African bee richness hotspots at a 11  11 km grid cell resolution: (1) Winter rainfall bee hotspot: along the west coast
and incorporating parts of the Fynbos and Succulent Karoo biomes, (2) Aseasonal rainfall bee hotspot: extending along the south coast
incorporating parts of the Fynbos, Succulent Karoo and Albany Thicket biomes, (3) Early-to-late summer rainfall bee hotspot: a narrow band along
the east coast extending inland into the high-altitude montane region of the eastern side of the Drakensberg Mountains, and including parts of
the Grassland and Savanna biomes.

summer rainfall zone and including parts of the Grassland and between species richness and range-restricted species richness
Savanna biomes. (Pearson’s R = 0.87, p < 0.001). Despite the high correlations among
National endemic richness (unweighted) is concentrated in the west- these three indices, there are some differences in the locations of the
ern parts of South Africa, around the Winter rainfall bee hotspot and the hotspots (Figure 2a–c). We also found a strong positive correlation
Aseasonal rainfall bee hotspot (Figure 2b). Hotspots of unweighted between the spatial patterns of national endemic richness and range-
range-restricted species (Figure 2c) and range-restricted endemic species restricted endemic richness (Pearson’s R = 0.93, p < 0.001; Figure 2b,d).
(Figure 2d) are also concentrated around these two richness hotspots, Comparisons across biomes and rainfall seasons showed that the
although for the latter category, the Winter rainfall bee hotspot is number of bee species was similar (Table 1); however, when values
extended as a hotspot of range-restricted endemic species. were weighted on area size, species density was higher in the smaller-
After weighting for richness of national endemics, both the Win- sized biomes and rainfall seasons (Figure 4a and Tables 2 and 3). The
ter rainfall bee hotspot and the Aseasonal rainfall bee hotspot become Fynbos, Succulent Karoo and Desert biomes, making-up 16% of
more conspicuous, whereas the Early-to-late summer rainfall bee hot- the total area of South Africa (Mucina & Rutherford, 2006), had
spot is less so, with only the eastern KwaZulu-Natal Drakensberg almost six times as many species per unit area than the three largest
Mountains being rich in both species and national endemics. When biomes (Savanna, Grassland and Nama Karoo), which make-up the
correcting for the richness of range-restricted species, the Winter bulk of South Africa (Table 2). The Desert biome showed the highest
rainfall bee hotspot is emphasised as an area with a high proportion of species density, with approximately six times more species per unit
bee species with narrow geographic ranges (Figure 2e). However, area than the Fynbos and Succulent Karoo biomes, and >20 times
when correcting for range-restricted endemic richness a somewhat more species per unit area than the Grassland, Savanna and Nama-
different pattern is retrieved (Figure 2f). The winter rainfall zone is Karoo biomes. A similar pattern was seen for rainfall seasonality, with
retained but extends further into the central arid interior and includes the winter and aseasonal rainfall areas (15% of the total area of
elements of the Aseasonal rainfall bee hotspot as well as areas around South Africa) having approximately two and five times more species
the Early-to-late summer rainfall bee hotspot. per unit area than the summer rainfall areas, respectively (Table 3).
There was a significant positive correlation between species richness Area was a poor predictor of bee species richness explaining 6.6%
and national endemic richness (Pearson’s R = 0.64, p < 0.001), as well as of the variation in bee richness (F1,184 = 14.13, p < 0.001) (Figure 4a).
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
HOTSPOTS OF SOUTH AFRICAN BEE DIVERSITY 479

FIGURE 2 Spatial patterns of South African bee diversity (a–f) at a 11  11 km grid cell resolution.

All South Africa’s biomes fell above the regression line and outside of Comparison of areas with similar arid and Mediterranean climates
the 95% confidence interval indicating that they have higher bee spe- revealed that the San Bernardino Valley (Arizona/ Sonora) holds the
cies density than predicted by area (Figure 4a,b). Although all biomes highest rank, followed in descending order by Pinnacles National Park
fell above the regression line, the smallest biomes had the largest dif- (California), Lesvos (Greece), Yosemite National Park (California), and
ferences in the number of species observed relative to the number of Grand Staircase Escalante National Monument (Utah) (Figure 4b).
species predicted (Figure 4b). These small-sized biomes are those with These rankings surpassed those of all South Africa’s biomes. Notably,
arid and Mediterranean climates. the Albany Thicket and Fynbos biomes outperformed analogous areas
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
480 MELIN ET AL.

FIGURE 3 Biomes (a) and rainfall seasonality (b) of South Africa.

such as Mediterranean Chile and Southwestern Australia (Figure 4b). rainfall regions and the third with the summer rainfall region
Additionally, the more arid Succulent Karoo and Desert biomes (Figures 1 and 3b). Hotspots of endemics were also predominantly
secured higher rankings than the Mojave National Preserve, San located within the winter and aseasonal rainfall areas of the Greater
Rafael Desert and Desert Chile. Cape Floristic Region (GCFR), except for a small high-elevation hot-
spot of endemism found in the summer rainfall Drakensberg Moun-
tains. Patterns of bee diversity have been assessed within
DISCUSSION South Africa to some extent, where Kuhlmann (2009) focused primar-
ily on the bee families Apidae and Colletidae, Bystriakova et al. (2018)
Using the widest taxonomic coverage of South African bees (877 spe- emphasised Colletidae, and Orr et al. (2021) used projected data for
cies across 85 genera) to date, we identified three main bee richness all families. This is the first fine-scale study using geospatial modelling
hotspots—two are closely aligned with the winter and aseasonal to assess patterns of richness, endemism and range-restricted taxa. In
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
HOTSPOTS OF SOUTH AFRICAN BEE DIVERSITY 481

TABLE 1 Number of bee species and national endemic species found in South African biomes and rainfall seasons.

Species richness (endemics)

Biomes
Fynbos Succulent Karoo Savanna Grassland Albany Thicket Desert Nama-Karoo
698.29 674.64 712 687.00 594.00 330.83 650.98
(233.50) (221.90) (215.00) (221.00) (183.25) (114.11) (200.00)
Rainfall seasonality
All year Winter Early-summer Mid-summer Late-summer Very late-summer
650.83 (208.62) 718.55 (246.92) 737.00 (229.00) 663.90 (201.95) 667.00 (206.00) 669.00 (205.00)

F I G U R E 4 Relationship between area and the number of bee species (log-transformed): (a) Each data point (n = 186; Data S1) represents a
country, biome, and sub-region. The solid black line represents the linear regression fit, and the grey shaded region around the line shows the
95% confidence interval. Data points that fall above the line show higher than expected bee richness for their area size. Labels for South African
biomes (circle with a star) and selected sub-regions have also been included for comparison and aesthetics; two countries have been highlighted:
South Africa (this study; circle with a cross) and Israel (region with the highest bee density; circle with a dot). (b) Barplot of the difference in the
number of bee species observed relative to the number of bee species predicted by the regression line plotted in panel A. Areas that recorded
more bee species than expected are shown in blue with the exception of one area shown in red that recorded less bee species than expected.
South African biomes are outlined in black. The following have been abbreviated: GSE National Monument, Grand Staircase Escalante National
Monument; SA, South Africa; for states in North America: AZ, Arizona; CA, California; UT, Utah. Full study details are given in Data S1.

addition, it is the first regional study to do this for any South African (Figure 4a,b). This western region is characterised by having a high
insect group. proportion of distinct bee genera: Scrapter, Fidelia and Patellapis
We attribute South Africa’s remarkably rich bee fauna in the Win- (Danforth et al., 2013; Hedtke et al., 2013; Litman et al., 2011;
ter rainfall bee hotspot to the Succulent Karoo and Fynbos biomes Michener, 1979), which elevates this hotspot’s profile as an important
(Figures 2a–f and 3a), and the Desert biome (Figures 2d–f and 3a). area for bee conservation. Added to this, the region is also of evolu-
Bee density in these winter rainfall biomes is considerably higher com- tionary significance for bees as a centre of diversity for Melittidae, a
pared with the larger summer rainfall South African biomes phylogenetically deep lineage of bees and sister to all other bees
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
482 MELIN ET AL.

TABLE 2 Comparisons of number of species density across South Africa’s biomes.

Biomes Fynbos Succulent Karoo Savana Grassland Albany Thicket Desert Nama-Karoo

Fynbos 1.00 1.02 4.83 4.30 0.39 0.16 3.17

Succulent Karoo 0.98 1.00 4.74 4.22 0.38 0.16 3.11
Savanna 0.21 0.21 1.00 0.89 0.08 0.03 0.66
Grassland 0.23 0.24 1.12 1.00 0.09 0.04 0.74
Albany Thicket 2.55 2.60 12.32 10.97 1.00 0.41 8.09
Desert 6.25 6.38 30.20 26.87 2.45 1.00 19.82
Nama-Karoo 0.32 0.32 1.52 1.36 0.12 0.05 1.00

Note: Comparisons are read left to right, and values indicate the ratio after dividing species density of one biome by another. Biome area size (km2) from
smallest to largest: Desert (6105.19); Albany Thicket (26862.81); Succulent Karoo (79367.41); Fynbos (80588.442); Nama Karoo (238102.22); Grassland
(340669.32); Savana (396837.03).

TABLE 3 Comparisons of number of species density across South Africa’s rainfall seasons.

Rainfall seasons All year Winter Early-summer Mid-summer Late-summer Very late-summer

All year 1.00 2.26 3.03 5.68 5.11 6.08

Winter 0.44 1.00 1.34 2.52 2.26 2.69
Early summer 0.33 0.74 1.00 1.88 1.69 2.01
Mid-summer 0.18 0.18 0.53 1.00 0.90 1.07
Late-summer 0.20 0.44 0.59 1.11 1.00 1.19
Very late-summer 0.16 0.37 0.50 0.93 0.84 1.00

Note: Comparisons are read left to right, and values indicate the ratio after dividing species density of one season by another. Rainfall season area size
(km2) from smallest to largest: All-year (52272.04481); Winter (136589.81); Early-summer (149991.16); Late-summer (257857.34); Very-late-summer
(305383.55); Mid-summer (318943.09).

(Danforth et al., 2013; Michez et al., 2008). The life history traits, such are important ecological factors in explaining insect diversity in this
as the ancestral state of not lining their nests as seen in Fidelia, may western region. The generally strong relationship between plant and
explain the restriction of some lineages to arid environments (Litman insect species diversity is likely due to both direct interactions
et al., 2011). The winter rainfall Succulent Karoo and Desert (desert and where both groups are responding to similar environmental fac-
climates) and Fynbos (Mediterranean climate) biomes are recognised tors (Colville et al., 2014; Procheş & Cowling, 2006).
globally for their exceptional plant diversity, which is disproportionally Similar ecological factors to the Winter rainfall bee hotspot are
high in comparison to its physical size and to other Mediterranean likely to influence the diversity of bees in the western part of the
and desert regions (Cowling et al., 1998, 1999; Manning & Aseasonal rainfall bee hotspot because it is composed of Fynbos and
Goldblatt, 2012; Mittermeier et al., 1998; Myers et al., 2000). Ecologi- Succulent Karoo vegetation (Cowling et al., 2017). This southern hot-
cal factors associated with the high plant diversity within these spot is characterised by having a high proportion (>50%) of species of
biomes include fine-scale climatic, edaphic and topographic heteroge- Compsomelissa and Allodapula; both have distinctive life history traits
neity, fire regime, and biotic interactions (e.g., plant-pollinator coevo- related to their sociality and progressive feeding (Michener, 1979;
lution) (Cowling et al., 2017). Areas within this region have also been Schwarz et al., 2003). The eastern part of the Aseasonal rainfall bee
strongly influenced by historical climatic stability, which is thought to hotspot is also likely to be influenced by a high degree of habitat het-
have lowered rates of extinction and allowed for higher niche speciali- erogeneity due to the transitional nature of several floristic elements
sation in plants (Colville et al., 2020; Cowling et al., 1998, 2015, found here: Succulent Karoo, Albany Thicket, Nama-Karoo, Savanna
2017). This area is also known as a global centre for species richness and Grassland biomes (Figure 3a). This region, also known as the
for several other insect groups, for example, monkey beetles (Hopliini Albany Centre, has a very old phytogeographic origin (Clark
(Colville et al., 2018)), grasshoppers (Lentulidae (Otte, 2020)), weevils et al., 2009; Cowling & Procheş, 2005; Van Wyk & Smith, 2001), indi-
(Curculionidae (Hévin et al., 2022)), heelwalkers (Mantophasmatodea cating an area of historical climatic stability (Colville et al., 2020; Potts
(Predel et al., 2012)) and several others (Colville et al., 2014). Findings et al., 2013), which may also contribute to its higher than expected
to date suggest that plant species richness (Procheş et al., 2009; bee diversity. Vertebrates (Perera et al., 2011) and several insect
Wright & Samways, 1998), plant phylogenetic diversity (Procheş groups, for example, butterflies (Carcasson, 1964), damselflies
et al., 2009), soil-type preferences for nesting (Gess, 1992; Gess & (Dijkstra, 2007) and grasshoppers (Dirsch, 1965; Gordon et al., 2023)
Gess, 2014) and temperature (Botes et al., 2007; Stuckenberg, 1969) also have high species diversity in this region. After adjusting for
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
HOTSPOTS OF SOUTH AFRICAN BEE DIVERSITY 483

species richness, this hotspot maintains a higher concentration of geo- and specialisation, and species turnover (beta diversity (e.g., Colville
graphically restricted taxa than would be expected based solely on et al. (2002)), are associated with the high number of fine-scale floris-
species richness (Figure 2c,d). Although we have factored in collector tic and edaphic niches of these areas (see also Cowling et al. (2015)).
effort through geospatial modelling, the Aseasonal rainfall bee hotspot In summary, the Mediterranean climate and arid areas in
might be the result of intensive collecting carried out by hymenopter- South Africa contain a high density of bee species and a large number
ists Sarah K. Gess and the late Friedrich W. Gess over a period of of endemic and geographically restricted bee taxa. Although our diver-
approximately 40 years (Gess & Gess, 2014; Melin & Colville, 2019). sity maps are based on 70% of South African bee species, adding
However, these hymenopterists conducted widespread sampling records for missing species would most likely strengthen our hotspots.
throughout the western parts of South Africa during this timeframe These hotspots have frequently been recognised as areas of high rich-
(Gess & Gess, 2014). ness for a range of insects and plants. This study goes further by illus-
Between the extremes of the Winter rainfall bee hotspot and the trating the entomofaunal distinctiveness of these hotspots in terms of
Aseasonal rainfall bee hotspot, we see a large proportion of the GCFR disproportionally high numbers of species, endemics and range-
as seemingly depauperate of bee diversity. Given that the Winter rain- restricted taxa. The strong correlations observed between diversity
fall bee hotspot and the Aseasonal rainfall bee hotspot largely fall metrics suggest that species richness is a suitable indicator of endemic
within the GCFR, with similar habitat and processes, it is conceivable and range-restricted bees in South Africa.
that bee diversity is most likely higher in this area, for example, high
species richness has been retrieved in well-sampled insects such as but-
terflies (Mecenero et al., 2013) in this area. This gap may be due to poor CONC LU SION
sampling (Melin & Colville, 2019) and therefore targeted surveying in
this area, particularly in protected areas as seen in recent bee surveys Understanding bee diversity at a regional scale can give insights into
(e.g., Meiners et al., 2019; Messinger Carril et al., 2018), could be used local scale patterns and can be highly informative for understanding
to establish if the Winter rainfall bee hotspot and the Aseasonal rainfall broader global patterns of bee diversity. It can also provide important
bee hotspot are contiguous across the south-eastern GCFR. information on local areas of insect diversity, such as hotspots for
The Early-to-late summer rainfall bee hotspot is dominated by conservation, areas of evolutionary interest where certain lineages are
more widespread generalist bee species and shares many of the taxa concentrated, and in understanding the historical biogeography of a
with other Afrotropical areas (e.g., Eucerini (Eardley, 1989) and Melec- region. The spatial patterns of species richness and endemism in
tini (Eardley, 1991)). The low numbers of endemic and range-restricted South African bee fauna presented here corroborate but vastly extend
species richness in this hotspot, particularly away from the Drakens- the results of previous studies and conform with global patterns of
berg area, confirm this pattern (Figure 2b–d). In contrast, Kuhlmann unusually high bee diversity in Mediterranean and arid areas. The next
(2009) using a restricted species dataset and at a coarse resolution steps in understanding these regional patterns will require an assess-
(200  200 km grid square) found the proportion of endemic bee ment of the underlying processes that shape them (Kaloveloni
species in the early-to-mid summer rainfall area to be almost equiva- et al., 2018), particularly in the context of specialisation and the
lent to the winter rainfall area, although recognising that the true pat- extraordinarily rich flora and the role of historical climatic stability.
tern is most likely considerably lower in the early-to-mid summer Future analysis, using our species distributional models, can also be
rainfall area and more in line with the results of this study. In the extended to calculate other patterns of diversity such as beta-
Early-to-late summer rainfall bee hotspot, other insects, such as but- diversity (species turnover) and functional diversity (e.g., Leclercq
terflies (Mecenero et al., 2013, 2015), dragonflies (Deacon et al., 2023; Marshall et al., 2023). Because of the high numbers of
et al., 2020), and dung beetles (Davis, 2002) exhibit high species rich- range-restricted taxa confined to small numbers of grid cells, habitat
ness. Similar to bees, these insects demonstrate overall low values of specialisation is most likely driving high species turnover and func-
endemism and share numerous widespread taxa with other Afrotropi- tional diversity in the Winter rainfall bee hotspot and the Aseasonal
cal areas. rainfall bee hotspot. As undertaken for plants (Cowling et al., 2015),
As observed in other parts of the world, our bee richness and comparisons across other global Mediterranean and arid areas of
endemism patterns conform to the general pattern that bee faunas exceptional bee richness would also provide deeper insights into the
are highest in arid or Mediterranean climates (Michener, 1979; Orr drivers of bee diversity and the role of habitat heterogeneity, includ-
et al., 2021). Across these regions, specific climatic factors and habitat ing floristic resources, and historical climate stability. In addition,
heterogeneity including high floristic diversity rather than area size determining the extent to which the existing protected area networks
per se appear the more important variables in determining bee rich- in South Africa, based predominantly on plant-based data
ness (but see Kaloveloni et al. (2018)). This scenario has also been (e.g., Skowno et al. (2018)), adequately cover the bee hotspots war-
hypothesized for other areas of exceptional bee richness, such as Pin- rants further scrutiny. This could be expanded by taking into consider-
nacles National Park in Central California (Meiners et al., 2019). Our ation the vulnerability of the South African bee fauna under several
results also show that these areas are not only disproportionally high scenarios of global change (Kuhlmann et al., 2012). Finally, refining
in bee species richness but also high in endemics and species with nar- the hotspot boundaries presented here would require long-term
row geographic ranges. This suggests that rates of bee diversification inventories to ground truth these patterns.
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
484 MELIN ET AL.

AUTHOR CONTRIBUTIONS Baddeley, A., Rubak, E. & Turner, R. (2015) Spatial point patterns: methodol-
Annalie Melin: Conceptualization; methodology; data curation; formal ogy and applications in R. CRC Press, Taylor & Francis Group, Boca
Raton, Florida.
analysis; funding acquisition; visualization; writing – original draft;
Barbet-Massin, M., Jiguet, F., Albert, C.H. & Thuiller, W. (2012) Selecting
writing – review and editing; investigation. Colin M. Beale: Methodol- pseudo-absences for species distribution models: how, where and
ogy; formal analysis; writing – review and editing; resources; investi- how many? Methods in Ecology and Evolution, 3, 327–338.
gation; data curation. John C. Manning: Writing – review and editing; Bartomeus, I., Ascher, J.S., Gibbs, J., Danforth, B.N., Wagner, D.L.,
Hedtke, S.M. et al. (2013) Historical changes in northeastern US bee
funding acquisition; resources. Jonathan F. Colville: Conceptualiza-
pollinators related to shared ecological traits. Proceedings of the
tion; methodology; data curation; formal analysis; investigation; visu- National Academy of Sciences of the United States of America, 110,
alization; funding acquisition; writing – review and editing. 4656–4660.
[Correction added on 2 February 2024, after first online publication: Bartomeus, I., Stavert, J., Ward, D. & Aguado, O. (2018) Historical colal-
phabeticallylections as a tool for assessing the global pollinator crisis.
The contribution of Jonathan F. Colville has been updated.].
Philosophical Transactions of the Royal Society B: Biological Sciences,
374, 20170389.
ACKNOWLEDGEMEN TS Bivand, R., Keitt, T. & Rowlingson, B. (2018) rgdal: bindings for the “Geos-
We wish to acknowledge the enormous contributions made by Sarah patial” Data Abstraction Library. R package.
Bivand, S.R., Pebesma, E.J. & Gomez-Rubio, V. (2013) Applied spatial data
K. Gess (Albany Museum), Connal Eardley (National Collections of
analysis with R, 2nd edition. Springer, New York, NY.
Insects) and Simon van Noort (Iziko Museums of South Africa) in man-
Botes, A., McGeoch, M.A. & Chown, S.L. (2007) Ground-dwelling beetle
aging and digitising the natural history collections. We thank Richard assemblages in the northern cape floristic region: patterns, correlates
Cowling for reading and commenting on the manuscript and Terry and implications. Austral Ecology, 32, 210–224.
Houston for advice on bee species numbers for Southwestern Bystriakova, N., Griswold, T.L., Ascher, J.S. & Kuhlmann, M. (2018) Key
environmental determinants of global and regional richness and
Australia.
endemism patterns for a wild bee subfamily. Biodiversity and Conser-
vation, 27, 287–309.
FUND ING INFORMATION Carcasson, R.H. (1964) A preliminary survey of the zoogeography. African
Annalie Melin was supported by Elizabeth Parker, the Mapula Trust, Journal of Ecology, 2, 122–157.
Chown, S.L. & McGeoch, M.A. (1995) South African terrestrial zoology:
the Natural Science Collections Facility, and National Research Foun-
strengths, weaknesses and opportunities. South African Journal of Sci-
dation (PDP postdoctoral research fellowship grant no. UID127738). ence, 91, 189–196.
Clark, V.R., Barker, N.P. & Mucina, L. (2009) The Sneeuberg: a new centre
CONF LICT OF IN TE RE ST ST AT E MENT of floristic endemism on the great escarpment, South Africa.
South African Journal of Botany, 75, 196–238.
The authors declare no conflicts of interest.
Coetzer, W. & Eardley, C.D. (2019) Insights into 260 years of taxonomic
research gained from the catalogue of Afrotropical bees. African
DATA AVAI LAB ILITY S TATEMENT Invertebrates, 60, 291–318.
The data that support the findings of this study are publicly available Colville, J.F., Beale, C.M., Forest, F., Altwegg, R., Huntley, B. &
Cowling, R.M. (2020) Plant richness, turnover and evolutionary diver-
from GBIF.org: GBIF.org (2023) GBIF Occurrence Download, Colleti-
sity track gradients of stability and ecological opportunity in a mega-
dae. https://doi.org/10.15468/dl.gktdhq, GBIF.org (2023) GBIF diversity centre. Proceedings of the National Academy of Sciences,
Occurrence Download, Melittidae. https://doi.org/10.15468/dl. 117, 20027–20037.
byf3rz, GBIF.org (2023) GBIF Occurrence Download, Megachilidae. Colville, J.F., Picker, M.D. & Cowling, R.M. (2002) Species turnover of
monkey beetles (Scarabaeidae: Hopliini) along environmental and
https://doi.org/10.15468/dl.87hk2g, GBIF.org (2023) GBIF Occur-
disturbance gradients in the Namaqualand region of the Succulent
rence Download, Andrenidae. https://doi.org/10.15468/dl.vr67q8,
Karoo, South Africa. Biodiversity and Conservation, 11, 243–264.
GBIF.org (2023) GBIF Occurrence Download, Apidae. https://doi.org/ Colville, J.F., Picker, M.D. & Cowling, R.M. (2018) Feeding ecology and
10.15468/dl.wnamdv, GBIF.org (2023) GBIF Occurence Download, sexual dimorphism in a speciose flower beetle clade (Hopliini: Scara-
Halictidae. https://doi.org/10.15468/dl.7rjdz2, GBIF.org. (2023) GBIF baeidae). PeerJ, 6, e4632.
Colville, J.F., Potts, A.J., Bradshaw, P.L., Measey, G.J., Snijman, D.,
Occurrence data for bee species of Southwestern Australia Download
Picker, M.D. et al. (2014) Floristic and faunal Cape biochoria: do they
https://doi.org/10.15468/dl.ksa8ng. exist? In: Allsopp, N., Colville, J.F. & Verboom, G.A. (Eds.) Fynbos:
ecology, evolution and conservation of a megadiverse region. Oxford
ORCID University Press, Oxford, United Kingdom, pp. 73–92.
Cowling, R.M., Bradshaw, P.L., Colville, J.F. & Forest, F. (2017) Levyns’ law:
Annalie Melin https://orcid.org/0000-0003-3393-1701
explaining the evolution of a remarkable longitudinal gradient in
Colin M. Beale https://orcid.org/0000-0002-2960-5666 Cape plant diversity. Transactions of the Royal Society of South Africa,
John C. Manning https://orcid.org/0000-0003-2886-8787 72, 184–201.
Jonathan F. Colville https://orcid.org/0000-0003-2176-3077 Cowling, R.M., Esler, K.J. & Rundel, P.W. (1999) Namaqualand, South
Africa—an overview of a unique winter-rainfall desert ecosystem.
Plant Ecology, 142, 3–21.
RE FE R ENC E S Cowling, R.M. & Lombard, A.T. (2002) Heterogeneity,
Araújo, M.B., Anderson, R.P., Barbosa, A.M., Beale, C.M., Dormann, C.F., speciation/extinction history and climate: explaining regional plant
Early, R. et al. (2019) Standards for distribution models in biodiversity diversity patterns in the Cape Floristic Region. Diversity and Distribu-
assessments. Science Advances, 5, 1–12. tions, 8, 163–179.
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
HOTSPOTS OF SOUTH AFRICAN BEE DIVERSITY 485

Cowling, R.M., Potts, A.J., Bradshaw, P.L., Colville, J.F., Arianoutsou, M., GBIF.org. (2023b) GBIF occurrence download, Melittidae. Available from:
Ferrier, S. et al. (2015) Variation in plant diversity in Mediterranean- https://doi.org/10.15468/dl.byf3rz [Accessed 24th July 2023]
climate ecosystems: the role of climatic and topographical stability. GBIF.org. (2023c) GBIF occurrence download, Megachilidae. Available
Journal of Biogeography, 42, 552–564. from: https://doi.org/10.15468/dl.87hk2g [Accessed 24th July
Cowling, R.M. & Procheş, Ş. (2005) Patterns and evolution of plant diver- 2023]
sity in the Cape Floristic Region. Biological Ecology, 55, 273–288. GBIF.org. (2023d) GBIF occurrence download, Andrenidae. Available from:
Cowling, R.M., Rundel, P.W., Desmet, P.G. & Esler, K.J. (1998) Extraordi- https://doi.org/10.15468/dl.vr67q8 [Accessed 24th July 2023]
nary high regional-scale plant diversity in southern African arid lands: GBIF.org. (2023e) GBIF occurrence download, Apidae. Available from:
subcontinental and global comparisons. Diversity and Distributions, 4, https://doi.org/10.15468/dl.wnamdv [Accessed 24th July 2023]
27–36. GBIF.org. (2023f) GBIF Occurrence download, Halictidae. Available from:
Crisp, M.D., Laffan, S., Linder, H.P. & Monro, A. (2001) Endemism in the https://doi.org/10.15468/dl.7rjdz2 [Accessed 24th July 2023]
Australian flora. Journal of Biogeography, 28, 183–198. Gess, S.K. (1992) Biogeography of the Masarine wasps (Hymenoptera:
Critchlow, R., Cunningham, C.A., Crick, H.Q.P., Macgregor, N.A., Vespidae: Masarinae), with particular emphasis on the southern Afri-
Morecroft, M.D., Pearce-Higgins, J.W. et al. (2022) Multi-taxa spatial can taxa and on correlations between Masarine and forage plant dis-
conservation planning reveals similar priorities between taxa and tributions. Journal of Biogeography, 19, 491.
improved protected area representation with climate change. Biodi- Gess, S.K. & Gess, F.W. (2014) Wasps and bees in southern Africa. Pretoria:
versity and Conservation, 31, 683–702. SANBI Biodiversity Series, SANBI.
Danforth, B.N., Cardinal, S., Praz, C., Almeida, E.A.B. & Michez, D. (2013) Gordon, M.L., Couldridge, V.C.K., Engelbrecht, A. & Colville, J.F. (2023)
The impact of molecular data on our understanding of bee phylogeny Delimiting zoogeographic centres for South African Orthoptera.
and evolution. Annual Review of Entomology, 58, 57–78. Frontiers of Biogeography, 15, 1948–6596.
Danforth, B.N., Sipes, S., Fang, J. & Brady, S.G. (2006) The history of early Gotelli, N.J. & Colwell, R.K. (2001) Quantifying biodiversity: procedures
bee diversification based on five genes plus morphology. Proceedings and pitfalls in the measurement and comparison of species richness.
of the National Academy of Sciences of the United States of America, Ecology Letters, 4, 379–391.
103, 15118–15123. Graham, C.H., Ferrier, S., Huettman, F., Moritz, C. & Peterson, A.T. (2004)
Davis, A.L.V. (2002) Dung beetle diversity in South Africa: influential fac- New developments in museum-based informatics and applications in
tors, conservation status, data inadequacies and survey design. Afri- biodiversity analysis. Trends in Ecology and Evolution, 19, 497–503.
can Entomology, 10, 53–65. Graler, B., Pebesma, E. & Heuvelink, G. (2016) Spatio-temporal interpola-
de Palma, A., Abrahamczyk, S., Aizen, M.A., Albrecht, M., Basset, Y., tion using gstat. R Journal, 8, 204–218.
Bates, A. et al. (2016) Predicting bee community responses to land- Hamer, M. (2012) An assessment of zoological research collections in
use changes: effects of geographic and taxonomic biases. Scientific South Africa. South African Journal of Science, 108, 1–11.
Reports, 6, 31153. Hedtke, S.M., Patiny, S. & Danforth, B.N. (2013) The bee tree of life: a
Deacon, C., Samways, M.J. & Pryke, J.S. (2020) Determining drivers of supermatrix approach to apoid phylogeny and biogeography. BMC
dragonfly diversity patterns and the implications for conservation in Evolutionary Biology, 13, 1–13.
South Africa. Biological Conservation, 245, 108548. Hengl, T., Sierdsema, H., Radovic, A. & Dilo, A. (2009) Spatial prediction of
Dijkstra, K.-D.B. (2007) Demise and rise: the biogeography and taxonomy species’ distributions from occurrence-only records: combining point
of the Odonata of tropical Africa. PhD thesis, Leiden University, pattern analysis, ENFA and regression-kriging. Ecological Modelling,
pp. 143–187. 220, 3499–3511.
Dirsch, V.M. (1965) Revision of the family Pneumoridae (Orthoptera: Hévin, N.M.C., Hansen, S., Addison, P., Benoit, L., Kergoat, G.J. & Haran, J.
Acridoidea). Bulletin of the British Museum (Natural History), 15, (2022) Late Cenozoic environmental changes drove the diversifica-
325–396. tion of a weevil genus endemic to the Cape Floristic Region. Zoolo-
Eardley, C.D. (1989) The Afrotropical species of Eucara Friese, Tetralonia gica Scripta, 51, 724–740.
Spinola and Tetraloniella Ashmead (Hymenoptera: Anthophoridae). Kaloveloni, A., Tscheulin, T. & Petanidou, T. (2018) Geography, cli-
Entomology Memoir Department of Agriculture and Water Supply mate, ecology: what is more important in determining bee diver-
Republic of South Africa, 75, 1–62. sity in the Aegean Archipelago? Journal of Biogeography, 45,
Eardley, C.D. (1991) The Melectini in subsaharan Africa. Entomology Mem- 2690–2700.
oir Department of Agricultural Development Republic of South Africa, Kearns, C.A. & Inouye, D.W. (1997) Pollinators, flowering plants, and con-
82, 1–49. servation biology. Bioscience, 47, 297–307.
Eardley, C.D. & Coetzer, W. (2023) Catalogue of Afrotropical bees. Version Klein, A.M., Vaissière, B.E., Cane, J.H., Steffan-Dewenter, I.,
1.22. Agricultural Research Council. Checklist dataset. 8 July 2023. Cunningham, S.A., Kremen, C. et al. (2007) Importance of pollinators
https://doi.org/10.15468/u9ezbh in changing landscapes for world crops. Proceedings of the Royal Soci-
Eardley, C. & Urban, R. (2010). Catalogue of Afrotropical bees (Hymenop- ety B: Biological Sciences, 274, 303–313.
tera: Apoidea: Apiformes). Zootaxa, 2455, 1–548. https://doi.org/10. Kougioumoutzis, K., Kaloveloni, A. & Petanidou, T. (2022) Assessing cli-
11646/zootaxa.2455.1.1 mate change impacts on Island bees: the Aegean Archipelago. Biol-
Elith, J. & Leathwick, J.R. (2009) Species distribution models: ecological ogy, 11, 552.
explanation and prediction across space and time. Annual Review of Kreft, H. & Jetz, W. (2007) Global patterns and determinants of vascular
Ecology, Evolution, and Systematics, 40, 677–697. plant diversity. Proceedings of the National Academy of Sciences of the
Garibaldi, L.A., Steffan-Dewenter, I., Winfree, R., Aizen, M.A., United States of America, 104, 5925–5930.
Bommarco, R., Cunningham, S.A. et al. (2013) Wild pollinators Kuhlmann, M. (2009) Patterns of diversity, endemism and distribution of
enhance fruits set of crops regardless of honey bee abundance. Sci- bees (Insecta: Hymenoptera: Anthophila) in southern Africa.
ence, 339, 1608–1611. South African Journal of Botany, 75, 726–738.
Gaston, K.J. (2000) Global patterns in biodiversity. Nature, 405, 220–227. Kuhlmann, M., Guo, D., Veldtman, R. & Donaldson, J. (2012) Conse-
GBIF.org. (2023a) GBIF occurrence download, Colletidae. Available from: quences of warming up a hotspot: species range shifts within a cen-
https://doi.org/10.15468/dl.gktdhq [Accessed 24th July 2023] tre of bee diversity. Diversity and Distributions, 18, 885–897.
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
486 MELIN ET AL.

Leclercq, N., Marshall, L., Caruso, G., Schiel, K., Weekers, T., Otte, D. (2020) Thirty-two new south African grasshopper species in the
Carvalheiro, L.G. et al. (2023) European bee diversity: taxonomic family Lentulidae belonging to eleven genera (Orthoptera: Acridoi-
and phylogenetic patterns. Journal of Biogeography, 50, 1244– dea). Transactions of the American Entomological Society, 146, 1–164.
1256. Patiny, S. & Michez, D. (2007) Biogeography of bees (Hymenoptera, Apoi-
Litman, J.R., Danforth, B.N., Eardley, C.D. & Praz, C.J. (2011) Why do leaf- dea) in Sahara and the Arabian deserts. Insect Systematics and Evolu-
cutter bees cut leaves? Bew insights into the early evolution of bees. tion, 38, 19–34.
Proceedings of the Royal Society B: Biological Sciences, 278, 3593– Pearce, J.L. & Boyce, M.S. (2006) Modelling distribution and abundance
3600. with presence-only data. Journal of Applied Ecology, 43, 405–412.
Manning, J.C. & Goldblatt, P. (2012) Plants of The Greater Cape Floristic Perera, S.J., Ratnayake-Perera, D. & Procheş, Ş. (2011) Vertebrate distribu-
Region 1: the Core Cape Flora. Strelitzia, Vol. 29. Pretoria: SANBI. tions indicate a greater Maputaland-Pondoland-Albany region of
Marshall, L., Ascher, J.S., Villagra, C., Beaugendre, A., Herrera, V., endemism. South African Journal of Science, 107, 1–15.
Henríquez-Piskulich, P. et al. (2023) Chilean bee diversity: contrast- Potts, A.J., Hedderson, T.A., Franklin, J. & Cowling, R.M. (2013) The last
ing patterns of species and phylogenetic turnover along a large-scale glacial maximum distribution of South African subtropical thicket
ecological gradient. Ecosphere, 14, e4535. inferred from community distribution modelling. Journal of Biogeogra-
Mecenero, S., Altwegg, R., Colville, J.F. & Beale, C.M. (2015) Roles of spa- phy, 40, 310–322.
tial scale and rarity on the relationship between butterfly species Predel, R., Neupert, S., Huetteroth, W., Kahnt, J., Waidelich, D. & Roth, S.
richness and human density in South Africa. PLoS One, 10, (2012) Peptidomics-based phylogeny and biogeography of Manto-
e0124327. phasmatodea (Hexapoda). Systematic Biology, 61, 609–629.
Mecenero, S., Ball, J.B., Edge, D.A., Hamer, M.L., Henning, G.A., Kruger, M. Procheş, Ş. & Cowling, R.M. (2006) Insect diversity in Cape Fynbos and
et al. (2013) In: Terblanche, R.F. & Williams, M.C. (Eds.) Conservation neighbouring South African vegetation. Global Ecology and Biogeogra-
assessment of butterflies of South Africa. Lesotho and Swaziland: Red phy, 15, 445–451.
List and Atlas. Safronics (Pty) Ltd., Johannesburg & Animal Demogra- Procheş, Ş. & Cowling, R.M. (2007) Do insect distributions fit our biomes?
phy Unit, Cape Town. South African Journal of Science, 103, 258–261.
Meiners, J.M., Griswold, T.L. & Messinger Carril, O. (2019) Decades of Procheş, Ş., Forest, F., Veldtman, R., Chown, S.L., Cowling, R.M.,
native bee biodiversity surveys at Pinnacles National Park highlight Johnson, S.D. et al. (2009) Dissecting the plant-insect diversity relation-
the importance of monitoring natural areas over time. PLoS One, 14, ship in the Cape. Molecular Phylogenetics and Evolution, 51, 94–99.
e0207566. QGIS Development Team. (2023) QGIS Geographic Information System.
Melin, A. & Colville, J.F. (2019) A review of 250 years of South African bee Open Source Geospatial Foundation Project. (Version 2.18.14).
taxonomy and exploration (Hymenoptera: Apoidea: Anthophila). Available from: https://qgis.org
Transactions of the Royal Society of South Africa, 74, 86–96. R Core Team. (2021) R: a language and environment for statistical comput-
Messinger Carril, O., Griswold, T.L., Haefner, J. & Wilson, J.S. (2018) ing. Vienna, Austria: R Foundation for Statistical Computing. Avail-
Wild bees of Grand Staircase-Escalante National Monument: rich- able from: https://www. R-project.org/
ness, abundance, and spatio-temporal beta-diversity. PeerJ, 2018, Ricklefs, R.E. (2004) A comprehensive framework for global patterns in
e5867. biodiversity. Ecology Letters, 7, 1–15.
Michener, C.D. (1979) Biogeography of the bees. Annals of the Missouri Robertson, M.P., Visser, V. & Hui, C. (2016) Biogeo: an R package for
Botanical Garden, 66, 277–347. assessing and improving data quality of occurrence record datasets.
Michener, C.D. (2007) The bees of the world, 2nd edition. Baltimore, MD: Ecography, 39, 394–401.
Johns Hopkins Press. Rosenzweig, M.L. (1995) Species diversity in space and time. Cambridge
Michez, D., Patiny, S. & Danforth, B.N. (2009) Phylogeny of the bee family University Press, Cambridge, United Kingdom.
Melittidae (Hymenoptera: Anthophila) based on combined molecular Scholtz, C.H. & Chown, S.L. (1995) Insects in southern Africa: how many
and morphological data. Systematic Entomology, 34, 574–597. species are there? South African Journal of Science, 91, 124–126.
Michez, D., Patiny, S., Rasmont, P., Timmermann, K. & Vereecken, N.J. Schulze, R.E. (2007) South African atlas of climatology and agrohydrology.
(2008) Phylogeny and host-plant evolution in Melittidae Water Research Commission, Pretoria, RSA, WRC Report
s.l. (Hymenoptera: Apoidea). Apidologie, 39, 146–162. 1489/1/06. Pretoria.
Minckley, R.L. & Radke, W.R. (2021) Extreme species density of bees Schwarz, M.P., Bull, N.J. & Cooper, S.J.B. (2003) Molecular phylogenetics
(Apiformes, Hymenoptera) in the warm deserts of North America. of Allodapine bees, with implications for the evolution of sociality
Journal of Hymenoptera Research, 82, 317–345. and progressive rearing. Systematic Biology, 52, 1–14.
Mittermeier, R.A., Myers, N., Tliomsen, J.B. & Olivieri, S. (1998) Biodiver- Skowno, A.L., Poole, C.J., Raimondo, D.C., Sink, K.J., van Venter, H. & van
sity hotspots and major tropical wilderness areas: approaches to set- Niekerk, L. (2018) National Biodiversity Assessment 2018: the status of
ting conservation priorities. Conservation Biology, 12, 516–520. South Africa’s ecosystems and biodiversity—synthesis report. Pretoria:
Mucina, L. & Rutherford, M.C. (2006) The vegetation of South Africa. Strelit- SANBI.
zia, Vol. 19. Lesotho and Swaziland: South African National Biodiver- Soroye, P., Newbold, T. & Kerr, J.T. (2020) Climate change contributes to
sity Institute, Pretoria. widespread declines among bumble bees across continents. Science,
Myers, N., Mittermeier, R.A., Mittermeier, C.G., Fonseca, G.A.B. & Kent, J. 367, 685–688.
(2000) Biodiversity hotspots for conservation priorities. Nature, 403, Steiner, K.E. & Whitehead, V.B. (1990) Pollinator adaptation to oil-
853–858. secreting flowers—Rediviva and Diascia. Evolution, 44, 1701.
Neff, J.L. & Simpson, B.B. (1993) Bees, pollination systems and plant diver- Stuckenberg, B.R. (1969) Effective temperature as an ecological factor in
sity. In: la Salle, J. & Gauld, I.D. (Eds.) Hymenoptera and biodiversity. southern Africa. Zoologica Africana, 4, 145–197.
Wallington: CAB International, pp. 143–167. van Wyk, A.E. & Smith, G.F. (2001) Regions of floristic endemism in south-
Ollerton, J., Winfree, R. & Tarrant, S. (2011) How many flowering plants ern Africa. A review with emphasis on succulents. Pretoria: Umdaus
are pollinated by animals? Oikos, 120, 321–326. Press.
Orr, M.C., Hughes, A.C., Chesters, D., Pickering, J., Zhu, C.-D. & Ascher, J.S. Whitehead, V.B. & Steiner, K.E. (1993) A new Rediviva bee (Hymenoptera:
(2021) Global patterns and drivers of bee distribution. Current Biology, Apoidea: Melittidae) that collects oil from orchids. African Entomol-
31, 451–458. https://doi.org/10.1016/j.cub.2020.10.053 ogy, 1, 159–166.
 17524598, 2024, 3, Downloaded from https://resjournals.onlinelibrary.wiley.com/doi/10.1111/icad.12715 by South African Medical Research, Wiley Online Library on [31/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
HOTSPOTS OF SOUTH AFRICAN BEE DIVERSITY 487

Wickham, H. (2009) ggplot2: elegant graphics for data analysis. Springer- SUPPORTING INF ORMATION
Verlag New York. Additional supporting information can be found online in the Support-
Wiens, J.J. (2011) The causes of species richness patterns across space,
ing Information section at the end of this article.
time, and clades and the role of “Ecological Limits”. The Quarterly
Review of Biology, 86, 75–96. Data S1. Supporting Information.
Williams, P.H. (1998) An annotated checklist of bumble bees with an
analysis of patterns of description (Hymenoptera: Apidae, Bom-
bini). Bulletin of the Natural History Museum Entomology Series, 67, How to cite this article: Melin, A., Beale, C.M., Manning, J.C. &
79–152. Colville, J.F. (2024) Fine-scale bee species distribution models:
Wright, M.G. & Samways, M.J. (1998) Insect species richness tracking
Hotspots of richness and endemism in South Africa with
plant species richness in a diverse flora: gall-insects in the Cape Flo-
ristic Region, South Africa. Oecologia, 115, 427–433.
species-area comparisons. Insect Conservation and Diversity, 17(3),
Zattara, E.E. & Aizen, M.A. (2021) Worldwide occurrence records suggest 474–487. Available from: https://doi.org/10.1111/icad.12715
a global decline in bee species richness. One Earth, 4, 114–123.