The phytogeography of moist forests across Eastern Zimbabwe

Background and aims – During the 1970s, a comprehensive survey of moist forest areas across Zimbabwe’s Eastern Highlands was carried out, from Nyanga in the north to Chirinda Forest in the south. All tree stems 8 cm diameter or greater in both canopy and sub-canopy layers were measured in 176 quarter-hectare plots, and plots were then classified into 12 altitude-related forest types. The aim of the present study is to categorise the woody species recorded in these plots in terms of their chorological status. The findings are compared to similar analyses from moist forests across Malawi. Material and methods – All tree species recorded during the initial survey were listed and placed into 12 described chorological (phytogeographical) categories based on their continental distribution. Their occurrence across the 12 previously described forest types is given. Key results and conclusions – A total of 211 tree species was recorded from 176 plots. Most species (86, or 40.8%) are Afromontane endemics or near-endemics, while 48 species (22.7%) are Guineo-Congolian linking, 31 (14.7%) are Eastern endemic or near-endemics, and 16 (7.6%) are sub-Afromontane endemics or near-endemics. The remainder comprise African linking and Zambezian species. This shows that species of the moist forests of eastern Zimbabwe form a complex phytogeographical mix. While Afromontane species dominate at high altitudes, Guineo-Congolian and Eastern species become more common at lower altitudes. In terms of proportions of chorological categories, results were found to be similar to those from forest studies in Malawi.


INTRODUCTION
Globally there has been much interest in tropical moist forests (often termed rain forests), including on the importance of their relatively high species diversity, the numbers of species of restricted distribution found there, and in moist forests being a habitat of conservation significance (e.g. Sayer et al. 1992;Burgess et al. 2004). In Africa, moist tropical forests comprise a wide range of species, with differing geographical distributions. Some are mostly confined to lowland areas while others are restricted to higher altitudes. As White (1978White ( , 1983a has noted, many woody forest species have relatively wide distributions, whether across the Congo Basin into West Africa, along the chain of mountains on the eastern side of the continent from Ethiopia to South Africa, or confined to lower-altitude areas closer to the Indian Ocean coastline. Especially in eastern and southern Africa, where many of the moist forest areas are scattered and associated with mountains or higher ground, it is not clear how unique each of these various areas are in terms of their species composition -whether each forest patch comprises a fortuitous association of species of differing geographical origin, or whether there is a more consistent or repeating set of species associations and ecologically coherent communities. Another important question is whether these associations change with latitude, altitude, or with substrate (Müller 1999). In addition, it would be interesting to understand what the distributional and evolutionary origins of moist forest species might be, and if they are similar to each other or quite varied across the region.
In the 1970s, one of the present authors, Tom Müller, undertook a wide-ranging study of what were considered to be the remaining areas of tropical moist forest across the Eastern Highlands of Zimbabwe (Müller 1999(Müller , 2006. The main objectives of this study were to try and determine the range and distribution of forest types, their species composition, and to see which readily measurable environmental factors (e.g. altitude, aspect, slope, soil type) might determine this. The study systematically recorded all tree species with stems ≥ 8 cm diameter within 176 quarterhectare plots. These plots spanned the full geographical and altitudinal range of forest in the country, with plots recorded from all larger forest patches. The 8 cm diameter limit was chosen to conform to locally accepted forestry practice at that time, and because the use of a lower diameter limit would have been far more time consuming. A summary of the distribution and extent of the 12 forest types identified has been produced (Müller 1999(Müller , 2006 but, apart from data on limited forest loss over the last 50 years (Timberlake & Müller 2021), no other quantitative findings have yet been published.
Given the availability of this dataset, the opportunity was taken to list all recorded tree species from the plots and allocate each to one of White's phytochoria or biogeographical units (see below). This was done to provide a reference source for future studies on the phytogeography of south-central African forests, and also to help determine what the broad affiliations for Zimbabwe's forests might be. A later paper will look at the relative proportions of each species in terms of its basal area across the 12 forest types; an initial attempt at doing this using a subset of data was presented at an AETFAT Congress in 2007 (Timberlake & Müller 2007). To date, the only other detailed study on the phytogeography of south-central Africa's moist forests is that by Dowsett-Lemaire (1988a, 1989, 1990 and White et al. (2001) for the forests of Malawi. Results of the Zimbabwe phytogeographical analysis are discussed and compared to this study.

STUDY AREA
Moist forests are localised in Zimbabwe, being found principally on the windward slopes of a 250 km stretch of mountains along the Zimbabwe-Mozambique borderlands of Manicaland Province from 17°44′S to 20°32′S ( fig. 1). These east-and south-east-facing slopes intercept moisture- bearing air flows from the Mozambique Channel, air flows that provide additional moisture during the long dry season. The forests range in altitude from 320 m to 2100 m and have a very disjunct distribution. They range greatly in size, many being patches of a few hectares, but can be blocks of up to a few square kilometres (Müller 2006); the largest patch at 2850 ha (28.5 km 2 ) is found on the slopes on Mt Nyangani.
There are four main centres of forest development ( fig.  1), which are (from north to south) the Nyanga mountains (Clark et al. 2017), the area around Stapleford Forest and the Bvumba mountains  including Banti, the mountains and uplands of Chimanimani District (Timberlake et al. 2016;Wursten et al. 2017), and finally the southern outlier of Chirinda Forest (Timberlake & Shaw 1994). Total forest extent is estimated at approximately 107 km 2 (Müller 2006), less than 0.03% of the country's total area.
Average annual rainfall across forest areas is at least 1200 mm, most falling in the warmer rainy season but with significant precipitation in the form of rain and mist from low clouds during the cooler dry season; this greatly reduces the physiological stress of the long dry season. The main environmental variables thought to control species distribution and the species composition of the 12 forest types (see table 1) are altitude (often a proxy for mean temperatures), amount of available moisture (especially during the dry season), past human disturbance and, to a lesser extent, soil type, aspect, and topography (Müller 1999).
Although there has been little previous work on their distribution and composition, the importance of these moist forests in terms of biodiversity has been widely recognised in Zimbabwe (e.g. Wild 1968 Keay (1959) and earlier workers, who based phytogeographical units on the distribution of genera and families, White (1976White ( , 1983a proposed a new chorological classification of African vegetation based on the patterns of distribution shown by species. He thus defined 18 main phytochoria, which, according to levels of endemism and species richness, are divided into three categories: Regional Centres of Endemism, Regional Transition Zones, and Regional Mosaics. A Regional Centre of Endemism has more than 50% of its vascular plant species confined to it and a total of more than 1000 endemic species (White 1983a). There are nine such regions in Africa, of which two concern Zimbabwe − the Zambezian region and the archipelago-like Afromontane region. Zimbabwe lies entirely within the Zambezian region: it is one of the richest regions in terms of species diversity, but the main vegetation types are different forms of woodland and thicket, hence the region is unlikely to be more than very poorly represented in the country's moist forests.
Within the Zambezian region sit enclaves of the Afromontane archipelago-like region, as shown in White's map (White 1983a). In contrast to the widespread but poorly represented Zambezian species, Afromontane endemics are expected to be important elements in the upland forests of eastern Zimbabwe. A small subset of Afromontane affinity included here, the sub-Afromontane element, applies to species associated with montane vegetation but occurring overall at a lower elevation than Afromontane endemics.
Regional Transition Zones and Regional Mosaics have fewer than 50% endemism and are transitional in various ways (White 1983a). Among Regional Mosaics, White (1983a) recognised two such regions of Eastern (coastal) elements, from north to south the Zanzibar-Inhambane and  Tongaland-Pondoland regions. During the preparation of the Evergreen Forest Flora of Malawi in the 1980s (published as White et al. 2001), this concept had to change to include a wider region of Eastern endemism, containing both of those coastal mosaics with an extension inland, from now on termed the Eastern region. Geographically, it comes west to the fringes of the Lake Victoria basin and the eastern half of the Zambezian region (Dowsett-Lemaire 1988a: 83;White et al. 2001: 16). The main vegetation types of the Eastern region are evergreen moist forest and related types and there is thus virtually no overlap with Zambezian or other drier vegetation. As with the Afromontane region, the Eastern (forest) region has a discontinuous, archipelago-like distribution with islands of vegetation scattered over a large area dominated by other regions; except for the coastal belt these islands are too small to be mapped. The proportion of endemic elements is less than 50% overall and decreases from east to west. In addition to species endemic to one of these regions, the forests of Zimbabwe also consist of species of wider distribution belonging to two or more phytochoria, termed 'linking species'. Species of wide distribution found in three or more phytochoria are termed African linking. Some of these linking elements are also ecological transgressors, occurring in more than one vegetation type. The most important category of linking species found in moist forest usually consists of Guineo-Congolian linking elements, i.e. species that are characteristic of Guineo-Congolian forest vegetation (including secondary forest) and which also occur in other forest types on the continent. In contrast, African linking species refer to taxa not represented in the Guineo-Congolian region, or only on its margins. The distinction between a Guineo-Congolian linking and an African linking taxon is sometimes difficult to establish: examples include Albizia glaberrima and Diospyros abyssinica subsp. abyssinica which are found in the drier forest types at the periphery of the Guineo-Congolian region, but also locally in semi-evergreen moist forest further inside this region. For the moment they are still classified as African linking but that decision could be reversed. The status of a few species classified by White et al. (2001) in one or other category has had to change in light of improved knowledge of their distribution and ecology (e.g. in Hawthorne & Jongkind 2006; William Hawthorne, University of Oxford, in litt.): thus Antidesma membranaceum, Bridelia micrantha, Margaritaria discoidea, and Psydrax parviflora are here changed from African linking to Guineo-Congolian linking as these pioneer species occur in secondary forest right in the heart of the Guineo-Congolian region, including the evergreen forest zone. Conversely, Ficus vallis-choudae and Vitex doniana have been moved from Guineo-Congolian linking to African linking as they are completely absent from Guineo-Congolian vegetation in West Africa and the Vitex is marginal at best on the southern fringes, as in southern Congo-Brazzaville. White (1978) subdivided the Afromontane archipelago into seven subregional groups, based on inter-island distances. These are 'West African' (mostly Cameroon/ Nigeria), 'Ethiopian', 'Imatongs-Usambara', 'Kivu-Ruwenzori', 'Uluguru-Mulanje', 'Chimanimani' (= eastern Zimbabwe and adjacent Mozambique), and 'Drakensberg'. The highlands of Angola were considered too impoverished to be included, but are nevertheless recognised by ornithologists, given a significant level of local endemism of the Angolan scarp (Dowsett 1986). White (1978White ( , 1983a has repeatedly stressed that, despite the wide intervals between mountains, the endemic montane flora shows a remarkable uniformity. He has also shown (White 1981(White , 1983b) that some Afromontane near-endemic trees − presumably more tolerant of climatic variations than stenotopic species − have satellite populations in mid-altitude forests in south-central Africa (on the Congo-Zambezi watershed) and in some upland areas of Congo-Brazzaville, Gabon, and Cameroon, indicating a possible linking route between the mountain floras of eastern and western Africa in relatively recent times.
The different chorological categories used in this analysis are listed below (see table 2). Not all species defined as belonging to a particular region are strict endemics (that is, strictly confined to a region): species belonging overwhelmingly to one region but also feebly represented elsewhere − either in neighbouring transition zones or in more distant satellite populations − are referred to as 'near-endemics' in the meaning of White (1979). Some Afromontane 'endemics' may occur marginally beyond the African continent, as in south-west Arabia or Madagascar, but they are still classified as endemics because we are concerned with a comparison of lowland and montane elements in an African context. If this is not done the terminology becomes too complicated and the situation in Africa becomes blurred.

METHODS
Over 200 georeferenced plots, each 0.25 ha (most being 50 × 50 m) and marked on aerial photos, were recorded during Müller's study in the 1970s. Of these, 176 plots have data on all tree stems larger than 8 cm dbh, by species, and were used in this paper. Although lianas were recorded, diameter measurements were not made and hence these species are not incorporated here. Altitude, slope, soil type, and other ecological aspects such as disturbance, were also recorded (Müller 2006). Additional information recorded (although not used in this paper) included species composition and cover-abundance in both sapling and ground layers. Where there was uncertainty as to species identity, a voucher specimen was collected. These were later identified (most by Tom Müller and Bob Drummond) at the National Herbarium in Harare (SRGH), although only a small number were ever formally accessioned. Species authorities are given in full in supplementary file 1.
Data from all 176 quarter-hectare plots were put into Excel spreadsheets from the original summary fieldsheets, with the total number of measured stems noted for each plot by species. Basal area was calculated from diameter or circumference (field data came in both formats at various times) using standard formulae. Any stems less than 8 cm dbh (or the equivalent circumference of 25 cm) were omitted. Plots were then grouped under each of Müller's forest types (Müller 1999 and unpublished notes, table 1) and total basal area was calculated for each type. Comparable forest plot data were not available for Malawi. As no plots were recorded for the distinctive Widdringtonia nodiflora forest (type 3), this type was omitted from some of our results.
For each forest type, the component species were allocated to a chorological unit based primarily on those given in White et al. (2001), with White (1983a) and Friis (1992) being used where a species was not found in Malawi. Some small changes concerning Guineo-Congolian and African linking species are detailed above under definitions of chorological categories. Any species still lacking a chorological unit were assessed by us using mapped distribution data available on the African Plants Database (2020). The 12 categories recognised (plus uncertain) are listed and grouped in table 2.
Species nomenclature and family arrangement follow those used in the Flora of Zimbabwe website (https://www. zimbabweflora.co.zw/, Hyde et al. 2020) with some minor modifications, and many have been significantly updated from the names used during the original survey. Where appropriate, the species name under which it was originally recorded during the surveys is also given.
These summary data were then amalgamated into a single table (supplementary file 1) listing all forest species alphabetically by genus and species, indicating the lifeform (e.g. canopy tree, understory tree), chorological unit, total number of plots in which the species occurred, and the presence of a species within each forest type. In order to indicate which species were more significant, those that contributed 1% or more to the total basal area recorded for each type are indicated (X); the remainder (indicated by O) being relatively minor in this regard.

RESULTS
The number of plots recorded for each of Müller's forest types, the number of species in each and the mean basal area are shown in table 1. From this it can be seen that the coverage in terms of number of plots for each type was not even, and that total species number per type was also not uniform. Apart from the obvious lack of plots in type 3, sampling density ranged from around one plot per 8.8 ha for type 8 to one plot per 93 ha and per 149 ha for types 5 and 6, respectively. Medium altitude forest (type 11) was the most species-rich, although also with the largest number of plots recorded, while the highest altitude forest types (types 1, 2), regenerating forest (type 9) and Craibia brevicaudata forest (type 8) were the least species-rich. Basal area did not differ greatly across types with a mean of 43.7 m 2 /ha, although lowland forest (type 12) was surprisingly low in this regard at 35.4 m 2 /ha.
Although 35 species were present in only one plot, three species (Aphloia theiformis, Macaranga mellifera, and Xymalos monospora) were present in 100 or more plots (supplementary file 1). Skewness of species distribution is also shown by only 72 (34%) species being found in more than 10% (18) of total recorded plots.
Not surprisingly, the main chorological category of Zimbabwe's moist forest tree species comprises Afromontane (near-)endemic species with 86 species, and when 16 sub-Afromontane elements are added to the total, these constitute nearly half (48.4%) of the floristic composition (table 2, fig.  2). The Guineo-Congolian linking category with 48 species comes next (including two Guineo-Congolian/Afromontane linking species) and, together with Afromontane and sub-Afromontane species, these comprise 71.1% of the total. There are also 31 Eastern (near-)endemic species.

Afromontane (near-)endemic species
Of the 86 Afromontane species, a few are endemic only to the Chimanimani montane group: Faurea rubrifolia, Maytenus chasei, Pavetta umtalensis, and Canthium oligocarpum subsp. angustifolium (the species as a whole being present in five montane subregions  Afromontane near-endemics as they occur not only on the Southern Track but also very locally in the tropical lowlands, but they are infinitely more common in, and characteristic of, montane vegetation. Most of these widespread trees produce fruit dispersed by birds, and it may be significant that a key disperser, the Afromontane near-endemic Olive (Rameron) Pigeon Columba arquatrix (Dowsett-Lemaire 1988b), undergoes migrations outside the breeding season and is also present in very small numbers along the Southern Track (as in Mwinilunga District in north-western Zambia, Dowsett et al. 2008). There is an almost perfect match between the ranges of Olive Pigeon and Olea capensis, one of its most important food plants (Dowsett-Lemaire & Dowsett 1998, 2006. Finally, some of the near-endemics include species that extend into the Cape region, e.g. Cassinopsis ilicifolia, Cussonia spicata, Podocarpus latifolius, Rapanea melanophloeos, and Widdringtonia nodiflora.

Sub-Afromontane (near-)endemic species
Most of these 16 species occur at medium altitudes. Newtonia buchananii is one of the more characteristic trees of this forest type, being locally abundant in forests of central and southern Malawi between 1000 and 1600 m; in Zimbabwe, with increasing latitude this is slightly lower (below 1400 m, cf. forest type 11). Several of the near-endemic species occur locally on the Congo-Zambezi watershed. Among them Cyathea manniana, Newtonia buchananii, Suregada procera, and Trichilia dregeana all reach the highlands of Angola; Newtonia and Trichilia also reach Cameroon and even (Trichilia) the uplands of Guinea and Liberia.
Calodendrum capense is predominantly of Afromontane affinity in the northern part of its range (Uganda and Kenya) but less so from Malawi southwards. By contrast, Bersama swynnertonii and Strychnos mellodora are endemic to the Chimanimani subregion, whereas Ochna arborea and Vangueria esculenta extend from Zimbabwe to South Africa. Thus the 'origins' or distribution patterns of plants of this category are as varied as for Afromontane species.

Eastern (near-)endemic species
The 31 species listed are characteristic of evergreen forest and thicket at low and medium altitudes on the eastern side of the continent, several reaching as far inland as eastern Congo-Kinshasa and one or two have satellite populations in Nigeria or Angola. Some show Afromontane intrusions (i.e. they occasionally reach higher mountains or high altitudes), such as Ficus scassellatii and Gymnosporia mossambicensis.
Only one species, Synsepalum chimanimani, is endemic to eastern Zimbabwe and adjacent Mozambique. Although Afromontane endemic/near-endemic

Zambezian (near-)endemic species
Tree species endemic (or nearly so) to the Zambezian region are normally encountered in woodland and dry thicket, vegetation types drier than evergreen or rain forest. The 12 listed here are rather marginal, with the exception of Mimusops zeyheri. Uapaca lissopyrena is considered conspecific with U. guineensis by White et al. (2001) and the latter is a Guineo-Congolian linking species, normally encountered in swamp or fringing forest.

Guineo-Congolian linking species
Not surprisingly, the majority of the more widespread species falling outside a regional centre of endemism are to be found in this category, the Guineo-Congolian region being originally dominated by moist forest. These 48 species are found mainly in lowland or mid-altitude forest types, but may ascend into the lower levels of Afromontane forest, even to the highest level in the case of Clausena anisata.

Afromontane/Zambezian linking species
There is only one Afromontane/Zambezian linking species, Heteromorpha arborescens, which is common in and characteristic of both the Zambezian and Afromontane regions.

African linking species
These 17 species occur in three or more phytochoria, and most have a circum-Congo distribution. Breonadia salicina is a good example, being characteristic of riparian vegetation from Mali east to Ethiopia, south to KwaZulu-Natal and west to Angola; it also extends to Yemen and Madagascar. Several of these circum-Congo species could marginally be considered Guineo-Congolian linking as they occur in the drier types of Guineo-Congolian forests or at its fringes (e.g. Diospyros abyssinica subsp. abyssinica). Others are typical only of very dry forest in the Guineo-Congolian region and extend into riparian forest or moist woodland in savanna to the north, e.g. Ekebergia capensis, which then ascends to high altitudes on the eastern side of Africa from Ethiopia south to South Africa. This and several other species are ecological transgressors, occurring in more than one vegetation type over a wide range, e.g. Gymnospora buxifolia (Ethiopia to the Cape, west to Namibia and Angola). Mystroxylon aethiopicum is a versatile ecological transgressor, being characteristic of upland vegetation from Guinea and Cameroon to Ethiopia and south to South Africa, but is also found in riparian forest, woodland and bushland on termite mounds at low altitudes. Moreover, it extends to several Indian Ocean islands, including Madagascar.

DISCUSSION
It is recognised that plot coverage was not particularly high with a total sampled area of 44 ha or just 0.4% of the total forested area. However, at the time the plots were carefully chosen to represent the full range of forest types and geographical locations. In addition, the number of woody stems measured was 29,100, giving comprehensive coverage. Even though there were no plots recorded for the distinctive Widdringtonia forest (type 3) compared to the high number of plots that were recorded for Medium altitude forest (type 11), it is apparent that Mixed sub-montane forest and Medium altitude forest (types 7 and 11, respectively) were particularly species-rich. As expected, there was a general tendency to greater species richness with decreasing altitude. As plot data were not available for the Malawi study, or for other similar forests in the region, useful comparisons cannot be readily made at this stage.
The proportion of Afromontane species was expected to increase with altitude, and that of 'lowland' species, especially Guineo-Congolian linking species, to decrease. This was clearly shown by Timberlake & Müller (2007) using a subset of basal area data from Müller's Zimbabwe study. They showed that the total basal area comprised of Afromontane elements was around 41 m 2 /ha in Muller's type 1 and type 2 forests, declining to 17 m 2 /ha in Medium altitude forest (type 11), and 0.7 m 2 /ha in Lowland forest (type 12). In contrast, Guineo-Congolian linking elements totalled just 1.3 and 0.1 m 2 /ha in type 1 and type 2 forests, respectively, rising to 10.7 and 11.0 m 2 /ha in type 11 and type 12 forest at lower altitudes.
In Malawi, the proportion of Afromontane species was used to define the three main forest types of the country: thus Afromontane forests are defined as such when they have in excess of 50% Afromontane (near-)endemic species while 'mid-altitude forests' have between 25 and 50% and 'lowland forests' less than 25% (Dowsett-Lemaire 1989, 1990White et al. 2001: 60−64). On Mount Mulanje, the only large mountain with a near-continuum of forest from low to high altitude (Dowsett-Lemaire 1988a), the proportion of Afromontane elements was found to increase from 22% at 650−950 m to 44% in mid-altitude forest (up to 1350 m) and to 76% in Afromontane forest (1600−1950 m on upper slopes and plateaux). Elsewhere in Malawi, the elevation of mid-altitude forests is somewhat higher as conditions are warmer than on the wet southern slopes of Mount Mulanje; they lie mostly between 1100-1200 m and 1500-1600 m. In Zimbabwe, no attempt was made to define mid-altitude forest using these criteria so a direct comparison is difficult, but Medium altitude forest (Müller's type 11) has just 29% Afromontane elements.
There is a general tendency for montane trees to occur at lower altitudes with increasing latitude and a shift from tropical to temperate conditions. For example, in South Africa and especially in the Cape, Podocarpus latifolius, Olea capensis, and other montane near-endemic trees occur only a short distance above sea level. Although not specifically looked at here, this phenomenon has also been noted for some individual species distributions between Zimbabwe and Malawi (Françoise Dowsett-Lemaire pers. obs.), and in addition Zimbabwe appears to have a lower diversity of montane forest species than Malawi or Tanzania (cf. Dowsett-Lemaire 1989: 118).
With the proviso that the Malawi study included a wider range of species (with lianas, slender climbers and smaller shrubs), very similar figures were obtained for a sample of 660 species (excluding 24 elements of uncertain taxonomic and chorological status) over an altitudinal range of (500-)600−2500 m (White et al. 2001: 50 and last column in table 2). Thus, we get a similar figure of 47.6% Afromontane elements (including 4.6% of sub-Afromontane (near-)endemics), followed by 17.8% of Guineo-Congolian linking species and 15.5% of Eastern (near-)endemics. The remainder comprise African linking and Zambezian species.
To conclude, species constituting the evergreen moist forests of eastern Zimbabwe form a very complex phytogeographical mix, ranging from narrow Afromontane endemics to pan-African ecological transgressors. Not surprisingly, the proportions of the different phytochoria represented are very similar to those observed in the Malawi forest flora, and change markedly with altitude. There are few species restricted to these forests (i.e. local endemics), but these include some belonging to the Afromontane region and one Eastern (lowland) species.