Vertebrate distributions indicate a greater Maputaland-Pondoland-Albany region of endemism

How to cite this article: Perera SJ, Ratnayake-Perera D, Procheş Ş. Vertebrate distributions indicate a greater MaputalandPondoland-Albany region of endemism. S Afr J Sci. 2011;107(7/8), Art. #462, 15 pages. doi:10.4102/sajs. v107i7/8.462 The Maputaland-Pondoland-Albany (MPA) biodiversity hotspot (~274 316 km2) was primarily recognised based on its high plant endemism. Here we present the results of a qualitative biogeographical study of the endemic vertebrate fauna of south-eastern Africa, in an exercise that (1) refines the delimitation of the MPA hotspot, (2) defines zoogeographical units and (3) identifies areas of vertebrate endemism. Initially we listed 62 vertebrate species endemic and 60 near endemic to the MPA hotspot, updating previous checklists. Then the distributions of 495 vertebrate taxa endemic to south-eastern Africa were reviewed and 23 endemic vertebrate distributions (EVDs: distribution ranges congruent across several endemic vertebrate taxa) were recognised, amongst which the most frequently encountered were located in the Eastern Escarpment, central KwaZulu-Natal, Drakensberg and Maputaland. The geographical patterns illustrated by the EVDs suggest that an expansion of the hotspot to incorporate sections of the Great Escarpment from the Amatola-Winterberg-Sneeuberg Mountains through the Drakensberg to the Soutpansberg would be justified. This redefinition gives rise to a Greater Maputaland-Pondoland-Albany (GMPA) region of vertebrate endemism adding 135% more endemics with an increase of only 73% in surface area to the MPA hotspot. The GMPA region has a more natural boundary in terms of EVDs as well as vegetation units. An accurate delimitation of this hotspot, as well as a better understanding of biogeography in the region, would greatly benefit conservation planning and implementation. Towards these aims, we used EVDs to delimit non-overlapping zoogeographical units (including 14 areas of vertebrate endemism), facilitating numerical biogeographical analyses. More importantly, this study opens up possibilities of refining hotspot delimitation and identifying local conservation priorities in regions of the world where data do not allow numerical analyses.


Introduction
Despite criticism over their selection and delimitation, biodiversity hotspots 1,2 have undeniably become a popular approach for prioritising conservation efforts globally, 3 as well as in South Africa. 4Of the 34 global biodiversity hotspots, 3 are either within South Africa or extend marginally into neighbouring countries: the Cape Floristic Region, the Succulent Karoo and the Maputaland-Pondoland-Albany (MPA).In a broader context, southern Africa, defined as the area south of the Cunene, Okavango and Zambezi Rivers, 5 fully encompasses these three hotspots together with the southern parts of the Coastal Forests of Eastern Africa hotspot. 1 The major biological criterion for the designation of biodiversity hotspots is floristic endemism, that is, the area must contain at least 0.5% of the world's vascular plant species (1500 species) as endemics. 1,2his means that animal endemism per se is not critical for hotspot selection, although vertebrates are most likely to become hotspot flagship species. 6e MPA biodiversity hotspot stretches along the eastern coast of southern Africa from Maputo in the north-east to Port Elizabeth in the south-west, extending inland towards the Great Escarpment. 7In compliance with the criteria for defining biodiversity hotspots, it is an important centre of plant endemism, being home to 1900 endemic species. 5,7The delimitation of floristic regions by White 8 and van Wyk and Smith 5 provided the basis for the original recognition of this hotspot.White's Tongaland-Pondoland regional mosaic has been extended further inland by van Wyk and Smith 5 as the Maputaland-Pondoland region of floristic endemism, to include the Afromontane elements below 1800 m a.s.l.along the Great Escarpment.In delimiting the MPA hotspot, Mittermeier et al. 2 followed the suggestion of van Wyk and Smith 5 and extended the Maputaland-Pondoland region in a south-westerly direction to the Albany centre, giving rise to a hotspot encompassing all three centres (Maputaland, Pondoland and Albany).Nevertheless, the hotspot has not yet been well documented in terms of its animal endemism, except for the species accounts given by Mittermeier et al. 2 , those in the terrestrial vertebrate species database of Conservation International 9 and in the draft ecosystem profile of the hotspot. 10en though the vertebrate fauna of southern Africa is relatively well studied, greater emphasis on vascular plants (with 60% species endemism) 5 and perhaps on large widespread mammals has overshadowed the broader picture of southern African vertebrate endemism (especially with regard to smaller vertebrates) in the scientific literature.A relatively strong correlation in global endemism patterns has been observed between plants and vertebrates, 11 although the congruence is not always high at regional scales (e.g.eastern Madagascar 12 and the Cape Floristic Region 13 ).In the tropical forest hotspots of Africa (i.e. the Eastern Arc and the Coastal Forests of Eastern Africa), the patterns of plant endemism are matched by those of vertebrates, but this congruence is lower in the Cape Floristic Region, which has a Mediterraneantype climate. 1 It is yet to be tested whether the floristic endemism in south-eastern Africa, ranging from subtropical to temperate latitudes, will show congruence with patterns of vertebrate endemism.
Early phytogeographical studies of Africa have used intuitive chorological approaches to delimit floristic regions (phytochoria) and centres of floristic endemism, 5,8 and recent studies have tested their validity using multivariate cluster analyses. 14,15,16In turn, vertebrate biogeography has addressed patterns in different vertebrate classes across Africa, initially by grouping species ranges qualitatively through visual sorting and matching, 17,18 and later by numerical cluster analyses on distribution data. 19,20,21,22,23,24,25owever, the databases used for most of the numerical analyses were incomplete and/or biased because various subjective factors were involved as a result of the lack of uniform sampling throughout the study areas.All these analyses were based on distribution patterns of taxa in single vertebrate classes (frogs, 17,18,23,25 large mammals 22 and birds 19,20,21,24 ).Crowe 26 summarised faunal zones for several terrestrial vertebrate groups through numerical analyses, but used larger and different grid quadrats (~110 km × 110 km for water birds, ~220 km × 220 km for frogs, snakes and lizards, and ~400 km × 400 km for larger mammals and terrestrial birds), which may have been too coarse to pick up patterns of narrow endemism.No attempt has yet been made to delimit zoogeographical units or areas of endemism for the entire vertebrate fauna of southern Africa.
Therefore, this paper documents vertebrate diversity and endemism in the MPA hotspot as currently defined, using the latest available literature.Subsequently, distribution ranges that are congruent across several endemic vertebrate taxa in south-eastern Africa are identified and used (1) to discuss the adequacy of the current boundary of the MPA hotspot, and (2) to propose a set of non-overlapping geographical units for a rigorous numerical analysis of south-east African zoo-(bio-) geography.The results we present here are not based on any numerical analysis, and as such our methods can be replicated in regions of the world where data are scarce.

Study area
This paper focuses on south-eastern Africa, delimited in the west by the Nelspoort interval in the southern Great Escarpment 27,28 (about 24º E), and in the north by the northernmost loop of the Limpopo valley (about 22º S).The study area fully encompassed the MPA hotspot 7 and the Maputaland, Pondoland, Albany, Drakensberg, Barberton, Wolkberg, Sekhukhuneland and Soutpansberg centres of floristic endemism, 5 together with the Highveld and Bushveld bioregions, 29 to the west and north-west of the MPA, respectively, and the Lowveld and Mopane bioregions 29 in the north-east.

Assessment of vertebrate endemism
Vertebrate species richness and endemism within the MPA hotspot, and endemism only for south-eastern Africa, were assessed based on distribution data available in the literature.We used the latest available atlases as the primary sources of data for amphibians, reptiles and birds (with data at quarter degree square 30 scale, hereafter QDS), supplemented with other data where QDS data were not available.The use of data other than atlases is of particular importance (1) as sources of distribution data beyond the atlas regions of 'South Africa, Lesotho and Swaziland' for the herpetofauna, (2) to assess their endemicity in south-eastern Africa (see below for our definition of south-east African endemism) and (3) to assess whether the atlas records are of breeding populations or migrants in the case of birds.Freshwater fish distributions were sourced from Skelton 31,32 and further validated using FishBase (Froese and Pauly 33 ).The amphibian atlas of Minter et al. 34 , mirrored by the online Virtual Museum of the Animal Demography Unit, University of Cape Town, was supplemented with data from du Preez and Carruthers 35 and similarly the reptile atlas of the Southern African Reptile Conservation Assessment, available online (Animal Demography Unit 36 ) was supplemented with data from Branch 37 .The original bird atlas of Harrison et al. 38 , and updated atlas-based maps in Hockey et al. 39 were used as sources of bird data, whilst the comprehensive treatment by Skinner and Chimimba 40 was used for mammals, as the mammal data in the online Virtual Museum of the Animal Demography Unit is as yet far from complete.Museum databases were not considered in the cases of freshwater fish and mammals, as they are particularly incomplete and collection biases are prominent, especially in freshwater fish.Species taxonomy follows the latest treatment mentioned above for each taxon, that is Froese and Pauly 33 for freshwater fish, du Preez and Carruthers 35 for amphibians, Animal Demography Unit 36 (as at 31 January 2010) for reptiles, updated following Kelly et al 41 for the family Lamprophiidae, Hockey et al. 39 for birds, and Skinner and Chimimba 40 for mammals, whilst other sources 42,43,44 were used in updating familial-level taxonomy.

Vertebrate species richness and endemism in the MPA hotspot
We compiled checklists of vertebrate species for the MPA hotspot as currently defined, updated those available in the terrestrial vertebrate species database of Conservation International, 9 and produced a first checklist of freshwater fish species for the region.The list also includes the secondary freshwater species (coastal/estuarine species that also occur in fresh water, excluding the stragglers or sporadic marine fishes that are sometimes found living, but not breeding, in inland waters 31 ).Besides endemic species, near-endemic species were also listed for the hotspot, as they are important in recognising biogeographical links between the hotspot and surrounding regions.MPA near endemics were defined as those species with more than 50% of their range within the hotspot, and extending outside the hotspot only marginally (e.g.bush blackcap, Lioptilus nigricapillus and Natal red rock rabbit, Pronolagus crassicaudatus) and/or with small distant satellite populations outside it (e.g.Natal mountain catfish, Amphilius natalensis; Phongolo suckermouth, Chiloglanis emarginatus; and laminate vlei rat, Otomys laminatus).In order to minimise subjectivity in including or excluding marginal species, those that occurred along the boundary of the hotspot were included only if they showed a prominent association with habitat types that occur within the hotspot.

Vertebrate taxa endemic to south-eastern Africa
In order to correct for possible errors induced by having a latitude-longitude boundary for the study area when dealing with natural distribution ranges, we defined south-east African endemics as taxa with more than 50% of their extent of occurrence (at the QDS scale when data were available and approximately elsewhere) south of 22˚S and east of 24˚E.Therefore, even the taxa that cross 22˚S and 24˚E latitudelongitude boundaries were included in the analysis, as long as they were endemic to south-eastern Africa in a broader sense, and met the above criterion.Taxa endemic to the Cape Floristic Region and the Coastal Forests of Eastern Africa hotspots were excluded.
A database was compiled comprising all taxa endemic to south-eastern Africa.Taxa referred to here are species, whilst infraspecific taxa (subspecies, subspecies complexes and populations) were also taken into account wherever there was a considerable geographical disjunction.More specifically, a disjunct subspecies or subspecies complex isolated by a gap of more than one degree (~100 km) or a disjunct population (not recognised as a subspecies) isolated by a gap of more than 2 degrees (~200 km) or (in the case of freshwater fish) when found occurring in different major river systems that were not in contact under current drainage patterns, were included as a separate taxa.An exception was made for subspecies of range-restricted species endemic to south-eastern Africa, where subspecies were accepted as disjunct with a gap of more than 50 km.
The extent of occurrence of a given taxon was determined using records of naturally occurring breeding populations, excluding migrant and vagrant records as well as introduced (deliberately or unintentionally) and relocated populations extralimital to their natural range.Marine taxa were excluded from the analysis.The separation of terrestrial and freshwater taxa from marine taxa was self-explanatory in the case of marine mammals, pelagic sea birds and marine reptiles.In the case of fish, all taxa listed by Skelton 31 as freshwater or secondary freshwater taxa were considered.Any ambiguity in taxonomy or distribution, and difficulties in delimiting range boundaries (e.g. because of recent taxonomic revisions or single historical records not verified by recent extensive surveys) resulted in disqualifying the taxon or a particular record.The undescribed species listed by the Animal Demography Unit 36 and du Preez and Carruthers 35 were included when distribution data were provided and matched our criteria.
Precautions were taken when plotting bird distributions to minimise the bias caused by their higher dispersal ability compared to other vertebrates.Therefore, single outlying occupied QDSs were considered to be part of a taxon's extent of occurrence only if the reporting rate was more than 2% (see methods in Harrison et al. 38 ).Single outlying occupied QDSs more than 200 km away from core populations were omitted from the extent of occurrence, and not considered as isolated populations.Isolated records/subpopulations of bird species that were widely but sparsely distributed, especially in the case of aquatic birds (e.g.spotted crake, Porzana pusilla and lesser jacana, Microparra capensis) were not regarded as disjunct.Bats (Order Chiroptera) were treated only at the species level, considering isolated records as parts of a single scattered population because they are long-distance flyers that are not easily detected and hence may have been overlooked.Disjunct populations of small mammals were also disregarded when disjunctions were likely to be the result of insufficient observations.

Identification of endemic vertebrate distributions
Distribution ranges that are congruent across several endemic vertebrate taxa are defined as endemic vertebrate distributions (EVDs).They often overlap each other, and hence cannot be used as exclusive biogeographical units, or considered for a hierarchical geographical analysis, but serve here as a first step in the qualitative grouping of distribution ranges.The EVDs were used in (1) identifying spatial relationships amongst endemic vertebrate assemblages, in an effort to refine the delimitation of the MPA hotspot and (2) delimiting non-overlapping geographical units for a hierarchical biogeographical analysis.
EVDs were detected through an extensive review of the distribution data followed by plotting, visual sorting and matching of distribution ranges.This qualitative method was developed based on the classical biogeographical methods of White 8 and van Wyk and Smith 5 for the flora, and Poynton 17,18 for amphibians.The subjectivity of the results obtained through this method can be reduced with wider experience, a critical outlook and possession of a good eye and memory. 45hilst White 45 emphasises the importance of subsequent rigorous analysis for intuitively defined biogeographical regions, subsequent analytical confirmations of such regions (e.g.Linder 14 and Linder et al. 15 for White's 8 phytochoria of Africa; Procheş 46 and Kreft and Jetz 47 for Wallace's 48 zoographical regions of the world) have added credibility to such methods.Where our study area is concerned, even though Poynton's 18 intuitively defined regions of amphibian biogeography are not exactly matched by those identified in the subsequent numerical analysis of Alexander et al. 25 , the general biogeographical patterns were largely congruent.Further to that, it has been noted 49 that the frog atlas 34 representing the database for Alexander et al. 25 has some serious sampling biases and is of limited use for fine-scale analysis.Alexander et al. 25 themselves accepted this as a fact, using half-degree squares for their analysis, whilst the atlas provides data at the QDS scale.Thus, the discrepancies between Poynton's 18 and Alexander et al.'s 25 results may be attributed to data-related limitations, rather than to different methods.
Once the EVDs were delimited, every taxon selected for the study was then assigned to the narrowest of the EVDs encompassing all of its accepted breeding records, to assess the frequency of occupation of each EVD.The EVDs were then used in refining the delimitation of the MPA hotspot.

Delimitation of zoogeographical units
A set of non-overlapping zoogeographical units was designed based on (1) the core regions of EVDs (delimited adopting Linder and Mann's 50 method of 'mapping individual species distributions and seeking areas of overlap for range restricted taxa'), ( 2) the overlapping margins of EVDs and (3) the patterns of narrower endemism within broader EVDs.When borders of adjacent units thus delimited did not align with each other, the coverage of relevant EVDs was taken into consideration, and the borders were further adjusted using the boundaries of biomes and bioregions. 29Amongst the zoogeographical units delimited, those harbouring two or more narrow-range endemic vertebrate species characteristic of them were treated here as areas of vertebrate endemism (AOVEs), following Nelson and Platnick's 51 definition for 'areas of endemism'.

Vertebrate species richness and endemism in the Maputaland-Pondoland-Albany biodiversity hotspot as currently defined
A list of 1217 regularly occurring indigenous species of vertebrates (representing 586 genera in 184 families), including 62 (5.1%) endemic and 60 (4.9%) near-endemic species is provided for the MPA hotspot (Table 1).In addition, at least 51 bird species were recorded from the hotspot as vagrants, and at least 25 alien species were introduced to the hotspot and have well-established populations.Birds show the highest species richness within the MPA hotspot (668 naturally occurring species including vagrants), whilst the herpetofauna show the highest endemism (21.3% in amphibians and 14.3% in reptiles).Endemic and nearendemic vertebrates of the MPA hotspot are listed in Appendix 1, whilst a complete checklist of all vertebrates occurring in the hotspot is available as online supplementary material to this paper (vagrants and alien species excluded).
The ichthyofauna of the MPA hotspot includes narrowrange endemics such as the Amatola barb, Barbus amatolicus and the Border barb, B. trevelyani of the Amatola Mountains, the Eastern Cape rocky, Sandelia bainsii of Albany, the redtail barb, Barbus gurneyi and Tugela labeo, Labeo rubromaculatus of KwaZulu-Natal and the pennant-tailed suckermouth, Chiloglanis anoterus of Maputaland, together with the secondary freshwater endemics Sibayi goby, Silhouettea sibayi in Maputaland and the golden sleeper, Hypseleotris dayi along the coast of Maputaland and KwaZulu-Natal.The barbs in genus Barbus show the highest diversity and endemicity amongst fishes in the MPA hotspot with 18 indigenous species, 3 of which (mentioned above) are endemic and 2 are near endemic to the hotspot.The amphibian fauna of the MPA hotspot includes two endemic genera: the monotypic genus Natalobatrachus represented by the kloof frog, N. bonebergi (restricted to the KwaZulu-Natal coastal belt and Pondoland), and the genus Anhydrophryne, represented by the Natal chirping frog, A. hewitti and the mistbelt chirping frog, A. ngongoniensis, narrow-range endemics in KwaZulu-Natal and the KwaZulu-Natal Midlands, respectively, together with the Hogsback chirping frog, A. rattrayi, largely restricted to the Amatola Mountains.Amongst the reptiles, the monotypic endemic genus Macrelaps is represented by the Natal black snake, M. microlepidotus, restricted to Maputaland, KwaZulu-Natal and Podoland.Prominent reptile genera showing a great deal of diversification and narrow endemism within the hotspot are flat geckos in the genus Afroedura (15 species: 6 endemic and 7 near endemic), dwarf burrowing skinks in the genus Scelotes (11 species: 7 endemic and 3 near endemic) and dwarf chameleons in the genus Bradypodion (9 species: 8 endemic and 1 near endemic).Even though the avifauna shows the highest species richness in the MPA hotpsot, there are no endemic birds in the hotspot.However, 14 bird species are listed in Appendix 1 as near endemic to the MPA hotspot, with ranges extending towards the southeastern escarpment and/or Knysna.The hotspot also forms a part of BirdLife International's 'South-east African Coast' endemic bird area 52 with species extending north from the MPA hotspot along the coastal belt.The only three species of mammals endemic to the MPA hotspot are two golden moles (the giant golden mole, Chrysospalax trevelyani and Marley's golden mole, Amblysomus marleyi, narrowly endemic to the Transkei coastal belt and Maputaland, respectively), and one shrew (Sclater's forest shrew, Myosorex sclateri in KwaZulu-Natal).However, the numerous non-endemic species make the hotspot an important area for small mammals, especially for golden moles (with seven species representing five genera, two of which are endemic and one is near endemic in the MPA hotspot).

Vertebrate taxa endemic to south-eastern Africa
A total of 495 vertebrate taxa (259 species and 236 infraspecific taxa) were found to have restricted ranges endemic to southeastern Africa (see last panel in Figure 1), making the area an important region of vertebrate endemism, especially for reptiles (211 endemic taxa).Birds, as expected, showed higher endemism in infraspecific taxa (88) than in species (44), as their mobility retards isolation-driven speciation, whilst greater numbers of endemic species compared to infraspecific taxa were recorded for the herpetofauna (with endemic species: endemic infraspecific taxa ratios of 131:80 in reptiles and 33:09 in amphibians), although it can be argued that many more taxa await description.

Endemic vertebrate distributions
A total of 23 EVDs were identified within the study area based on the distribution ranges of 495 endemic taxa.Figure 1

The greater Maputaland-Pondoland-Albany region of vertebrate endemism
There are seven narrow-range EVDs in south-eastern Africa: Maputaland, KwaZulu-Natal, Transkei Coastal Belt and Albany Coastal Belt along the south-eastern coast, and the Eastern Escarpment, Drakensberg and Amatola-Winterberg-Sneeuberg along the south-eastern Great Escarpment (Figure 2a).he area added by the Inhambane and Mopane and the Knysna transitional extensions overlaps with the Lowveld and Bushveld bioregions, and the Cape Floristic Region, respectively.If the overlaps were reduced to include only forest patches in these areas (where most of the elements characteristic of the GMPA region are found), then the actual increase in area would be even smaller.Even if the GMPA region excluded these transitional extensions, it would still include 125 endemic species (a 103% increase in endemism from the MPA hotspot), with an increase in area of only 50%.
It is particularly remarkable that the GMPA region harbours 15 endemic bird species in contrast to the MPA hotspot, which has no endemic bird species.Three reptile genera extralimital to the MPA hotspot (genera of the Woodbush legless skink, Acontophiops; the Swazi rock snake, Inyoka and the cream-spotted mountain snake, Montaspis), together with the amphibian genus of the Natal cascade frog Hadromophryne and the bird genus of the bush blackcap Lioptilus, all monotypic and extending towards the escarpment from the MPA hotspot, are endemic within the GMPA region.This results in a total of eight genera (three amphibian genera -Anhydrophryne, Hadromophryne and Natalobatrachus, four reptile genera -Acontophiops, Inyoka, Macrelaps and Montaspis and the bird genus Lioptilus) endemic to the GMPA region, in contrast to the only three vertebrate genera (Anhydrophryne, Natalobatrachus and Macrelaps) endemic to the MPA hotspot.Furthermore, two golden mole genera, namely Chrysospalax and Neamblysomus, are near endemic to the GMPA region (each genus having two species, one of each endemic to the GMPA region and the others extending to the Northern Mesic Highveld and Central Bushveld bioregions, respectively 29 ).Appendix 1 lists all the vertebrate species endemic to the GMPA region, their status within the MPA hotspot and the EVDs they are characteristic to.

Zoogeographical units and areas of vertebrate endemism
Using the core regions of EVDs, their overlapping margins and the patterns of narrower endemism within them, we demarcated 24 non-overlapping zoogeographical units within the GMPA region (Figure 2c), including 13 AOVEs (Table 2).
Overall, 37 zoogeographical units (Figure 3) can be recognised for south-eastern Africa (south of 22˚S and east All taxa with restricted distributions endemic to south-eastern Africa

Methods revisited
We used the distribution ranges of endemic vertebrate taxa to qualitatively identify a region of conservation priority as well as zoogeographical units for south-eastern Africa, given that the distribution databases for some taxa are far from complete and not particularly accurate.The available distribution data for different vertebrate classes in southeastern Africa vary considerably in their completeness and resolution, as a result of varied intensities of collection and different scales of mapping.Some of the available atlas data sets do not completely cover our study region (most being limited to South Africa, Lesotho and Swaziland), and other references from which these data can be supplemented give distribution maps at different resolutions.Therefore, we used a qualitative approach to delimit EVDs congruent amongst vertebrate classes.From these, a set of zoogeographical units were developed in order to incorporate the available data into a subsequent rigorous numerical analysis, reducing both data incompleteness (as compared to QDSs), and the arbitrary nature of boundaries (compared to half-or fulldegree squares if QDS data were to be pooled).However, data are far scarcer in most of the world's biodiversity hotspots (e.g. the Guinean Forests of West Africa, 54 the Himalayas, 55 Indo-Burma, 56 Western Ghats and Sri Lanka 57 ).Hence we propose EVDs as a tool for understanding biogeographical patterns, refining the boundaries of biodiversity hotspots and identifying local conservation priorities within them, where distribution data are inappropriate for comprehensive numerical analyses.

Vertebrate endemism in the Maputaland-Pondoland-Albany biodiversity hotspot as currently defined
Although the MPA hotspot is identified as a biodiversity hotspot primarily based on its floristic endemism, it is evident that vertebrates also show an increased endemism in and towards the hotspot, albeit with comparatively lower degrees of endemism (5.1%) and near-endemism (4.9%) than is the case for plants, which have a degree of endemism of 23.5%. 2 Vertebrate endemism in the MPA hotspot is comparable with that of the Succulent Karoo (4.2%) but lower than that of the Cape Floristic Region (10.3%), the other southern African biodiversity hotspots.It is certainly low by global standards, with the MPA hotspot ranking 31st out of the 34 biodiversity hotspots currently recognised. 2 The figure for vertebrate  ).Here we took a conservative approach when dealing with marginal occurrences of species, taking their habitat preferences into consideration.

South-east African vertebrate endemism and its congruence with vegetation types
As one would expect, the EVDs described here showed a considerable degree of congruence with vegetation patterns (biomes and bioregions). 29Most of the south-east African endemic vertebrates are associated with either forests or grasslands, and a few of them with thicket (in Albany and azonal fire-free habitat patches elsewhere) and with savanna (especially in northern South Africa).But, in general, most narrow endemic species occupy azonal microhabitats (e.g.rock outcrops and marshy areas) within those biomes and hence are not quite characteristic of the biomes as such.Some bioregions, such as the Dry Highveld Grasslands and Eastern Kalahari Bushveld, are poor in south-east African endemics.
Several narrow endemics of forest affinities fall within the south-eastern coastal belt, represented in patches of sand, scarp and coastal forest. 29In addition, mistbelt forests 29 found along the sub-escarpment belt together with Afrotemperate 29 (Afromontane) forest patches along the Drakensberg-KwaZulu-Natal Escarpment, are also rich in endemic vertebrates.The Northern Middleveld, Southern Maputaland, KwaZulu-Natal Midlands, Drakensberg-KwaZulu-Natal Escarpment, Albany Coastal Belt and the Amatola-Winterberg stand out as the AOVEs with the highest numbers of characteristic narrow endemics within the GMPA region (that is, more than five; see Table 2).Interestingly, the narrow endemics within these richest AOVEs show major habitat affinities to forests, grasslands and thicket vegetation, emphasising the importance of these habitats for vertebrate endemism in the region.
The structural features of the flora determine the faunal assemblages in a region, 5 resulting in the narrow AOVEs coinciding well with vegetation units such as Ngoye, Pondoland and Albany Coastal Belt, even though broader EVDs tend to cut across different vegetation types.It is also evident that the patterns derived from vertebrates are largely congruent with van Wyk and Smith's 5 centres of floristic endemism, the most notable difference being in the delimitation of the Albany region, which is a single centre of plant endemism but encompasses two distinct EVDs.Whilst the prominent EVD of KwaZulu-Natal has not hitherto been formally recognised as a centre of plant endemism, 5 there are numerous narrow endemic plant species in the area 58 (e.g.dune aloe, Aloe thraskii; woolly calpurnia, Calpurnia woodii; sticky star-apple, Diospyros glandulifera; parrot tree erica, Erica psittacina; Tugela spikethorn, Gymnosporia macrocarpa; Wood's spikethorn, G. woodii; Rudatis' dwarf currant, Searsia rudatisii and Tugela bush-milkwood, Vitellariopsis dispar), albeit some have only recently been described.

Congruence of endemic vertebrate distributions and the greater Maputaland-Pondoland-Albany region with published zoogeographical regions
The GMPA region, as well as the narrow-range EVDs presented here, shows congruent patterns with published zoogeographical regionalisation schemes developed with numerical analyses for single vertebrate classes (Table 4).
The faunal zones identified by Crowe 26 for southern African vertebrates also captured similar patterns, with frogs, lizards, snakes and birds having increased species richness towards the GMPA region, whilst the herpetofauna also showed an increased endemism towards the GMPA region.

TABLE 4:
Congruence of the Greater Maputaland-Pondoland-Albany (GMPA) region and some endemic vertebrate distributions (EVDs) illustrated here with published zoographical regionalisations for amphibians and birds.

The greater Maputaland-Pondoland-Albany region of vertebrate endemism and its conservation significance
The importance of the zoogeographically delimited GMPA region as a more comprehensive area for conservation prioritisation needs to be understood by comparing it with the MPA hotspot as currently defined.The coastal section of the GMPA region is largely congruent -if narrower -with the MPA hotspot.Inland, even though the hotspot encompasses most of the Amatola-Winterberg-Sneeuberg EVD and the southern parts of the Eastern Escarpment EVD, these extend well beyond the current boundary of the hotspot.Furthermore, some taxa occupying these EVDs continue their ranges along the Drakensberg Escarpment, which lies almost completely outside the hotspot.This incongruence explains the inconsistency between the current boundary of the MPA hotspot and the patterns of vertebrate endemism (Figure 2a).Only 82.3% of the MPA hotspot's endemism is captured within the coastal section of the GMPA region (Table 5), the remainder of the endemic species have ranges that extend towards the south-eastern Great Escarpment (mostly towards the Eastern Escarpment and a few towards the Amatola-Winterberg-Sneeuberg and Drakensberg EVDs: see Appendix 1).Hence, we propose that a GMPA region that merges coastal and escarpment EVDs represents a more natural region of vertebrate endemism.
The GMPA region can be further validated by vertebrate species near endemic to the MPA hotspot.Amongst the near endemics of the MPA hotspot, 53.3% were found extending from the coastal section towards the south-eastern Great Escarpment and hence are endemic to the GMPA region.Moreover, when the GMPA region is considered to include the Inhambane and Mopane and the Knysna transitional extensions, it encompasses 86.7% of the MPA hotspot's near endemics (Table 5 and Appendix 1).The eight MPA near endemics that are not endemic to the GMPA region nevertheless have their core populations restricted to the GMPA region, with disjunct satellite populations (namely.Barratt's warbler, Bradypterus barratti and Gurney's sugarbird, Promerops gurneyi with satellite populations in the Chimanimani-Nyanga Mountains of eastern Zimbabwe; the Natal mountain catfish, Amphilius natalensis in the Chimanimani-Nyanga Mountains and southern Malawi; Fornasini's worm snake, Afrotyphlops fornasinii, the golden

Conclusions
This study emphasises south-eastern Africa as an important region for endemic vertebrates, especially with respect to herpetofauna.It also provides evidence for congruent endemic distributions amongst different vertebrate taxa, deriving patterns largely congruent with recognised centres of floristic endemism.The importance of the MPA hotspot is assessed for vertebrate endemism, providing less convincing evidence when compared to its flora.But the fact that vertebrate endemism in south-eastern Africa is concentrated towards the coastal belt and adjacent sections of the Great Escarpment provides an option to expand the boundaries of the MPA hotspot to include relevant Afromontane AOVEs.Supported by the zoogeographical links between the coastal and escarpment EVDs, the GMPA region of vertebrate endemism is proposed to serve as a better and more natural region of conservation significance.Simultaneously, the use of EVDs is proposed as a qualitative approach to identify conservation priorities, especially in situations where distributional data do not facilitate a numerical biogeographical analysis.
As suggested by White 45 and van Wyk and Smith 5 , biogeographical patterns detected by means of intuitive perception should be followed by rigorous analysis to establish the extent of biogeographical regions and centres of endemism.Hence, the zoogeographical units presented in this paper are meant to be used in rigorous numerical analyses (e.g.Born et al. 63 -for the Greater Cape Floristic Region).Even though it is far from complete, the understanding of invertebrate diversity and distribution in southern Africa is better than in many parts of the world.
Hence, the distributional data on invertebrates also need to be incorporated into such a rigorous analysis to visualise the broader picture of faunal endemism -invertebrates forming an incomparably higher share of the fauna than vertebrates.Finally, linking the AOVEs to the phylogenetic relatedness of the congruent, range-restricted taxa occupying them will provide clues to the origin and evolution of the faunal endemism in south-eastern Africa, revealing its historical biogeography.Simultaneously, the incorporation of these patterns of endemism in systematic conservation planning is envisaged, prioritising the AOVEs with high congruence across taxa.
the numbers of endemic species (black bars) and disjunct infraspecific taxa (grey bars), fitting each endemic vertebrate distribution (F, freshwater fish; A, amphibians; R, reptiles; B, birds; M, mammals).

TABLE 1 :
2ertebrate species richness and endemism in the Maputaland-Pondoland-Albany (MPA) biodiversity hotspot, as currently defined, and according to data from two additional sources, Mittermeier et al.2and Conservation International 9 .
Amblysomus hottentotushave distributions almost identical to the extent of the GMPA region.The Maputaland, KwaZulu-Natal, Transkei Coastal Belt and Albany Coastal Belt EVDs form the coastal section of the GMPA region, largely matching the borders of the MPA hotspot, whilst the Eastern Escarpment, Drakensberg and Amatola-Winterberg-Sneeuberg EVDs are situated inland, forming the Great Escarpment section of the GMPA region.

Table 2
Middleveld and KwaZulu-Natal Midlands transitional units are also AOVEs).The characteristic narrow endemics of each AOVE are listed in Table3.
and 1 AOVE represents the Waterberg), whilst 7 are transitional units because taxa occupying them show considerable links with units on either side of them: the Northern Mopane, Southern Mopane, Inhambane, Northern Middleveld, Southern Middleveld, KwaZulu-Natal Midlands and Knysna (the Northern Middleveld, Southern

TABLE 2 :
2he in the MPA hotspot is largely stable in comparison with prior assessments (Table1), although many details have changed since, because of the discovery of new species (e.g.cacos, Cacosternum spp., flat geckos, Afroedura spp.and dwarf chameleons, Bradypodion spp.) and taxonomic revisions.The number of MPA-endemic fish species in Mittermeier et al.2seems to be erroneously higher than the actual values.This, together with the increase in species richness as a result of the availability of better distribution data, especially in birds, Number of endemic species + endemic disjunct infraspecific taxa for each area of vertebrate endemism is given in parentheses.FIGURE 3: Proposed zoogeographical units for south-eastern Africa delimited based on endemic vertebrate distributions.endemism
Note: Please see the full reference list of the article, Perera SJ, Ratnayake-Perera D, Procheş Ş. Vertebrate distributions indicate a greater Maputaland-Pondoland-

TABLE 5 :
62presentation of the Maputaland-Pondoland-Albany (MPA) hotspot endemic and near endemic vertebrates within the Greater Maputaland-Pondoland-Albany (GMPA) region of vertebrate endemism.Percentage of the Maputaland-Pondoland-Albany hotspot endemics, endemic within the coastal section of the Greater Maputaland-Pondoland-Albany region.b,Percentage of the Maputaland-Pondoland-Albany hotspot near endemics, endemic within the Greater Maputaland-Pondoland-Albany region.blindleglessskink,Typhlosaurusaurantiacusand the speckled quill-snouted snake, Xenocalamus transvaalensis in the Lowveld region; Phongolo suckermouth, Chiloglanis emarginatus in the Bushveld region; and the laminate vlei rat, Otomys laminatus in the Western Cape).Scelotes mirus; giant Swazi flat gecko, Afroedura major and four undescribed flat gecko species, A. sp.nov.'mariepi', A. sp.nov.'rupestris', A. sp.nov.'rondavelica' and A. sp.nov.'granitica',together with four freshwater fish species: the shortfin barb, Barbus brevipinnis; Treur River barb, B. treurensis; Incomati suckermouth, Chiloglanis bifurcus and Incomati chiselmouth, Varicorhinus nelspruitensis).These species are indeed restricted to a Middleveld strip along the slopes of the Eastern Escarpment, which we identify as a transition zone between the Bushveld and Lowveld bioregions.This transition zone descends to an elevation of about 1000 m a.s.l. and extends south through western Swaziland to north-western KwaZulu-Natal (west of the Lebombo Mountains).Whilst the aforesaid species occurring in the northern parts of the transition zone are near endemic to the MPA hotspot, species occupying its southern parts, like the undescribed flat geckos, Afroedura sp.nov.'lebomboensis' and Afroedura sp.nov.'pongolae', are endemic to the hotspot, providing further evidence for the inconsistency of the hotspot boundary.In contrast, the GMPA region, with a more natural boundary, encompasses all of them as endemics.Along the south-western corner, the boundary of the MPA hotspot follows the Albany centre of plant endemism.In the demarcation of the Albany centre, van Wyk and Smith 5 incorporated a number of different floristic elements, creating a mosaic of bioregions within it.In contrast, the Albany Coastal Belt EVD is much narrower than the Albany centre of plant endemism.Here, the GMPA region encompasses two well-delimited EVDs, the Albany Coastal Belt and the Amatola-Winterberg-Sneeuberg, with coastal and escarpment affinities, respectively.The existence of the GMPA region is also supported by plants, as shown by the distribution patterns of some species in the thicket genera: Encephalartos (Zamiaceae), Rhoicissus (Vitaceae) and Cussonia (Araliaceae),59whilst van Wyk's 60 notation on outliers of some Maputaland endemic plants supports the Inhambane and Mopane extension.Redefining Africa's biodiversity hotspots, Küper at al.61also proposed a region similar to the GMPA region (except for the Inhambane and Mopane extension) as a hotspot of higher plant endemism.The conservation importance of the area covered under the GMPA region is also supported by Olson and Dinerstein62, who prioritised Maputaland-Pondoland Dry Forests as an example for the Tropical, Subtropical, Dry and Monsoon Broadleaf Forest Ecoregion, and the South African Montane Grasslands and Shrublands as an example for the Tropical Montane Grassland and Savanna Ecoregion.Whilst both the above ecoregions are represented within the GMPA region, the MPA hotspot alone only encompasses the Maputaland-Pondoland Dry Forests.
a ,