A biogeographical appraisal of the threatened South East Africa Montane Archipelago ecoregion

Recent biological surveys of ancient inselbergs in southern Malawi and northern Mozambique have led to the discovery and description of many species new to science, and overlapping centres of endemism across multiple taxa. Combining these endemic taxa with data on geology and climate, we propose the ‘South East Africa Montane Archipelago’ (SEAMA) as a distinct ecoregion of global biological importance. The ecoregion encompasses 30 granitic inselbergs reaching > 1000 m above sea level, hosting the largest (Mt Mabu) and smallest (Mt Lico) mid-elevation rainforests in southern Africa, as well as biologically unique montane grasslands. Endemic taxa include 127 plants, 45 vertebrates (amphibians, reptiles, birds, mammals) and 45 invertebrate species (butterflies, freshwater crabs), and two endemic genera of plants and reptiles. Existing dated phylogenies of endemic animal lineages suggests this endemism arose from divergence events coinciding with repeated isolation of these mountains from the pan-African forests, together with the mountains’ great age and relative climatic stability. Since 2000, the SEAMA has lost 18% of its primary humid forest cover (up to 43% in some sites)—one of the highest deforestation rates in Africa. Urgently rectifying this situation, while addressing the resource needs of local communities, is a global priority for biodiversity conservation.


Definition, extent and context
We recognise at least 30 sites in the core of the SEAMA ecoregion (nine in Malawi, 21 in Mozambique; Fig. 1 and Table 1), each reaching an elevation of at least 800 m (elevation range is 500-3002 m) above sea level, and with high humidity (aridity index > 0.65).These sites host remnants of humid evergreen forest and upland grasslands, have ancient substrates, share similar climatic influences, and exhibit high levels of shared and single-site endemism compared with the ecosystems that directly surround them.More precisely, we define the SEAMA ecoregion as "a range of ancient granitic inselbergs in southern Malawi and northern Mozambique, climatically isolated by topography and trade winds, hosting humid evergreen forest, montane grassland and shrublands notable for their high levels of endemism across multiple taxa." Most mountains within the SEAMA were formed ca.600-126 million years ago (Mya) as a result of igneous, rock-forming batholiths intruding into older, softer metamorphic rocks that subsequently weathered away 62,63 .The intrusions are mostly granitic and syenitic in composition 63,64 .Their mineralogy is characterised by low (typically < 10%) ferromagnesian mineral and high quartz and feldspar contents, rendering them relatively resistant to erosion.One aspect that contributes to this is the low degree of jointing in the rocks: SEAMA intrusions are characterised by homogeneous, randomly orientated crystals, which render them much less prone to jointing, water ingress and weathering, compared with, for example, some nearby gneisses (typically of planar fabric resulting from deformation and strain at an earlier time).
The distinct nature of the SEAMA ecoregion is further clarified through contemporary patterns of rainfall and humidity, which clearly show the ecoregion falling under its own climatic envelope with a break to the north and connections to the coast (Fig. 2).This suggests that a defining factor is the south-east trade winds, funnelled up through the Mozambique Channel, carrying moisture to the mountains throughout the year.Mountains further north outside of the SEAMA (including Mts Njesi, Yao, and Mecula) are in the rain shadow of Madagascar, and historically have been more influenced by fluctuations in the water level of Lake Malawi.Further south (e.g., Mt Gorongosa), the air is cooler and so holds less moisture, especially in the dry season.
The SEAMA has distinctly higher annual rainfall and humidity, especially in the dry season, compared to surrounding regions (Fig. 2).This is similar to the Eastern Arc Mountains (EAM) and also the Eastern Highlands of Zimbabwe (EHZ).For plants, this equates to a shorter period of physiological stress, and hence a potentially longer growing season and the existence of mesic vegetation types such as humid evergreen forest.Moreover, since montane forests also extract moisture directly from the air via local orographic precipitation, the estimates  www.nature.com/scientificreports/ of macroclimate presented here (Fig. 2) likely underestimate how much moisture the SEAMA forests receive relative to the lowlands 69 .
The convex hull of the ecoregion is defined by the farthest extents of the core sites identified through the selection criteria outlined by the methodology, plus other inselbergs at the periphery where elevations exceed 800 m and the aridity index exceeds 0.65 (humid climate).The interior of the convex hull also spans arid lowland habitats which, during cooler climes, would have periodically connected the now isolated inselberg habitats, as well as smaller inselbergs that no longer support humid evergreen forest, either due to lower elevation (< 800 m) or because the forests on these sites have already been cleared by humans (some of the core sites have also been largely deforested, e.g., Chiradzulu, Thyolo).
Accordingly, the SEAMA ecoregion has a core area (sum of the areas of the 30 individual sites) of 336,200 ha with a total extent of occurrence (convex hull) close to 10 million ha (Table 1).In relation to the WWF Global 200 ecoregions 9,70,71 , the SEAMA has a smaller core area than some of its neighbours, such as the Eastern Arc Montane Forests EAM (2.4 million ha), the Southern Rift Montane Forest-grassland Mosaic SRMFM (3.3 million ha), the East African Montane Forest ecoregion EAMF (6.5 million ha), and the Ethiopian Montane Forests EMF (24.9 million ha).However, it is a similar size to other mountain ecoregions such as the East African Montane Moorlands EAMM (330,000 ha) and the Knysna-Amatole Montane Forests KAMF (310,000 ha).The SEAMA incorporates the South Malawi Montane Forest-grassland Mosaic, which represents the Malawi component on this larger SEAMA ecoregion, and the Mulanje-Namuli-Ribáuè sub-Centre of Plant Endemism 55 .It also forms part of the Africa-wide Afromontane Archipelago botanical Centre of Endemism 2 .In relation to Key Biodiversity Areas (KBAs) in the Eastern Afromontane Biodiversity Hotspot 72 , the SEAMA is located between the Northern Lake Nyassa Catchments (Southern Rift montane forest-grassland mosaic) and the Chimanimani-Nyanga Catchments (Eastern Zimbabwe montane forest-grassland mosaic).The SEAMA ecoregion encompasses nine Important Bird Areas (IBAs), eight Important Plant Areas (IPAs) from Mozambique 73 (yet to be determined for Malawi), and 12 Key Biodiversity Areas (KBAs) in Malawi and Mozambique 74 (Table 1).

Comparative ecoregion endemism
Compared to five neighbouring mountain ecoregions, levels of endemism in the SEAMA are higher than most in the taxonomic groups surveyed.Notably the number of strictly endemic reptile species is higher than the much larger Albertine Rift (AlbRft) ecoregion (Table 2).When area is taken into consideration using a speciesarea function (see "Methods"), the SEAMA is a close third in ecoregion endemism to the EAMs and the AlbRft across most taxa, and second in comparative endemism for reptiles and crabs (Fig. 3).

Relative survey effort per site
Historically, relative survey effort (defined as sampling intensity relative to mountain blocks and site area 75 , see "Methods") has been much greater in Malawi than in Mozambique (Fig. 4).Overall, the mean relative survey effort for the whole of the SEAMA is 0.22, while the Malawi sites have a mean relative survey effort of 0.41, compared to 0.12 for Mozambique.The most comprehensively surveyed site in Malawi is Mount Mulanje, followed by Mangochi, Ndirande, Soche, and Thyolo.In Mozambique, the most comprehensively surveyed sites are Mabu, Namuli, and Lico, and nine Mozambique sites have never been scientifically surveyed.

Biological uniqueness
The SEAMA hosts endemic genera and species assemblages (Fig. 5a−k), which characterise the ecoregion (Table SI1).Currently, 217 endemic taxa (192 strictly endemic species, plus 25 subspecies and races) are recognised, with a greater number of endemic plants than animals (127 vs 90 taxa, Table 3).The number of endemic taxa is notably high for such a limited geographic area.We expect the number of known endemics, especially amongst the fauna, to grow substantially with additional research, given the low sampling effort for some groups (e.g., although reptiles have been relatively well surveyed, small mammals have not) 76 .Overall, of the groups surveyed, levels of endemism are proportionally highest among reptiles, amphibians, mammals, crabs, www.nature.com/scientificreports/and butterflies.Most endemic animals are forest specialists, as opposed to endemic plants which are typically restricted to high-elevation grasslands and lithophyte communities.Taxonomic influences are evident from all surrounding ecoregions, however there is greater influence from the north (Tanzania) amongst the faunal groups, which suggests a historical continuous humid forest belt stretching from eastern to southern Africa.

Plants
A subset of the SEAMA sites has recently been recognised as forming a distinct botanical sub-centre of endemism 55 , part of the Africa-wide Afromontane archipelago-like Centre of Endemism, as defined by White 2 .Among the best studied sites in the SEAMA are Mts Mulanje 16,77,78 , Zomba 79,80 , Namuli 30,34,81 , Mabu 29,54,82 , Chiperone 31,34 , and Ribáuè 55,83 .Plants account for 117 strictly endemic species and a total of 127 endemic taxa (Table 3, Fig. 5i−k, and Table S1), with the main single areas being Mts Mulanje (48 endemic taxa) and Namuli (19 endemic taxa) (Table S1).Of those assessed for the IUCN Red List, 37 (29%) are listed as threatened (VU, EN, CR).Approximately 69% of endemic plant taxa occur on rocky crags and/or grassland, and only 28% are limited to forest or forest margins.Only three endemic plant taxa show any link to woodland.It is likely that endemic plant taxa remain to be scientifically discovered in the SEAMA from sites currently not surveyed, and from the forests and forest margins of Mt Mabu.
. Comparative ecoregion endemism according to the species-area function.Within each taxonomic group and ecoregion, we plot the number of endemics divided by A z , where A is the core area (ha) of the ecoregion and z = 0.25 is taken as a representative value of the slope of the species-area curve.This yields a metric of endemism to estimate the number of endemic species per hectare, which diminishes the impact of larger areas.

Mammals
Mammals account for four strictly endemic species (Table 3 and Table S1).All are small (< 250 g in mass), and all are forest-dependent.Of those assessed, two (50%) are listed as threatened (VU, EN, CR) (IUCN 2022), while the remainder await formal description.Given narrow ranges and high rates of forest loss, the remaining two mammals will likely be considered Threatened by the IUCN once they are formally described.Because of the relatively low sampling effort for small mammals, the expectation is that many species remain scientifically undiscovered, and so diversity and richness in this group is likely to be significantly higher than is currently known 84 .

Birds
Birds account for eight endemic taxa, three strictly endemic species and five endemic subspecies (Table 3 and Table S1).All endemic taxa are forest-dependent, and most are considered highly threatened.Four (50%) are listed as Near Threatened (NT) or threatened (VU, EN, CR) on the IUCN Red List.Sampling effort has been good on some mountains (Mts Mulanje, Zomba, Mangochi, Liwonde, Namuli, Mabu, and Chiperone) but most of the other mountains are yet to be surveyed by ornithologists.However, the chance of finding new endemic bird taxa is low, since the remaining forest fragments across unsurveyed sites are generally too small to expect new ornithological scientific discoveries.

Reptiles
There are 22 strictly endemic reptile species (Table S1).Of these, 19 (86%) are forest-dependent species, and the others occur mainly in upland grasslands and on rock faces.As they predominantly occur in forest, SEAMA's endemic reptiles are highly threatened by habitat loss.Of those assessed for the IUCN Red List, 12 (55%) are listed as Near Threatened (NT) or threatened (VU, EN, CR), while at least six species await formal description, after which they will almost certainly qualify for an IUCN Threat category.Sampling effort has been adequate for certain sites (Mts Mulanje, Mabu, Namuli, Chiperone, Ribáuè), but many others remain unsampled.Given that new species have been scientifically discovered on each of the surveyed mountains, it is likely that other mountains harbour undescribed endemics, many of which will probably be considered threatened by ongoing forest loss.

Amphibians
There are 11 strictly endemic amphibian species (Table 3 and Table S1).Approximately half of these occur in forest and of those assessed for the IUCN Red List, seven (64%) are listed as Near Threatened (NT) or threatened (VU, EN, CR).Sampling effort has been adequate for some mountains (Mts Mulanje, Mabu, Namuli, Ribáuè) although some fossorial species are probably under-sampled.There are many other mountains that remain poorly sampled or unsampled.New species have been scientifically discovered on some of the recently surveyed mountains 32,46,85 , suggesting that further efforts on other mountains will likely reveal new species or new populations of endemic amphibians, and increased representation on the IUCN Red List.

Freshwater crabs
There are six strictly endemic freshwater crab species to the SEAMA ecoregion, two of which are undescribed species (Table 3 and Table S1).Currently, most of the recognised endemic diversity occurs in the newly established Maritonautes group (Fig. 5h).Four of the endemic species are found inside forest (67%), although none

Butterflies
There are 39 endemic butterfly taxa in the SEAMA ecoregion including 30 strictly endemic species (Table 3, Fig. 5e, and Table S1).The majority (75%) occur in forest and given the degree of forest loss due to slash and burn agricultural expansion [28][29][30][31] ; these species are under threat.Of those assessed for the IUCN Red List, seven (18%) are listed as threatened (VU, EN, CR).Sampling effort has been adequate in certain sites (Mts Mulanje, Mangochi, Zomba, Mabu, Namuli, and Lico), with single collecting visits made to some other sites (Mts Chiperone, Socone, Pewé, Nállume), leaving many others that remain scientifically unexplored.Given that new species have been scientifically discovered on most mountains, there is a high likelihood of finding more species of butterflies new to science.

Origins, connections, and divergence
Many of the plant endemics, including several genera, have their closest relatives in the Southern Afromontane region (stretching from southern Tanzania to the Western Cape province of South Africa), as opposed to elsewhere in East Africa.Although there are some links between the endemic plants of Mt Mabu and the Southern Highlands of Tanzania (e.g., Helixanthera schizocalyx), there appear to be more links with areas to the south, such as the Nyanga-Chimanimani area, while a few appear to be more strongly linked to the East African coastal belt.Others have more general affinities with southern African forms (especially those of drier formations) or even the wider Afrotropics.Within vertebrates, the closest relatives of the endemic mammals are mainly East African (e.g., with Rhinolophus mabuensis (Fig. 5a) close to R. hildebrandtii).Amongst birds, the affinities are mainly with Tanzania to the north and Zimbabwe to the west.For reptiles, the closest relatives are to the north or west depending on the genus, and so taxonomic influences are from all surrounding regions.The closest relatives of endemic amphibians are mainly from east Africa, with just two species associated with southern Africa.
Amongst invertebrates, the closest relatives of the endemic butterfly fauna are East African (88%), with just four species with affinities to central and southern Africa.Therefore, the influences are predominantly northern.The closest relatives of endemic crabs are mainly East African, with 40% connected to Zimbabwe.Therefore, the influences are from the north and the west.Divergence time estimation for endemic freshwater crab species suggest cladogenic activity was initiated during the late Miocene and continued to the Pliocene 50,86 .
Species-level divergences for some of the genera in the SEAMA have been estimated, with divergence estimates ranging from relatively recently (e.g., Plio-Peistocene) to fairly ancient (e.g., Mid-Miocene), Table SI2.
As of yet, no common patterns have emerged, in part due to information on the timing of divergence being lacking for most genera.Nevertheless, for some of the small mammal species, especially the bats (Rhinolophous mabuensis, Fig. 5a), species-level divergences are relatively recent, e.g., 1-2 Mya.The divergence dates between butterfly species have yet to be examined, although a new species of Cymothoe diverged from sister taxa approximately 4 Mya 49 .Within the freshwater crabs (Potamonautes and Maritonautes) species-level divergence dates are estimated at 2.5-8 Mya 50 .In contrast, estimated divergence dates for reptiles and amphibians are more ancient, with the earliest diverging lineages dating to the Mid-Miocene.For example the bush viper Atheris mabuensis (Fig. 5d), the sole representative of the genus from the SEAMA, diverged from its East Africa sister taxa around 15 Mya 33 .The pygmy chameleons (Rhampholeon spp.) are better represented in the SEAMA, with eight of the 25 described species (plus probably several undescribed) occurring on isolated mountains.Six of these form a monophyletic clade, having diverged from the East African species more than 10 Mya 37,87 .The two remaining Rhampholeon species are sister to East/Central African species and likely diverged even earlier.Similarly, specieslevel diversification within the SEAMA endemic amphibian genus Nothophryne (Fig. 5g) are ancient, between 7.5 and 18 Mya 85 .Unlike the chameleons, Nothophryne shares a common ancestor with southern African taxa, not East African 86 suggesting that the SEAMA has linkages to both East and southern Africa.Although additional studies are needed to assess common patterns, the reptile and amphibians essentially exhibit some of the oldest species known from the SEAMA (e.g., Nothophryne broadleyi, 18.65 Mya; Atheris mabuensis, 15 Mya).
Amongst the chameleons, Rhampholeon and Nadzikambia (Fig. 5f,g), most species are endemic to a single mountain, apart from R. tilburyi, which has been recorded from at least five mountains (and is therefore a SEAMA regional endemic).Mts Namuli, Inago, Mabu, Chiperone, Mulanje and the Malawi Hills all have an Overall across taxa, dating estimates suggest that the initiation of allopatric speciation through vicariance of forest patches began in the mid-Miocene for some groups but with other groups diversifying more recently.Thus, vicariance events do not appear to be coeval suggesting the SEAMA taxa have a complex evolutionary history that has melded to form the rich diversity of the region.For example, species-level divergences within the amphibian genus Nothophryne from Mts Mulanje, Namuli and Ribáuè can be dated to approximately 18 Myr, 13 Myr and 7 Myr, respectively 85 and the one SEAMA endemic clade of Rhampholeon shows a similar pattern.Thus, vicariance of the forests on these mountains may have been sequential, resulting in isolated populations that diverged in allopatry.Forest specialists or non-vagile species may have been prone to early vicariance events, becoming isolated in forest patches.The overall patterns to date suggest that Mt Mulanje became isolated first, followed by the eastern SEAMA mountains such as Mts Namuli, Mabu, Chiperone and Mt Ribáuè.Clearly, the high species richness that defines the SEAMA was formed as an amalgamation of both ancient and also recently diverged lineages (crabs, butterflies, and bats), suggesting the processes that contribute to speciation and diversification have been ongoing over tens of millions of years.
Dated phylogenetic analyses show that the earliest SEAMA lineages date back to the Mid-or Early-Miocene.Therefore, we can assume that this ecoregion's forest, started to become isolated from at least the Mid-Miocene.Given that species-level diversification is evident throughout the Mid-to-Late Miocene and into the Plio/Pleistocene, vicariance of these forests would have continued with subsequent isolation of species.These dates are similar to published data on the breakup of the continuous forest belt that covered this part of Africa at this time 2 .Throughout the Miocene, despite periods of wet and dry cycles, the climate became drier overall, resulting in a reduction of forest and an increase in woodland and open grassland.Forests in eastern and southern Africa were consequently confined to areas with higher precipitation, and especially to mountains 1,11,78,88 , resulting in a fragmented forest landscape with upland grasslands in which endemic species emerged as a result of the repeated isolation.

A threatened ecoregion
The major cause of montane forest loss in the SEAMA is slash and burn shifting agricultural practices, typically used for subsistence food production by local communities, along with charcoal production, for household cooking and as a source of revenue (sold on for use in urban areas) 16 .The fertility of the forest soil is valued, which results in smallholder plots inside or on the margins of forest being cleared for crops (especially maize, cassava, and Irish potatoes).Fire from such agriculture affects the forest edges around these cleared plots.Most of these forest patches are naturally small (< 1000 hectares) and edge effects therefore have a disproportionate impact on their ecological integrity.Where forest edges are proportionally high and degraded, the intact forest interior can be compromised due to drying effects.The upland grasslands are also threatened by increases in fire frequency, sometimes associated with flushing animals into traps.Conservation agriculture practices and sustainable alternative livelihoods are required to address these threats throughout the SEAMA ecoregion.
Official protection mechanisms vary between Malawi and Mozambique, as well as among the individual mountains.In Malawi, all the SEAMA sites are gazetted as national Forest Reserves under the management of the Department of Forestry, although this does not seem to afford adequate protection for either the forests or their natural resources, as regulated and unregulated deforestation for timber extraction and charcoal production is rife 16 .Thus, the wet forest on Thyolo Mountain (the type locality for several endemic taxa, including birds such as Chamaetylas choloensis (Fig. 5b) and the endemic race belcheri of Cryptolybia olivacea) was eliminated approximately 20 years ago, with the exception of a small (0.27 ha) forest patch on private land.The same fate affected Ndirande Mountain (in the 1990s) and Soche Mountain (2010s), as also most of Chiradzulu Mountain and the Malawi Hills 89 .In Mozambique, only one of the SEAMA sites, Mt Ribáuè (which includes Serra Mpàluwé) is a gazetted Forest Reserve.However, most forest reserves were established to regulate the harvesting of timber, not for establishing protected areas in terms of conservation 90 , and therefore Mt Ribáuè lacks any tangible, formal protection.Although all other sites lack any formal protection through national legislation, a project is currently underway to declare Mt Mabu as a 'community-conservation' protected area.Some SEAMA sites in Mozambique, such as Mt Lico, are protected due to their natural inaccessibility, while some others, such as Mt Mabu and Mt Pewé have remained in relatively good condition, possibly because these forests have spiritual value to the local communities.Overall, however, the majority of SEAMA sites lack meaningful national protection in contrast to conservation initiatives implemented in neighbouring ecoregions, such as the Eastern Arc Mountains in Tanzania.
At the level of taxonomic groups, extinction risks vary according to their reliance on the forest or upland grassland habitats (Table 2).Most of the small mammals, birds, reptiles, and amphibians, are forest-dependent Vol:.( 1234567890 www.nature.com/scientificreports/and therefore the proportion of threatened species is generally high due to the acute forest destruction on most mountains.Therefore, many of these forest-dependant species are listed as Near Threatened (NT) or threatened (VU, EN, CR) on the IUCN Red List, for example the Atheris mabuensis (Fig. 5d), Rhinolophus mabuensis (Fig. 5a), Apalis lynesi, Rhampholeon chapmanorum, and Paraxerus vincenti-see Supplementary Material for a complete list.For the endemic plants (e.g., Fig. 5i−k) the majority (69%) are found in open habitat which are threatened to varying degrees by increased fire frequency 30,91 .However, forest plant species such as Helixanthera schizocalyx, the two new Polysphaeria species from Mts Ribáuè and Mabu 42 , and the forest tree Faurea racemosa, are all listed as Endangered on the IUCN Red List.The greatest threat for the majority of sites is deforestation and increased fire frequency to the upland grasslands.Within the area defined (convex hull) for the SEAMA ecoregion, approximately 18% of primary humid forest above 800 m in elevation was lost between the years 2000 and 2022 (Table 1).This, for the same period, is far greater than in other African mountain ecoregions that contain primary humid montane forest 91 , including the Eastern Arc, Cameroonian Highlands, and Mt Cameroon and Bioko montane forests (< 5%), and the Albertine Rift and Guinean montane forests (ca.10%).Therefore, the SEAMA is considerably more threatened by the rate of primary humid forest loss.Within the SEAMA, the extent and timing of forest loss varies between sites, with some such as Chiradzulu, Ndirande, and Thyolo Mountains in Malawi having lost all their forest cover prior to the year 2000, while the Malawi Hills are estimated 89 to have lost nearly 80% of forest cover since the 1980s.Several other sites have suffered severe declines in the period 2000 to 2022 (Table 1), including Mts Nállume 43%, Inago 39%, Ribáuè 35%, Namuli 30%, Socone 18% and Chiperone 14%.The sites with the least loss of primary humid forest over the last 22 years are Mts Meluli and Mabu (both 3%), and Mt Lico (< 2%), this latter figure is because the Mt Lico forest is essentially inaccessible..Note that these estimates only account for loss of primary humid forest, not secondary forest, or woodland, such that the actual habitat loss including the lower mountain slopes is likely to be much higher [91][92][93] .More generally, rates of montane forest loss in Mozambique, at close to 30% since 2000, are among the highest in tropical Africa 91 .

Conclusions
An ecoregion has shared biotic and abiotic characteristics that distinguish it from surrounding areas 9,70 .In the case of the SEAMA, a distinct climatic envelope clearly characterises a specific range of mountains, and defines its boundaries.Within these boundaries, we find unique species assemblages, characterised by an abundance of endemic lineages.Although survey effort has improved in recent decades for the SEAMA, compared to other neighbouring regions (e.g., Eastern Arc Mountains) its Relative Sampling Effort is still comparatively low (Table 1).Despite this, the SEAMA has more strictly endemic reptile species than the much larger and betterknown Albertine Rift ecoregion.Where biological sampling has been relatively high within the SEAMA, e.g., Mt Mulanje, significant levels of endemism have been found, with a total of 48 endemic plant species (modified from Strugnell 15 ) and a range of endemic fauna 13,14,16 .
The South East Africa Montane Archipelago (SEAMA) 11,21,78,88 might be Africa's newest and most threatened ecoregion with one of the highest deforestation rates (18%); however, the evidence base now exists for its global recognition as a priority site for conservation.

Definition, extent and context
For the purposes of data collation and mapping, we bounded the ecoregion in two ways.First as a collection of core sites, each individually named and spatially delimited.These are the sites of greatest known biological interest.Second, we defined a convex hull around the core sites, plus other inselbergs with the requisite elevation (> 800 m) and humidity (aridity index > 0.65).Thus, the convex hull encompasses the core sites, other inselbergs (with or without forest remaining), as well as lowland connections in between.We defined 'endemic taxa' as those found only within the convex hull of the ecoregion and associated with one or more of the 30 core sites (as per Fig. 1 and Table 1).Our definition of endemic taxa includes strictly endemic species, subspecies, and races.Our analysis is based on the taxonomic groups we have listed.It is recognised that levels of endemism will further increase when additional taxonomic groups are sampled.
We bounded the core sites as close to the base as possible, selecting the lowest 100-m elevational contour that distinguishes only the target features.Due to variation in base elevations across the SEAMA extent, different contours were appropriate for different sites (Table 1).The chosen contours were converted to polygons, buffered by 500 m and then smoothed to aggregate features and simplify the topology.We used digital elevation data from the Shuttle Radar Topography Mission (SRTM) v3 at 1 arc-second (~ 30 m) spatial resolution, gap-filled using a bilinear resampling of the 3 arc-second version of the same product (which has no gaps).
For the aridity index, we used the 30 arc-second (~ 1 km) surface provided by CGIAR-CSI 66 .The aridity index is defined as the ratio of mean annual precipitation to potential evapotranspiration, where values < 0.2 are indicative of an arid or hyper-arid environment, 0.2-0.5 semi-arid, 0.5-0.65 dry sub-humid, and > 0.65 humid 67 .Evapotranspiration was based on the FA0-56 Penman-Monteith Reference Evapotranspiration equation, using rainfall and temperature estimates from WorldClim version 2 65 .In Fig. 2, we mapped this aridity index, and three other climatic variables that we expect to correlate with ecoclimatic stability 94 : annual rainfall (BIO12 in WorldClim2), rainfall seasonality (BIO15) defined as the coefficient of variation in rainfall across months, and maximum water deficit, a measure of dry season water stress 67 .We computed the maximum water deficit across consecutive months that experience rainfall < monthly ET0, over which the shortfall in rain was accumulated 68

Comparative ecoregion endemism
To place SEAMA (336,200 ha) in context, we compared the total number of strictly endemic species from various taxonomic groups with five other mountain ecoregions in east and southern Africa that are found within the Tropical & Subtropical Moist Broadleaf Forests category of the Afrotropical ecoregions according to the WWF Terrestrial Ecoregions Global 200 assessment.These are the Eastern Arc Mountains forests (EAM) at 2,380,000 ha; the Albertine Rift (AlbRft) at 10,390,000 ha; the East African Montane Forests (EAMF) at 6,563,700 ha; the Ethiopian Montane Forests (EMF) at 24,930,200 ha; and the Knysna-Amatole Montane Forests (KAMF) at 310,800 ha.In this context, in all taxonomic groups surveyed, 'endemic species' refer to strict endemic species only, while 'endemic taxa' refer to strict endemic species, endemic subspecies and endemic races.
Although direct comparisons between the absolute number of strictly endemic species can be made (Table 2), there are orders of magnitude differences in area between these ecoregions.To account for these differences in area, we applied a species-area function to estimate the number of endemic species per hectare 95 by dividing richness (strict endemic species) by A z , where A is the core area (ha) of the ecoregion and z is the slope of the species-area curve (= 0.25), which diminishes the impact of larger areas.

Relative survey effort per site
We estimated relative survey effort per site using a composite index that takes into account the area of each SEAMA site (A), and sampling intensity per site (SI) 75 .Survey effort reflects the amount of time and resources allocated to assess the biophysical aspects of each site within SEAMA.As such it allows comparison of sampling effort between ecoregions which adds context to the known levels of endemism and the likelihood of increasing these levels.We scored sampling intensity using a 0-1 scale as follows: 0 = no known studies in that site; 0.2 = limited and non-systematic survey over few localities in the site; 0.5 = systematic survey over many localities in the site; and 1 = extremely thorough and systematic survey covering all habitats in the site.
We adapted the original formulae and calculated the mean sampling intensity (SI) for all the taxonomic groups studied at each site.In our case I indicates the number of mountain blocks per site thus including all habitat types to account for the endemic species found in the moist forests, upland grasslands, and rocky subtracts.

Primary forest loss since 2000
We estimated forest loss using a Global Forest Change dataset provided by the Global Land Analysis and Discovery (GLAD), University of Maryland 93 .These data have reasonable accuracy in large, dense evergreen forests, but accuracy decreases for seasonal forests, at low canopy density, and for earlier years (2001-2010).Loss rates are relative to estimated tree cover extent for the year 2000, including plantations.For these reasons, we restricted the estimates of forest loss to an estimate of primary humid tropical forest extent 92 .Nonetheless, we note that there may be inaccuracies owing to the global nature of these forest segmentation/classification algorithms, and that losses in secondary/disturbed are potentially higher.

Figure 1 .
Figure 1.Location and extent of the South East Africa Montane Archipelago (SEAMA) showing core sites in red, and an outline boundary of the convex hull of the ecoregion (created using QGIS version 3.28.12LTR https:// qgis.org/ en/ site/).

Figure 2 .
Figure 2. Contemporary macro-climatology of the South East Africa Montane Archipelago.Also labelled is the Eastern Arc Mountains (EAM) ecoregion to the north, and the Eastern Highlands of Zimbabwe to the southwest (EHZ).Annual rainfall (a) is measured in mm.Rainfall seasonality (b) is the coefficient of variation across months65 .Aridity index (c) is the ratio of annual rainfall to potential evapotranspiration66 with values above 0.65 considered humid67 .Maximum water deficit (d) is a measure of water stress defined across the most arid months of the year, with lower values conducive of potentially dense, evergreen canopy structure68 .Bioclimatic layers were extracted from WorldClim version2 (https:// www.world clim.com/ versi on2) and presented in QGIS version 3.28.12LTR (https:// qgis.org/ en/ site/).

Figure 4 .
Figure 4. Relative survey effort across core sites in the SEAMA.Bars plot a comparative measure of sampling intensity, relative to the area of each site75 .Score varies between 0 and 1, where 0 is the minimum (no sampling at all) while 1 is the maximum effort possible to obtain.

Table 1 .
Details of the SEAMA core sites.Base contours were used for site delineation, the results of which are shown in Fig.1.Under recognition/reserve status: Important Bird Area (IBA); Key Biodiversity Area (KBA); Important Plant Area (IPA); Forest Reserve (FR).Dashes indicate no data for primary forest loss, and no data for SEAMA endemics.

Table 2 .
Number of strictly endemic species by taxonomic group in mountain ecoregions in east and southern Africa.DD Data Deficient, *Expert judgement (ABRI).*Expert judgment was engaged to assess the endemic butterflies for three of the neighbouring ecoregions based on published and unpublished data by the Executive Director of the African Butterfly Research Institute (ABRI).Due to the very comprehensive butterfly database at ABRI, this was considered acceptable and reasonably accurate in this instance.ABRI houses the largest global collection of African butterflies.

Table 3 .
50mber of endemic taxa known from the South East Africa Montane Archipelago and their conservation status.The last column also indicates the percentage of species considered Threatened (VU, EN, CR) on the IUCN Red List (accessed September 2022).Rhampholeon, with an endemic Nadzikambia recorded for each of Mabu (Fig.5g), Chiperone, and Mulanje.Chameleons on these inselbergs probably shared a common ancestor in the Mid-Miocene37.Phylogenetic analysis has shown that R. bruessoworum is the oldest diverging lineage, dated to the Mid or Early-Miocene (ca.11-30Mya).This may have been followed by the isolation of R. platyceps on Mt Mulanje and R. tilburyi on Mts Namuli, Pico Mulli, Nallume, Ribáuè, and Pewe in the Mid to Late Miocene.Divergence times for Rhampholeon from Mts Mabu (R. maspictus, Fig.5f), Chiperone (R. nebulauctor), and the Malawi Hills (R. chapmanorum) are more recent (4-9 Mya) within the Late Miocene or Early Pliocene, suggesting these forests remained connected until most recently37.The two most closely related species are R. nebulauctor from Mt Chiperone and R. chapmanorum from Malawi Hills, which are just 75 km apart and may have been connected through the Late Pliocene37.A recent phylogenetic study50showed that initial diversification with the SEAMA endemic crabs (Potamonautes and Maritonautes) dates to just over 7 Mya, with species-level diversification mainly within the Pliocene, e.g., Maritonautes namuliensis, 5.4 Mya (Fig.5h); Potamonautes mulanjeensis, 2.84 Mya; Maritonautes licoensis/Maritonautes choloensis, 2.5 Mya.