Vegetation of the Tembe Elephant Park, Maputaland, South Africa

A hierarchical classification, vegetation map, description and ecological interpretation of the plant communities of the Tembe Elephant Park and surrounding areas are presented . The study area falls within the Maputaland Centre of Endemism, which is part of the Maputaland-Pondoland Region , a cen tre of plant diversity rich in endemic plants and anima ls . Sixty-four sample plots were distributed in a stratified manner throughout the study area , A TWINS PAN classification , refined by Braun-Blanquet procedures, revealed eight distinct, mainly woodland plant communities , Much of the vegetation distribution can be ascribed to vegetation dynamics and the geomorphological evolution of the region . Dynamics of water in the landscape either directly or indirectly, through its role in moisture levels and soil formation, plays a role in the determination of plant communities at the higher levels. The proposed classification and general descriptions can be used for


Introduction
Maputal and, a region of grea t scenic beauty and rich biodiversity. shows considerable ecotourism potential. Several con servation areas have been proclaimed in the region, including the Grea ter St Lucia Wetland Park (a World Heritage site) , Tembe El ephant Park, Maputo El ephant Reserve, Ndumo Game Reserve. Mkuzi Game Reserve an d Kosi Bay Coas ta l Foresl Reserve, Recently, Moza mbique granted a co ncession compri sing approxim ately 230 OOOh a for ecotourism devel opment in south ern Mozambique. Currentl y initiatives for establishing a tran s-fro ntier conservation area ('peace park') between northern KwaZulu-Natal and southern Mozambique are underway, effectively linking Tembe Elephan t Park with some of Ihe Mozambi ca n parks and co nservation areas.
An area com prising southern Mozambique and the northern part of KwaZu lu-NataJ is recognised as a Centre of Plant End em ism, namely the Maputaland Centre [MC] (Van Wyk 1994, 1996, High levels of endemism are spread across virtually the who le taxo nomic spectrum. involving both plants and animals. The Maputa land region is also of exceptional nature reserve management, land-use planning and further research . Ordinations based on floristic data give an indication of floristic and associated habitat and environmen tal relationships. Soil character and moisture differences are usually slight, hence the sandbased communities do not have distinct boundaries, except for Sand Forest. Indications are that in the case of Sand Forest the distinct boundary zone of nearly bare soil may well be the result of allelopathy, Eighty-four Maputaland Centre plant endemics/near-endemics were recorded for the study area , with Sand Forest being the most prolific habitat for such taxa, many which appear to be neoendemics. Because of its hypothesised origin (especially lack of specific historic events) and vegetation dynamics, the inclusion of Sand Forest in conservation areas is no guarantee of the futu re perSistence of this rare vegetation type. biogeographical inleresl because of the sharp biogeographical tran sformati on of both plants and animals in the reg ion (Poynton 1961 , Bruton andCo oper 1980). Mapu tal and is at the southern end of the tropics in Africa and many organisms reach the southernmost limit of th eir range here.
Tembe Elephant Park [TEPJ was proclaimed in 1983, II lies within the core area of the Me and conta ins many rare pl ant and animal sp ecies. Noteworthy is the presence in the park of a rare forest type, locally referred to as Sand Forest and in Mozambique as Ucuati Forest. Sand Forest is very disti nctive , with a unique combin ation of plant and an imal species. As fa r as is known, th is vegetation type is more or less restricted to ancient coastal dt:Jnes in north ern KwaZulu-Natal and the extreme southern portion of Mozambi que (Maputaland). Good examples of Sand Forest ca n be seen in the Greater St Lucia We tlan d Park, Ndumo and Mkuze Game Reserves, and particularly th e TEP, Sand Forest harbours many rare and unusual plant and animal species, including several Me end emics. Because of its restricted occurrence and unusual species complement, Sand Forest     7 (42.5, 19761 tha n 30m beneath the surface dune sands. The topography in the study area is defined by high , linear, narth-sau1h oriented dune cordons with poorly preserved parabolic dunes superimposed on the surface. Sand comprising these dunes represents the Plio-Pleistocene weathering products of the underlying Uloa/Umkwelane Formation rocks with localised aeolian as well as sheetwash a nd reworking of sand off high-lying areas, The colour and textural characteristics of surface geological units in the study area are dominated by post-depositional, pedogenic weathering of the sands. Distribution of the sands, which represents a broad soil chronosequence, is characteristic of the entire coas tal plain. The oldest Plio-Pleistocene weathering profiles are red and dominan t in the western areas or in localised places where old sands rise above the surrounding deposits. Lesser-developed, yellowish and grey profiles formed in younger redis tribu ted sands during the Pl eistocene and Holocene, occur in degraded dune fields or areas dominated by a high water table. Calcrete deposits have formed adjacent to the Muzi Swamp and thick peat deposits have accumulated in the pe rmanent swamps.
Lateral ground water migration towards interdune depressions or the Muzi Swamp base level has resulted in formation of clay-rich , slightly saline or calcareous duplex soils in low-lying sites. Where the dune sands border on the Muzi Swamp, narrow areas of sodic Estcourt Form soils (SoH Classification Work ing Group 1991 ) wilh prismatic subsoil structure are common , These natric soils can be classified broadly as planosols (FAO-UN ESCO 1974).
Permanent swamps in the Muzi system are cha racterised by gleying conditions with pea t formation and development of organic-rich histosols (FAO-UNESCO 1974, Soil Survey Staff 1996 or soils of the Champagn e Form (Soil Classification Working Group 1991).

Hydrology
Water ta ble and ground water movements play an important role in relation to vegetation pattern::; in most parts of Maputaland, including Ihe TEP, as the area is covered by deep sand. From borehole data, water table depths are on average approximately 35m , but deplhs of 60m below surface have been recorded. In some places , water table depths can exceed 90m (borehole of 90m without reaching water). An elevated water table is prese nt in a few places such as the season al pans that occur on the duplex soils. Ground water is almost exclusively replenished by rainwater. After ra ins , quick fluctuations in local water levels can be experienced for a period. From work done by Kruger (1986) in the neighbouring areas to Sileza Nature Reserve, which have very similar soils as TEP, it is estimated that vertical seepage rates are in the order of 0 .1 mfday. Wa ter movements through the sands are at an average transmissivity of 20 m 2 /day and co-efficient of storativity of 1-10" . Permeabilities are highly variable, because the Pleistocene sediments underlying the coversands show extreme eastwest lateral variability (Hattingh 1998). In general, the swamps and marshes are surface expressions of the groundwaler table, with little or no evidence of perched groundwater horizons occurring in the area. The only permanent waler in the TEP (most years) is the Muzi Swamp , in the east of the reserve. The flow direction of the Muzi Swamp in this area is north , into Mozambique. The only othe r sources of su rfa ce water in the reserve , are the seasonal pans that occur on the duplex soil s. Th ese pans are usually dry during the winter period from April 10 September.

Methods
Sixty-four sample plots (10 x 10m; in accordance with the plot size used by Matthews et a/. 1999) were distributed in a stratified random manner throughout the study area . As far as possible, the plots were equally distributed in the different physiographical-physiognomically homogeneous units, distinguished on the basis of physical environmen t, physiognomy, dominant pla nt species composition and abundance.
Sampling was carried out from January to May 1996. Scientific names of tax a follow Arnold and De Wei (1993).

577
(1983). The following information was recorded for each sample plot: 1, total floristic compos ition and cover-abundance val ue for each pla nt species , according to the Braun-Blanquet cove r-abun dance sca le as described by Mueller- Dombois and Ellenberg (1974); 2, wate r table depth at the time of sampling, using a 2m soil auger; 3, soil Iypelform (FAO-UNESCO 1974, Soil Classification Working Group 1991 and basic colour; 4, amount of organic material ranked as distinct or indistinct and its depth of occurrence; 5, topographical position based on terrain Iypes (Land Type Survey Staff 1986), namely (1) crest, (2) scarp, (3) mid slope, (4) foots lope and (5) valley bottom or floodplain; 6, geology, according to 1 :250 000 geological survey maps and relevant li te rature (Geological Survey 1986, South African Commiltee for Stratigraphy 1980, Botha 1997) and locally at a larger sca le, based on personal observations in the field ; 7, land type according to 1 :250 000 land Iype map (Land Type Survey Staff 1986); 8, a 50 point herbaceous species survey using the step-poin t method (Mentis 1981) Largely unpublished soil analysis data were used in the descri ption of the soils. Th ese were based on soil samples taken from what was considered a representative example of the different planl commun ity types as distinguished in this study. Not all sub-communities were sampled due to the cost involved in delailed soil analyses. Ch emical analysis (pH , macro and micro nutrients) of the soH was done at the surface and at 1 m. 2m, 3m and 4m depths. The pooled resu lls of these different level samples were used in the discussions that follow.
Two-way indicator species analysis (TWINSPAN) was applied to the basic fioristic data set to derive a first approximation of the possible plant communities (Hill 1979a).

Refin ement of th is classification was done by the application of
Braun-Blanquet procedures (Behr andBredenkamp 1988, Bredenkamp et al. 1989 Wyk (1996). The term 'endemic' is used to refer to a taxon limited in its range to a restricted geographical area , or a particular substrate. When a taxon is predominantly confined to a particular area or substrate. but is also marginally present elsewhere (sometimes in Ihe form of distant salellite populalions), it is called a 'near-endemic' (Ma tthews et a/. 1993).
Plant communities recognised in the study area are classified as follows : 1. Drypetes arguta-Uvaria lucida subsp. virens Sand Forest of deep sandy areas (dry sands) . A major community which is mostly associated with dunes (crests, slopes but also in some interdune depressions), on the orange-brown arenosols and dystric regosols. Soils are sand with very little clay and low pH values (pH 5.6). Results of a separate soil analysis showed Sand Forest soils to be more acidic than those of the other regic sand communities.
This plant community is characterised by species group C (   (Table 4).
Structurally, this plant community is a forest, in some instances having a low canopy of approximately 5m and in other places extending up to approximately 15m. The species composition is typical of what has been called Licuati or Sand Forest (Myre 1964, Moll and White 1978, Ward 1981, Matlhews et at. 1999, with species present such as Dialium sch/echteri, Pte/eopsis myrtifolia, Ptaeroxy/on ob/iquum, Groton pseudopu/chel/us and Newtonia hildebrandlii. Protruding crowns ot some of the larger Iree species in Sand Forest are covered by many epiphytes, such as Microcoelia exilis and va rious lichens including Usnea spp.
This major community is subdivided into two sub-communities based on structure and floristic composition.

Cola greenway/-Balanites maughamii tall Sand Forest
A sub-community which is mostly associated with dunes (slopes but also in some interdune depressions), on the orange brown arenosols and dystric regosols. Soils are sand with very litlle clay and acidic (pH 5.6).

Psydrax fragranlissima-Hyperacanlhus microphy/-Ius short Sand Forest
A sub-community which is mostly associated with dunes (crests, slopes, but also in some interdune depressions), on the orange brown arenosols and dystric regosols. Soils are sand with very Iitlle clay and acidic (pH 5.6) . Reasons ror the floristic differences between this community and the previous one are not clear, but may well be due to vegetation dynamics .

Spirostachys africana-Berchem/a zeyheri closed woodland on duplex soils (clays)
A major community which is found on the clay-rich duplex soils normally associated with botlomlands of dunes as well as the borders of the Muzi Swamp. Perennial pans are to be found interspersed throughout these clay areas. Many of these pans have developed from mud wallows use by game species, especially elephant. This community is found on the grey-brown planosols.
This plant community is characterised by species group D l.    (Table 4) .
In structure this plant community varies from areas of dense vegetation to lhickets associated with termitaria, which are on clay (duplex soils) . Species composition of these th ickets on termitaria is similar to that foun9 in other areas in South Africa, such as Roodeplaat Dam Nature Reserve , and are sometimes referred to as bush clumps (Van Rooyen 1984).

Strychnos madagascariensis-Combretum molle woodland of deep sandy areas
A community which is found on dune crests, slopes, and interdune depressions, on the grey dystric regosols. This community is the most dominant and widespread community in the study area.

a
Th is plant community is characterised by species group L (Table 3). Diagnostic consistent species include the tall to short shrub Strychnos madagascariensis, and the trees Combretum molle and Acacia burkei. Prominent species (>50% constancy) include the tree Sclerocarya birrea , the shrubs Strychnos spinosa, Dichrostachys cinerea and the creeper Rhoicissus revoilii. On average, 39 species were recorded per sample plot. Thirty-seven MC endemiclnearendemic species were reco rd ed of which thirteen are restricted to this community, making this the second richest community in the TEP for Me endemicfnear·endemic species (Table 4).
This community is the dominant vegetation type of TEP, thus the diagnostic species represent the general species for th is area and as such would not be very consistent. Structurally this plant community is a wood land of varying density, from closed to open . Canopy height varies from approximately 5m, extending up to 10m in places. This major community is what Granger (1996) ca lled Subhumid Lowveld Bushveld, although his species composition described for this unit is not very typical.
This major community is subdivided into two sub-communities based on vegetation density, structure and floristic compos ition.

Acacia burkei-Euclea natalensis closed woodland and thicket
A sub-comm unity which is found on dune crests, slopes and interdune depressions, on the grey dystric regosol. Soils are acidic (pH 5.5) sands with very little clay, although sometimes may have very narrow bands of clay present. This is one of the most widespread sub-communities in the study area.
This SUb-community has no diagnostic species. It is characterised by the absence of species group K and the presence of species group H. Although species group H is also associated with thicket on clay, these species are linked to the presence of narrow clay bands (Table 3). Prominent species (>50% constancy) include the sh ru bs Euclea natalensis, Grewia caffra, Catunaregam spinosa and Bridefia cathcartica (species group H). Another species showing a strong affini ty to this community is the tree Acacia burkei, which has its highest constancy in th is community (species group L). On average 41 species per plot were recorded , Ihis community having the second highest number of species among the studied plant communities. Twenty-eight Me endemic/near-endemic species we re recorded , of which one was restricted to this community (Table 4). This plant community ca n be distinguished based on density, wh ich in most cases varies from continuous to sub-con tinuous crown gaps, with a canopy from 8m to approximately 12m high .
This sub-community is subdivided into two variations, one being a generally dense plant commu n i t y~ and the other an ecotonal form which is less dense and having affinities with the more open woodlands.

Zanthoxylum capense-Clausena anisata closed woodland and thicket on sand
A community which is found on dune crests, slopes and interdune depressions throughout the TEP, on the grey brown dystric regosols , which sometimes may have very narrow clay bands present.
This plant commun ity is cha racterised by species group F (Table 3). Diagnostic consistent species include the tree Zanthoxylum capense and the shrub Clausena anisata. Prominent species (>50% constancy) include the trees Vepris lanceolata and the forb Achyranthes aspera. Other species showing a strong affinity to this community are the shrubs Phyffanthus reticulatus, Grewia caffra and the tree Acacia burkei (species group L). The shrubs are from species groups G and H, showing an association with thicket vegetation, Acacia burkei has il highest constancy in this community. On average 42 species per plot were recorded , making this the second richest ptant community.

Panicum max;mum~Acacia burke; ecotonal closed woodland
A community wh ich is found on dune crests, slopes and interdune depressions throughout TEP, on the grey brown dystric re905015 , which may sometimes have very narrow clay bands present.
This plant communi ty is characterised by the absence of species groups F and G (Table 3) and the presence of Hand L. This community has no diagnostic or prominent species as it is an ecotonal community associated with less dense areas with essentially no clay in the soil. Species showing a strong affinity to this community are the shrub/tree Combretum molle and the grass Panicum maximum. The tree Acacia burke; attains its highest cover-abundance in th is community. Other species showing an affinity to this commu ni ty is the tree Terminafia sericea and the grass Digitaria eriantha, both belonging to species group 0 (species associated with open to sparse vegetation) . All the above species belong to more generally distributed species groups and species more strongly associated with less dense vegetation. On average 35 species were recorded per sample plot.

Salacia kraussii~ Themeda triandra open woodland on sand
A community which is found on dune crests, slopes and interdune depressions throughout TEP, on the grey brown to orange grey dystric regosols. Th is is also one of the more widespread communities of the study area.
This plant community is characterised by the presence of species group K (Table 3). Diagnostic consistent species include the geoxylic-suffrutex Safacia kraussii, the grasses Themeda triandra , Pogonarlhria squarrosa and Eustachys paspa/oides. Not all these species have high constancy. The tree Terminafia sericea attains its highest cover values in this community and is also the most consistent species (100% present) (speci es group 0 , Table 3). Species group 0 is associated with open to sparse vegetation cover. Prominent species (>50% constancy) include the shrubs Vangueria infausta , Brachy/aena disc%r and the grass Cymbopogon validus. Another species that shows a strong affinity to this community is Strychnos madagascariensis, which has its highest cover-abundance values in this commu nity. On average 37 species per sample plot were recorded. Twenly-two Me endemic/near-endemic species were recorded, of which six are restricted to th is community (Table 4) .
This community can be distinguished based on density, which in most cases varies from open to sparse woodland, with few tall trees from approximately 8m to 10m. This community is subdivided into two sub-communities based on vegetation density and floristics.

3,2.1 Andropogon gayanus~Albizia versicolor open woodland on sand
A sub-community which is found on dune crests , slopes, and interdune depressions throughout TEP, on the orange grey dystric regosols. Soils are acidic (pH 5.34) and sandy with very little clay. This is also one of the more widespread subcommunities of the study area .
This plant community is characterised by the presence of species group I (Table 3). Diagnostic consistent species include the grass Andropogan gayanus and the tree A/bizia versicolor. Prominent species (>50% constancy) include the shrubs Sapium integerrimum, Garcinia livingstonei and the forb Hoslundia opposita. Another species that shows a strong affinity to this community is St!},chnos madagascariensis , which has its highest abundance values in th is sub-community, The grass Panicum maximum also attains very high abundances in this sub-community. Other species are the trees Combretum mof/e, Terminalia sericea, the geoxylic-suffrutex Eugenia mossambicensis and the grasses Digifaria eriantha, Diheteropogon amplecfens and Perotis patens. The last two species come from species group N, this group representing grassland areas associated with Sand Forest edges. On average 37 species per sample plot were recorded.
This community has few tall trees from approximately 8m to 10m.

Helichrysum kraussii-Dichapetalum cymosum sparse woodland
A sub-community which is found mainly on the fiat areas between dunes, but also occurring to a lesser degree on dune slopes and crests throughout TEP, on the grey brown dystric regosols. Soils are slightly acidic (pH 5.8) sand with very lillie clay, the lowest percentage clay of all soils examined.
Th is plant community is characterised by the presence of species group J (Table 3). Diagnostic consistent species include the woody forb Helich!},sum kraussii and the geoxylic suffrutex Dichapeta/um cymosum. Promi nent species (>50% constancy) include th e shrub Da/bergia abovata , the grasses Trichoneura grandig/umis, Pan/cum ka/aharense and the geoxylic-suffrutex Parinari capensis subsp. incohata . Another species showing a strong affinity to th is community is Terminalia sericea, which attains its highest abundance in this sub-community. Other species are the shrub Strychnos madagascariensis, the geoxylic-suffrutices Sa/a cia kraussii, Eugenia mossambicensis, the grasses Pogonarthria squarrosa, Themeda iriandra , Diheteropogon amp/ectens, Perotis patens, Aristida stipitata subsp. spicata, Digitaria eriantha, the forb Carchorus junodii and the woody creeper Rhoicissus revoilii. Three of the last-mentioned species come from species group 0 , th is group representing grassland areas associated with Sand Forest edges, a clear link between wood land and these areas . On average 37 species per sample plot were recorded. This community is similar to what is called 'woody grassland' (Matthews et at. 1999) -a type of grassland characterised by an abundance of species with the geoxylic-suffrutex growth form .

Andropogon schirensis-Tephrosia longipes subsp. longipes grassland associated with Sand Forest
A community which ;s found directly neighbouring on the Sand Forest patches, throughout TEP, on the grey brown dystric regosols. Soils are slightly acidic (pH 5.47) sand wi th very little ctay. Th ese soils have been shown to possibly retain some allelopathic properties that are presen t in th e Sand Forest comm unity (unpubli shed results).
This plant community is characterised by the presence of species group M (Table 3) . Diagnostic consis tent species include the grass Andropogon schirensis and the forb Tephrosia longipes subsp. longipes. Prominent species (> 50 % constancy) include the forbs Indigo!era inhambanensis. Trachyandra ct. salti, the grasses Bewsia biflora , Urelytrum agropyroides and the sedge Cyperus obtusiflorus. Another species showing a strong affinity to this community is the grass Perotis patens that is very consistent (100%) and attains its highest cover-abundance value in th is community. Pero/is patens ;s a characteristic pioneer species. Strychnos madagascariensis also has an affinity to this commu nity, but takes the form of a stunted ·shrub. Other species are the grasses Diheteropogon amp/ectens, Aristida slipitata subsp. spicata , and Digitaria eriantha. The first two mentioned species come from species group N, which includes species common to open wood land , showing that th is community has a close association with the open woodland areas. On average 23 species per sample plot were recorded. Ten Me endemic/near-endemic species were recorded, of which two are restricted to this community (Table 4) .
This community is open gras sland with little or no trees and shrubs , and is always found directly bordering on the Sand Forest comm unity. These characteristic grasslands do not exlend very fa r from Sand Forest edges and exh ibit a gradient; further fro m the Sand Forest patches the se gras slands gradually acquire an open woodland character.

Nidorella auriculata-Justicia flava hygrophilous grassland on clay-rich soils
A communi ty which is often found next to pans and marshy areas as well as clay-based thickets, throughout TEP, on the grey brown to brown orange planosols. This community has soils with the highest percentage clay in the study area.
Thi s plant community is characterised by the presence of species group S (Table 3). Diagnostic consistent species include the forbs NidorelJa auriculata , Justicia flava and Kyllinga erecta. Prom in en t species (>50% constancy) include the grasses Panicum maximum , Aristida congesta and the forb Commelina d. erecra. On average 26 species per sample plot were recorded. One MC endemic sp ecies has been recorded in this commun ity, to wh ich it is also restricted ( Table 4).
This commu nity can be associa ted with 'wet areas' as water collects in these clayey areas after rain, resulti ng in temporary pans. Therefore , much moister salls occur here than are foun d on the sands where the water drains away. The Muzi Swamp on the other hand is also clay-based but is fed mostly by outside ground wate r sources.

587
Based on floristics , this co mmunity is subdivided in to two sub-communities.

Waltheria indica·Sporobolus pyramidalis grassland on clay between thicket and pan marsh edges
A sub -community which is often found nex t to or surrounding pans and marshy areas as well as clay-based thickets, throughout TEP, on the grey brown planosols. Th ese soils have a high percentage of clay.
This sub-community is characterised by the presence of species group Q (Table 3). Diagnostic consistent species include th e forbs Waltheria indica and the grass Sporobolus pyramidaiis. Prom inent species (>50% constancy) include the grass Eragrostis cifiaris, the forbs Monechma debile and an Indigo!era sp On average 29 species per sample plot were recorded.

Acacia borleae-Kohautia virgata shrubland on clay soils often associated with calcrete
A sub-community which is found next to marshy areas as well as the clay-based thickets associated with the Mu zi Swamp, on the orange grey brown planosols. These soils are alka line (pH 9) wi th a high percentage of clay.
This sub-community is characterised by the presence of speCies group R (Table 3). Diagnostic co nsistent species incl ude the shrub Acacia borleae and the forb Kohautia virgata . Prominent species (>50% constancy) include the grass Eragrostis heteromera that attains its highest cover/abundance values in this SUb-community. Eragrostis heteromera associated with species group U, which represents the common species of the Muzi Swamp. On average 23 species per sample plot were recorded.
6. Imperata cylindrica~Eragrostis lappula grassland on sandy clay soils associated with Muzi Swamp and other marshlpan areas A community which is found neighbouring the Muzi Swamp as well as marshy areas associ ated with sand, on the dark brown grey humic gleysols to planosols .
This plant commun ity is characterised by the presence of species group T (Table 3). Diagnostic consistent species include the grasses Imperata cylindrica, Eragrostis /appu/a and Dactyloctenium geminalum. Prominent species (>50% constancy) include the shrub Acacia nilo/ica , the grass Panicum genuflexum and the forb Nidorella anoma/a. Other species showing a strong affinity to this communi ty are the grass Eragrostis heteromera (species group U) and the sedge Cyperus nata/ensis. Cyperus nata/ensis comes from species group W, linking this community with the clay-based communities. On average 21 species per sample plot were recorded. Seven MC endemic/near-endemic species were recorded of which six are restricted to this community (Ta ble 4).

Oxalis cf. semiloba-Nymphaea nouchali aquatic veg~ etation of standing water in marshes and pans
A community which is found associated with the many perennial and pe rmanent pans and marshes occurring in the clay-based areas (duplex soils), throughout TEP. This community is found in the water of inundated planosol soils.
This plant comm unity is characterised by the presence of species group V (Table 3)

Phragmites australis reed bed s of the Muzi Swamp
O ne particu lar commu nity tha t was sampled but not i nclud~ ed in Table 3 is the reed beds of the extensive Muzi Swamp, which crosses the eastern side of the reserve and extends northwards to Maputo Bay in Mozambique. This is a welldescri bed commu nity that is widespread and associated with wa ter bodies th roughout South Africa and the world. Here along the entire length of the Muzi Swamp, dense stands of Phragmites australis occur, although in some places these are mixed with Typha capensis. Some of these areas are utilised by the local community for harvesting of reeds. The Muzi Swamp is the only sou rce of permanent wate r in TEP, even in the driest years. T his community is ra ther similar to the Phragmito-Magnocaricetea Klika in Klika and Novak (194 1) descri bed for Europe (Mucina 1997).

Ordination
Distribution of the sample plots (releve s) along the X and Y1 axis as well as X and Y2 axis of ord ination is given in the form of sca tter diagrams (Fig ures 4 and 5). The scatter diagrams illustrate a moisture gradient along the horizontal axis, with the wetter habitats to the right and the drier habitats to the left. Arenosols and regosols communities base themselves on the left of the diagram, while planosols and histosols communities base themselves on the right. On the X and Y 1 axis , a distinct discontinui ty exists between the commun ities of the Sand Forest (A). woodlands (6), Muzi Swamp grassland and the perennial pans (H). Grassland comm un ities associated with clays (G) do not show a discontinui ty, but do form a strong group restricted to a specific area of the scatter diagram.
There is no discontinuity in the distribution of the sample plots representing the woodla nd and associated com munities on the X and Y1 axes. However, on the X and Y2 axis a clea r separation with disconti nuities can be seen with in the woodland communities, namely. closed wood land associa ted wi th duplex soils and grassland associated with Sand All the proposed commu nities are distinguishable in the field based on structure, growth form, general species composition and character species -th is is despite the gradual enviro nmental gradients that cause comm unities to merge into one ano ther, particularly so in the case of the woodlands. The different woodland sub-communi ties often ca n be diffi cult to distinguish based on floristics alone, but structure and density help 10 distinguish between them .
Except for the broad-scale class ification of the coasta l grasslands of southern Mozambique (Myre 1964(Myre , 1971) and the detailed classification of coastal and inla nd commun ities elsewhere in Maputaland (Lubbe 1996, Matthews et a/. 1999), the present study is a further attempt at a more detai led , larger scale phytosociological cla ssification of some of the Maputaland coastal plain communities. A correlation between the plant commu nities of the prese nt study and Ihe various syntaxa described by Myre (1964Myre ( . 1971 and other workers has been attempted. Myre (1964) did not fo rmally describe any of the closed woodland or forest and related communities , but rather concentrated on the grassland a nd related communities. He did, however, give a basic species composition list for what he called Licuati Forest.

Mozambiq ue corresponds closely with what is called Sand
Forest in South Africa. Incl uded in this spesies list are also some species associated with thicket vegetation on sand and clay, e.g. Spirestachys africana and Euphorbia ingens. This is also the case for the Sand Forest type described for Sileza Nature Reserve, namely the Artabotrys monteiroae-Dialium schlechteri forest, the coasta l form of Sand Forest (Matthews et a/. 1999). The last mentioned variation compares well with typical Sand Forest, although it also contains a few species which are associated with what is referred to as Ihicket or closed wood land in the present study (e.g. Commiphora neglecta). Such species are , however, very few in the case of the Artabotrys monteiroae-Oialium schlechteri forest. Lubbe (1996) found a degree of similarity with Sileza 'Sand Fores t' with what he described as Tricalysia delagoens;s-Dalium schlechteri inland forest.

dry < -------------------------
However, the latter inland forest type contains many coastal dune forest species and is obviously more closely associated with coastal dune lorest whilst containing a lew Sand Forest elements. The open to sparse woodlands described in this study do have similarities with the Themedelo-Salacietum M. Myre (1964) woody grasslands of Myre (1964) and Matthews et al. (1999). The typical woodland communities, as described in the present study. are limited in distribution and in variety in Sileza Nature Reserve (Matthews et a/. 1999) .
The Muzi Swamp communities described in this study are also rather similar to the Sileza Nature Reserve's Ischaemum fasciculatum-Eragrost;s inamoena hygrophilous grasslands and specifically to the Monocybium ceresiiforme-Eragrostis lappula grassland of areas not regularly inundated with water (pan edges) . However, in TEP, which has drier conditions with much deeper water table levels, these communities are limited, and species compositions ditfer accordingly.

Major determinants of ptant communities
A simple situation exists in the MC with the important determinants being the interconnected effects of water table (moisture), soil type and topography, upon which lire and vegetation dynamics have been superimposed. Casual observalions and comparison with other studies (notably Myre 1964, Lubbe 1996, Matthews at al. 1999 have nevertheless shown that the plant communities (and Ihus associated environmental factors) 01 the present sludy extend over most 01 the Maputaland Coastal Plain . In the case 01 TEP, which lies in a drier and topographically more elevated part 01 the Maputaland Coastal Plain, water table level is not a leading lactor, although soil moisture plays an important role. In the case 01 Sand Forest associated with the ancient dune cordons and driest soils, it is suggested that historic plant dynamics could be a maior determinant of the presenl day Sand Forest patches.
Regular fires are a natural phenomenon in these areas, although today most fires are anthropogenic. As a result of humans, fires are more regular (sometimes more than once a year) and are now taking place at times 01 year when the, vegetation would nol naturally, on a regular basis, be exposed to fire. In many managed areas, current lire Irequency and intensities are lower than in the past, resulting in increased woody biomass in these areas (Van Wilgen and Scholes 1997). There can be no doubt that fire is an important natural ecological factor in these communities. In the past, it is probable that fires would have taken place annually to tri-annually, depending on biomass build up . These fires would occur during th e dry season (winter to early spring), when the grasses are dormant and at that time of year when lightning frequencies are the highest Margins of Sand Forest are often exposed to fire, which in some cases has a low impact, but at other times can be quite destructive. Many of these 'negatively' fire-impacted Sand Forest patches and margins do not regenerate at all. This leads us to the assumption that Sand Forest is decreasing as a result of fire . However, a comparison of aerial photographs of TEP taken in 1942TEP taken in , 1959TEP taken in , 1969TEP taken in and 1975 indicates that the Sand Forest patches we see today are very stable. There has been no increase in Sand Forest distribution , this being in contrast to the areas neighbouring Sand Forest which show a distinct increase in woody plants. There is indication of slight reduction of Sand Forest patches, as would have been predicted. This reduction is very slight and depending on Sand Forest dynamics in other areas and over the long term, should not be looked on negatively. However, the negative impact of large herbivores, such as elephant, which in the past were not closely associated with Sand Forest, in comb ination with fire , could promote Sand Forest shrinkage. Under the time scale al which Sand Forest dynamics seems 10 operale, these added impacts could negatively affect Sand Forest conservation within TEP.

Endemism
About 2 500 species, but probably more, of vascular plants occur in Ihe Maputaland Centre of Endemism (Van Wyk 1996). Of these, at least 230 species or infraspecilic taxa and three genera are endemic or near-endemic to the region (Van Wyk 1996). Eighty-four of these Maputaland Centre endemics/near-endemics were recorded in the study area (Table 4), but the actual number present is expected to be slightly higher due to incomplele sampling. Mosl endemics were associated with only two of the seven major plant communities. Fourty-two taxa are associated (27 restricted) with Drypetes arguta-Uvaria lucida subsp. virens forest (Sand Forest) and 37 (13 reslricted) 10 the Strychnos madagas-

cariensis-Combretum molle woodland. Most of the Sa/acia kraussii-Themeda triandra open woodland sub-communities' endemics have the geoxylic suffrutex growth form.
Hence this commun ity shows a strong association with the Themedeto-Sa/acietum woody grasslands described from the Sileza Natu re Reserve and surrounding area (Matthews et a/. 1999). Themedeto-Sa/acietum woody grassland , which occurs closer to the coast. is another plant commun ity noted for its high numbers of MC endemics. Another community relativel y rich in endemics is the Spirostachys africana-Berchemia zeyheri closed woodland on clay, where 20 MC endemic/near-endemic species have been recorded . This is a restricted commun ity in TEP, many of the endemics having a wider distribution outside the park on similar claybased substrates. In th e study area, by far the majority of plant endemics are confined to non-hygrophilous plant communities. The observed low species diversity and paucity of regional endemics in hygrophilous grassland communities is a common phenomenon throughout southern Africa.
The geological history of Maputaland suggests that the current ecosystems in the region may be of recent derivation . Th is part of the Mozambican Coastal Plain was last inundated by the ocean approximately 3-1.8 million years ago. In fact. many endemic plant taxa comply with the concept of neoendemics (young ta xa). Thus, the MC com prises a unique environment in Africa in that it is geologically young with biological evolution (notably speciation) being actively driven .
Maputaland Centre endemics/ near-endemics recorded in the study area , and even more so for the Centre as a whole , represent a wide spectrum of growth forms, including trees, shrubs , suffrutices , llanas, forbs , geophytes and ann ual herbs (Van Wyk 1996). Maputaland Centre endemics are also well represented in Sand Forest, woodland and grassland, with Sand Forest, on a regional basis, being perhaps th e single richest community in MC plant endemics. In contrast, th e Pondoland Centre of Endemism, located on the Pondoland Plateau to the south of the MC , is particularly rich in woody palaeoendemics associated wit h forest (Van Wyk 1994). It is hypothesised that one of the reasons for this marked difference in growth form and vegeta tion-type partitioning of endemics between these two centres of endemism is the relatively youthful (Quaternary) age of the exposed sandy Maputaland Coastal Plain and its associated plant communities. The species and plant comm unities had to adapt to more arid conditions inland as a result of sea level changes and resultant decrease in rainfall and soil moisture (wa ter table drop); the lowering of the water table from deeper sand deposits rather than diminished rainfall.

The southern Ka lahari, Similarly associated with sand
deposits , is also a relatively youthful ecosystem.

Accumulation of Kalahari sands began towards the end of
the Cretaceous , approximately 65 million years ago (Partridge and Maud 1987). Today many Kalaha ri sand plant endemics are to be found on the mesic northern areas of these sand deposits (Angola/Zambia/Democratic Republic of Congo/Zimbabwe). In contrast, the arid southern Kalahari sands (South Africa/Namibia/Botswana) contain very few local endemics. The southern Kalahari has been subjected to two major arid events, the second beginning in the Late 591 Pleistocene, although the re have been intermitted wetter periods during this time (Malherbe 1984). The most recent arid period began approximately 10 000 years ago -consequently little time was available for speciation. The southern Kalahari also borders on ancient arid areas (e.g. Namib Desert and Karoo) ; thus there has been the opportunity for pre-adapted arid species establishing themselves th rough migration in Kalahari areas subjected to arid ification . The latter invasions may have hindered in situ speCiation and the development of neoendemics. On the other hand the southern portion of the Mozambican Coastal Plain (Maputaland) we re at the time of its most recent exposure from the ocean apparently not bordered by any Significant pre-adapted , arid, psammophilous floras . This may have resulted in the availabil ity of many vacant niches on the sandy coastal plains subsequently occupied be newly evolved taxa derived from the available regional floras , hence the preponderance of neoendemics in the MC.

Sand Forest (= Licuati Forest)
Floristically, White (1983) classified the southern Africa forests as mainly belonging to two phytochoria, namely the Afromontane Arch ipelago-like Regional Centre of Endemism (which extends to the northeastern African uplands) and the Tongaland-Pondoland Regional Mosaic (which is largely restricted to the eastern coasta l belt of southern Africa) . His mapping was not at a fi ne enough scale to separate the 'Sand Forest' of northern KwaZulu-Natal and southern Mozambique from other Tongaland -Pond oland vegetation types. Moll and White (1978) conside red many typical Sand Forest species (e.g. Balanites maughamii, Cleistanthus sch/echleri and Newtonia hildebrandtil) to be Zanzibar-Inhambane linking species (the Zan zibar-Inhambane Reg ion is the coasta l phytochorial mosaic extendi ng from southern Mozambique to east Africa). This phytochorion, together with the Tongaland-Pondoland Reg ional Mosaic, form White's (1983) Indian Ocean Belt. Low and Rebelo (1996), on the other hand , divide the southern African forest types , and is thus class ified as tropica l dry forest. Marker and Russel (1984) have stated that the presence of mixed forest in low rainfall areas of the Eastern Cape can partly be ascribed to the presence of fog and the water holding capacity of deep sand , which sustain forest growth. Sand Forest in Maputaland is also found on deep sand , periodically experiencing heavy dews and lOW-level mists, espe-cially in winter. This could effectively increase the rainfall. In southern Africa forest at a loca l and regional scale generally shows low levels of tree endemism (White 1983, Cowling 1993, MaUhews et at. 1993. Similarly, few forest tree species are rare (MacDevette et al. 1989), and uncommon species are generally those in ecosyste ms such as Sand Forest that just penetrate South Africa from the northeast. A notable exception is the forests of the Pondoland Centre of Endemism. These forests are rich in endemic plant species , several of which appear to be palaeoendemics (Van Wyk 1990, 1994, 1996. Tinley (1971) mapped the distribution of a dry semi-deciduous forest type (probably equivalent to Sand Forest) as a th in zone from northern KwaZulu-Natal up to Quissico (approximately 100km south of Inhambane) , Mozambique. The northern limit (from Maputo-Quissico) of th is dry semi-deciduous forest is still unclear, as Tinley (1971) noted a change in species composition in the area north of Maputo Bay. Our work has shown that Sand Forest is also an exception in being a forest type rich in endemic species (though in this case mainly neoendemics). These neoendemic species are centred in an area south of Maputo-Quissico (Mozambique) and north of St Lucia in northern KwaZu lu-Natal (the core area of the Maputaland Centre) . Hence, as far as is known , Sand Forest (as floristically defined in this paper) is more or less confined to the MC.
In many parts of southern Africa and rest of Africa, the boundary between forest and adjacent vegetation types (usually grassland or woodland, ra rely fynbos) is very abrupt and in nearly all cases th is transition does not follow environmental discontinuities. The abruptness of forest boundaries is chiefly ascribed 10 fire (Hopkins 1979, Midgley et al. 1997, an important determinant ot torest boundaries and forest edge composition. Forest rarely burn , the fires usually stops at the forest margin. In the case of Sand Forest, not only does it have distinct boundaries, but it also has a narrow zone (1-2m) of nearly bare soil immediately adjacent to it. There are indications that allelopathic effects may playa role in the establishment of these narrow fringes of plant inhibition. Sand Forest, through allelopathic effects , possibly establishes a unique environment for itself, which also results in specific grassland communities neighbouring the Sand Forest. The resulting zone of plant inhibition may also protect Sand Forest against fire by acling as a kind of natural tirebrea k. Coetzee (1983) found similar abrupt boundaries (bu t no zone of plant inhibition) to one of the communities that he described from the central areas of the Kruger National Park, the Androstachetum. This communi ty is composed of dense stands (forest) of the evergreen tree Androstachys johnsonii (Euphorbiaceae) on arid substrates in areas periodically experiencing heavy winter fog . The sharp boundaries observed here can not be attributed to fire, but were hypothesised to be due to the ability of the stands of Androstachys johnsonii to absorb water from the wi nter fogs , thus out-competing other species in this environment. It has now been established th at the hairs in the dense indumentum on the leaves and twigs of this species can take up moisture (Alvin 1987).
In most parts of Africa forest types form mosaics with moist grassland and woodland. Relic t patches of

Matthews, Van Wyk, Va n Rooyen and Botha
Afromontane and Coastal Forest may occasionally be associa ted with savanna, but this is because of a restricted refug e, and the type of savanna more mesic. In the Cape Floristic Region Afromontane Forest is associated with wet or mesic evergreen sclerophyllous shrubland (fynbos). However, in the case of Sand Forest it occurs in a mosaic with dry woodland and thicket. Midgley (1993) claimed that Sand Forest is dominated by species with wind-dispersed seeds, whe reas other Indian Ocean coastal forests are dominated by fleshy-fru ited species. From an analysis of unpublished data of dispersal patterns in the Sand Forest of TEP, it can now be shown that only 5% of the species are definitely wind-dispersed. Th is also applies to the more coastal Sand Forest types . Leaf attributes are generally fa irly similar in all forests . Sand Forest is generally evergreen, non-spinous, with the leaves predominantly simple, micro-or mesophyllous and entire. Semi-deciduous trees do occur in Sand Forest, but these species are not dominant.

Threats and conservation
Conservation efforts in the Maputaland Centre have hi therto centred mainly on areas conta ining species of large game. The vegetation of nearly all existing nature reserves and game reserves in KwaZulu-Natal is dominated by va rious types of sava nna. A notable exception is the TEP that contains well-preserved stands of MC endemic-rich Sand Forest.
Outside conservation areas , the main threa ts to the Sand Forest habitat are uncontrolled fires used in clearing areas for agriculture, as well as selective species utilisation for traditional medicines, construction , fuel and other purposes. Examples of species selected for are Cassipourea mossambicensis that is used for cosmetic purposes (Brenton-Styles pers. comm.) and Cleistanthus schlechteri that is used in the wood carving trade. Unconlrolled cattle grazing , random uncontrolled fires , as well as selective species utilisation also have an effect on the other wood land areas outside conservation areas .
Inclusion of Sand Forest in conservation areas is no guarantee of the future persistence of this vegetation type. The effects of large herbivores, regular fires and Sand Forest dynamics can lead to disappearance of this vegetation type. One such threat to Sand Fores t within a conservation area is highlighted in TEP, the conservation area in Sou th Africa with the greatest amount of protected Sand Forest. The two primary reasons for the proclamation of TEP were that of Sand Forest conservation and the protection of the natural elephant population of Maputaland. Th is has lead to a clash of interests between Sand Forest and elephant. Although these elephant prefer plant species from woodland habitats, they are increasingly utilising Sand Forest species. This is mainly due to the fact that old movement patterns have been tenced off and the elephant are now confined to a re latively small reserve. Since the old elephant movement routes were closed off in 1989, negative effects on Sand Forest and associa ted vegetation types in TEP have increa sed markedly (MaUhews and Page unpublished results).