Online appendix 1:Vegetation of the eastern communal conservancies in Namibia: I. Phytosociological descriptions

Since Namibia’s independence there has been a substantial increase in efforts to develop the country’s communal farming areas. Many of these areas have been, and still are, subjected to poor and indifferent management practices resulting from a lack of sound ecological (baseline) data and hence a poor understanding especially of the vegetation resource base – its biodiversity, dynamics, possibilities of utilisation and its limitations (Hines 1992; Mendelsohn et al. 2002). This is also the case when looking at the communal lands in eastern Namibia, which consist of what was formerly known as Bushmanland, Hereroland West and Hereroland East. A phytosociological study on the vegetation of Bushmanland (part of the present Tsumkwe district) was conducted by Hines (1992). Information regarding the vegetation of the former Hereroland could thus far only be approximated from the preliminary vegetation map of Namibia (Giess 1998), as well as the Relative homogenous farming areas report (Department Landbou Tegniese Dienste 1979). The above report lists farming possibilities and constraints of the commercial farming area and includes a grazing-capacity map based on expert opinion. This map has been found to be outdated (Espach, Lubbe & Ganzin 2006; Lubbe 2005) mainly because the concept of carrying capacity has since changed from a fixed rate per area to a biomass concept, which is an (annually changing) estimate of the amount of fodder produced during the previous season’s rainfall (Bester 1988; Espach et al. 2006; Espach, Lubbe & Ganzin 2010). The preliminary vegetation map of Namibia is regarded as relatively accurate as far as the delineation of vegetation types is concerned, but some anomalies have been found. Furthermore, its descriptions of Namibian vegetation types are very basic. Both publications were based on limited observations and some photographic records, but no actual ecological surveys were conducted when they were compiled. Few attempts have since been undertaken to describe the vegetation of Namibia further (Burke & Strohbach 2000; Strohbach & Jürgens 2010). Page 1 of 18


Introduction
Since Namibia's independence there has been a substantial increase in efforts to develop the country's communal farming areas. Many of these areas have been, and still are, subjected to poor and indifferent management practices resulting from a lack of sound ecological (baseline) data and hence a poor understanding especially of the vegetation resource base -its biodiversity, dynamics, possibilities of utilisation and its limitations (Hines 1992;. This is also the case when looking at the communal lands in eastern Namibia, which consist of what was formerly known as Bushmanland, Hereroland West and Hereroland East. A phytosociological study on the vegetation of Bushmanland (part of the present Tsumkwe district) was conducted by Hines (1992). Information regarding the vegetation of the former Hereroland could thus far only be approximated from the preliminary vegetation map of Namibia (Giess 1998), as well as the Relative homogenous farming areas report (Department Landbou Tegniese Dienste 1979). The above report lists farming possibilities and constraints of the commercial farming area and includes a grazing-capacity map based on expert opinion. This map has been found to be outdated (Espach, Lubbe & Ganzin 2006;Lubbe 2005) mainly because the concept of carrying capacity has since changed from a fixed rate per area to a biomass concept, which is an (annually changing) estimate of the amount of fodder produced during the previous season's rainfall (Bester 1988;Espach et al. 2006;Espach, Lubbe & Ganzin 2010). The preliminary vegetation map of Namibia is regarded as relatively accurate as far as the delineation of vegetation types is concerned, but some anomalies have been found. Furthermore, its descriptions of Namibian vegetation types are very basic. Both publications were based on limited observations and some photographic records, but no actual ecological surveys were conducted when they were compiled. Few attempts have since been undertaken to describe the vegetation of Namibia further (Burke & Strohbach 2000;Strohbach & Jürgens 2010).
The target area selected for this study was the former Hereroland, which now consists of communal and emerging commercial farms within the Omaheke and Otjozondjupa regions of central-eastern Namibia. In order to empower the farmers to sustainably utilise their natural resources, especially wildlife resources, the establishment of communal conservancies is promoted in Namibia (Weaver & Petersen 2008;Weaver & Skyer 2005). At the start of the study (2004) The goal of the Desert Margins Programme (Namibia) was the conservation and restoration of biodiversity in Namibia's desert margins. One of the major outputs was the establishment of baseline information on natural resources available to local communities in selected areas. In order to establish such baseline information, a survey of the vegetation types was commissioned for these communal conservancies as a baseline of the ecosystems to be managed. This is in line with generally proposed uses of such vegetation descriptions and maps (Brown et al. 2013;Strohbach 2012). The nomenclature throughout this article follows Klaassen and Kwembeya (2013).

Climate
Very little meteorological data has officially been recorded within the study area. Climate variables for the area have been extrapolated from records collected in Gobabis, Grootfontein, Tsumkwe and Ghanzi (in Botswana). Average long-term annual rainfall ranges from about 325 mm to 350 mm in the south-eastern corner (Rietfontein Block) and the westernmost part (around Okakarara) and from about 425 mm to 450 mm in the central part of the arearoughly from Okondjatu and Otjinene to the Mangetti Dune    Figure 1). An important aspect of this rainfall is its often erratic nature within the rainfall season and from year to year. Droughts are regarded as common (Botha 1998;Leser 1972;Mendelsohn & El Obeid 2002).
During winter, mean daily temperatures range from 2.5 °C (minimum) to 22.2 °C (maximum) (with 8 frost days) in the southernmost parts of the study area and from about 4.3 °C (minimum) to 23.6 °C (maximum) (with 2 frost days) in the northern parts. During summer, mean daily temperatures range from 17.4 °C (minimum) to 31.8 °C (maximum) in the southern parts and from 17.9 °C (minimum) to 30.2 °C (maximum) in the northern parts. However, daytime temperatures can soar to just below 40.0 °C, especially during November and December (Mendelsohn & El Obeid 2002).

Landscapes and geology
The study area is situated on the vast Kalahari System (Geological Survey 1980;Thomas 1988;Mendelsohn & El Obeid 2003). This ancient basin has been gradually filled up with silt and clay sediments and windblown sands during the Tertiary to Quaternary Periods (32-39 Ma) (Geological Survey 1980). The windblown sands dominate the surface soils of the study area. Occasionally, exposed rock formations can be found -these are mainly sandstones, limestones, schists and dolomites of the much older Karoo and Damara Sequences.
The sandy landscapes are generally flat to rolling (with 6° -9° slopes); these plains are incised by omuramba valleys or alternated with vegetated fossil (no longer actively moving) dunes. An omuramba (plural: omiramba) can best be described as a shallow watercourse with no visible gradient (King 1963). Leser (1972) differentiates between coarse-grained yellow sands, which were found mostly on plains, and finer-grained reddish sands, present especially on the fossil dunes. Apart from the omuramba valleys, which often have exposed or underlying calcretes and limestone ridges, occasional pans formed by the limited runoff in the area have grey clays to grey sands. A more thorough study of the soils and landscapes is presented by Strohbach et al. (2004).

Procedure
Surveying followed the general method employed for the vegetation survey of Namibia project (Strohbach 2001). Sampling of 422 plots throughout the study area took place during April and May 2004. Owing to the vastness of the area, and the limited access to it, emphasis was placed on the sampling of dominant vegetation types (rather than rare niche types) along roads, tracks and cutlines, following a stratified systematic approach. False colour LandSat Thermatic Mapper (TM) imagery was used for stratification. At each survey plot of 20 m × 50 m (1000 m 2 ), a Braun-Blanquet relevé was compiled. All occurring vascular plant species were noted down, as well as their typical growth forms and estimated percentage crown cover. Habitat descriptors included: the position by way of a GPS-reading (referenced to WGS84), the landscape, local topography, slope and aspect, lithology, degradation indicators, as well as a photograph. Unknown species and reference specimens were collected, identified and deposited at the National Herbarium of Namibia (WIND).
The relevé data was captured on Turboveg (Hennekens & Schaminée 2001) and forms part of the Namibian phytosociological database (GIVD ID: AF-NA-001) (Strohbach & Kangombe 2012;see Dengler et al. 2011). For the original report (Strohbach et al. 2004), this data was classified with PhytoTab (Westfall et al. 1982). The initial study identified three associations, which were represented by a limited number of relevés, but which are known also from other studies (Hines 1992, own unpublished data). For this reason, relevant relevé data available in the Namibian phytosociological database (GIVD ID: AF-NA-001) from the Tsumkwe district (25 relevés from Hines 1992 surveyed on 400 m 2 plots) and from the southern Mangetti area near Tsintsabis in the Grootfontein district (five relevés from own unpublished data surveyed on 1000 m 2 plots) was extracted and included in the dataset. The final dataset consists of 451 relevés, with 448 species. For the present study the classification was repeated using Juice 7 (Tichý 2002). Total inertia was used as diversity measure, whilst pseudospecies cut levels set at 0%, 2%, 5% and 20%, were utilised because the gradients were often very gradual, resulting in strong overlapping of species distribution. For convenience, the dataset was split into two, using the first division of a Modified two-way indicator species analysis (TWINSPAN) classification (Roleček et al. 2009).
The two relevé groups were further classified, again using Modified TWINSPAN with the same settings as before. The highest level of divisions still yielding floristically wellcharacterised clusters was used. After inspection of the results, it was realised that further refinements were necessary on some of the clusters, based on both field observations and previous vegetation descriptions, in particular Cole and Brown (1976), Hines (1992), Leser (1972) and Strohbach et al. (2004). This was performed with Cocktail (Bruelheide 1997;Bruelheide & Flintrop 1994) using known character species for selection. In several cases, recombinations of clusters as subassociations or even associations were necessary to produce ecological significant results. The fact that such clusters could not be differentiated with diagnostic species but had similar habitat, structure and dominant species composition was taken as criteria for recombination.
Diagnostic species were determined using the phi co-efficient of association (Chytrý et al. 2002). For this calculation the relevé numbers were standardised following Tichý and Chytrý (2006). Species with phi ≥ 40 were considered as diagnostic and with phi ≥ 60 as highly diagnostic; however, species with a non-significant fidelity at α = 0.05 using Fisher's exact test were omitted. Species occurring at least with a 60% frequency were regarded as constant and with at least an 80% frequency as highly constant.
Further calculations to describe the structure and species richness were performed using the available growth form data and species richness and density. For the calculation of species density (number of species per 1000 m 2 ), the relevé data from Hines (1992) was excluded, as these were sampled on 400 m 2 plots, not 1000 m 2 plots as all other relevés.

Results
The classification results are depicted in Figure 2 as a dendrogram. The first division of the dataset represents two distinct habitat and physiognomic groupings: the 'sandveld' or broad-leafed savannas on very sandy habitats (265 relevés with 305 species) and the 'hardeveld' ('hard veld') on slightly to very loamy soils, occasionally even shallow rocky substrates with microphyll savannas (186 relevés with 402 species).
Within the hardeveld (Figure 2 top), the second cluster was split into two associations with the aid of Cocktail (Bruelheide 1997;Bruelheide & Flintrop 1994), using the presence of Felicia clavipilosa subsp. clavipilosa, Lycium bosciifolium, Ptycholobium biflorum subsp. angolensis and Xerophyta humilis versus Combretum imberbe. In addition to differences in composition, these two associations were recognised in the field as different in structure and habitat. A third association (no. 5), based on the dominance (> 5% cover) of Catophractes alexandri, but the absence of indicator species Eriocephalus luederitzianus and Hyphaene petersiana, was defined also using Cocktail. This is in line with various literature sources, in particular Cole and Brown (1976), Hines (1992), Leser (1972) and Strohbach et al. (2004). The initial clusters 8, 9 and 10 could not be differentiated through diagnostic species, only through subtle differences in constant and dominant species. Therefore clusters 9 and 10 were combined as a subassociation (10.2), which, in turn, was combined with cluster 8 (subassocaition 10.1) to form a larger association (10) because of the high similarity in permanent species composition, structure and habitat.
Within the group of relevés representing the broad-leaved savannas, again, Cocktail had to be employed to strengthen cluster 1, based on the presence of Pterocarpus angolensis and Burkea africana. This cluster was also split into two, based on the presence of Schinziophyton rautanenii and Strychnos pungens. These refinements were based on findings from the original study (Strohbach et al. 2004) as well as Hines (1992). Clusters 2 to 5 were recognised as subassociations of a single association, because of the very similar structure, dominance by Terminalia sericea and various Combretum species and similar habitat. In effect, these four subassociations represent a precipitation gradient (Strohbach & Kutuahuripa submitted).
The full phytosociological table is presented in downloadable form in Online Appendix 1 and, similarly, the synoptic table in Online Appendix 2. The classification yielded 11 communities, of which 10 are formally described associations, following the International Code for Phytosociological Nomenclature (Weber, Moravec & Theurillat 2000). Species with high fidelity (phi ≥ 60) and high constancy (frequency ≥ 80%) are indicated in bold. Structural description follows Edwards (1983), whilst habitat descriptions follow the soil and terrain database (SOTER) terminology (Food and Agricultural Organization [FAO] 1995).

Acacio fleckii-Terminalietum prunioidis ass. nov.
The type for this association is relevé 1336, sampled on 29 April 2004 at 20°12'40" S, 20°46'22" E. Only two relevés of this association were sampled in the present study area, but this association occurs extensively within the Nyae-Nyae Conservancy further north (Hines 1992). An additional 15 relevés from the Hines (1992) Hines (1992) refers to these thickets as Terminalia prunioides association and describes them as occurring on some sub-outcropping geological features, specifically fold structures (e.g. near Cin Qo) throughout the Nyae-Nyae Conservancy. Bearing out Hines's (1992) findings, these thickets were found on low limestone ridges, mostly covered a Source: Author's own creation  by shallow sand deposits with few stones on the surface (< 1% cover), north of Gam. Here the altitude ranges between 1110 m.a.s.l. and 1130 m.a.s.l., with generally rolling slopes. The soils are mostly shallow. As a result of the loose sand cover, few or no signs of crusting, but slight to moderate signs of erosion were observed. The thickets sampled in this study are situated near the resource-poor settlements of Gam and showed severe signs of degradation through overgrazing and felling of trees for timber.
Terminalia prunioides dominated the tree layer at between 40% and 60%. The shrub understorey was, on average, less than 10%, making this association a tall, closed woodland. Grass cover was generally variable, dominated by annual grasses (Figure 3a and Figure 4a). Only two perennial grass species were observed, these being Panicum maximum and Cenchrus ciliaris, indicative of the poor grazing resource these thickets represent. Although this could be attributed to overgrazing in the area, heavy shading by the overstorey would also prevent the growth of good grazing. Very striking are the tall Combretum imberbe (Leadwood) trees interspersed with smaller trees of Acacia tortilis in this association. The understorey of shrubs is variable, mostly less than 10% cover, making this a tall, semi-open bushland ( Figure 3b and Figure 4b). The Omuramba Omatako near Okakarara has a distinct wetland characteristic, with Eragrostis rotifer present in two, and Schoenoplectus senegalensis and Marsilea macrocarpa present in one, of the sites -all species common in areas with seasonal standing water. Although some of the patches are drier and have a higher cover of grasses, including some perennial grasses, this syntaxon is to be regarded as having a low grazing capacity. If left ungrazed during the rainy season (which it was evidently not), this could, like the oshanas in the central north, be valuable winter grazing. The Omuramba Omatako is at an altitude of between 1290 m.a.s.l. and 1310 m.a.s.l. south of Okakarara and it flows in a north-easterly direction towards Otjituuo. The slopes are flat to gently undulating, whilst the soils are strongly crusted. At one of the sites, signs of selective wood harvesting of Combretum imberbe was observed. This association has a consistent tree layer of about 6 m -8 m in height with a cover of about 12%. The shrub layer reaches a cover of about 24%, thus making the structure a short, semiopen thicket (Figure 3c and Figure 4c).

Ptycholobio biflori-Acacietum luederitzii ass. nov.
The Ptycholobio biflori-Acacietum luederitzii occurs typically on the floodplains around the Omuramba Omatako, south of Okakarara in the Ozonahi Conservancy, characterised by the erosion plains around the Omatako Omuramba throughout its catchment. These floodplains are gently undulating and generally at an altitude of 1290 m.a.s.l. -1330 m.a.s.l. The soils are shallow, with a hardpan layer at a depth of 30 cm, and generally strongly crusted. Only slight sheet erosion was observed. These floodplains are (relatively) undisturbed; although, a few show signs of bush encroachment. On one occasion severe overgrazing was observed near a village.
Grazing value is low, with only two perennial grasses (Eragrostis trichophora and Aristida congesta) being found consistently. Both are low producers of biomass, whilst Aristida congesta is also known as generally hard and unpalatable (Müller 2007). The annual grass layer is dominated by Eragrostis porosa and Aristida rhiniochloa. The latter can produce fairly high amounts of biomass, but is a very hard grass with a strong, pungent seed. Both these species are regarded as unpalatable. The poor grazing condition is aggravated by the consistent occurrence of the poisonous Geigeria ornativa. With a variable, but very low, tree cover and low shrubs between 40 cm and 120 cm in height dominating the landscape, the structure can best be described as a low, semi-open shrubland ( Figure 3d and Figure 4d).
This association occurs in omiramba with heavier loamy and clay soils, as well as a pan near Otjutuuo in the far north. The soils are often strewn, up to 15%, with calcrete fragments of sizes ranging from gravel to large stones. Rocks were even found on one site. The soil surface is moderately to strongly crusted, showing signs of slight sheet erosion. The slopes are generally flat to undulating. These omiramba occur at altitudes of 1180 m.a.s.l. -1330 m.a.s.l.
Although the Eragrostio echinocloideae-Eriocephaloetum luederitziani occasionally has a very high cover of grasses, this is not the norm. The dominant grass species are Eragrostis echinochloidea, Eragrostis porosa, Enneapogon desvauxii, Tragus racemosus and Stipagrostis uniplumis var. uniplumis. However, only the latter species is a bulk biomass producer. Enneapogon desvauxii in particular is a very short grass, about 5 cm high, often forming a fairly dense cover. It is a fast grower, producing leaves and seed within a short period of time (thus the vernacular name 'Agt dae gras' [eight day grass]). The combination of species -Enneapogon desvauxii, Eragrostis echinochloidea, Eriocephalus luederitzianus and Salsola tuberculata -is evident of a harsh microclimate dictated by soil conditions (high pH and/or sodic). The grazing production in this habitat is thus low and the grazing capacity must equally be regarded as low.

Acacio melliferae-Catophractetum alexandri ass. nov.
The type for this association is relevé 1260, sampled on 24 April 2004 at 20°35'44" S, 18°05'26" E. Twenty-one relevés were classified into this association, which is characterised by the occurrence of Catophractes alexandri as diagnostic species. This species also dominates the landscape with an average 20% -40% cover. It is constantly associated with Grewia flava, Acacia mellifera subsp. The Acacio melliferae-Catophractetum alexandri is strikingly dominated by more or less dense stands of Catophractes alexandri. Trees are scarce to totally absent, whilst an average of 40% -60% shrub cover of between 1 m and 3 m in height is characteristic, thus it can be described as a tall, closed shrubland ( Figure 3e and Figure 4e). Hines (1992) described a similar vegetation unit (the Catophractes alexandri subassociation of the Croton gratissimus-Grewia flava association) for the Nyae-Nyae Conservancy. This unit is described as occuring on soils with shallow underlying calcretes. As localised stands of dense Catophractes alexandri can be found throughout northern Namibia, always on relatively shallow calcrete soils, this community is recognised as a separate association, even though it is closely related to the Stipagrostio uniplumis-Acacietum melliferae.
Catophractes alexandri is regarded a typical indicator for petric Calcisols (i.e. shallow loamy soils with calcretes) (Cole & Brown 1976;Hines 1992). Such soils occur in some of the omiramba, in interdunal valleys and around larger pans and depressions. Generally, the slopes are flat to gently undulating, at altitudes of between 1220 m.a.s.l. and 1450 m.a.s.l. The soil surface is often covered, but never more than 2%, with fragments of the underlying calcretes that vary in size from small (gravel) to large (stones). Leser (1972) describes similar soil conditions for his Catophractes alexandri shrublands along the Black Nossob. The soil surface was also observed to be moderately to strongly crusted, showing signs of slight sheet erosion. In one case, slight rill erosion was observed. Bush encroachment is a common disturbance.
The most prominent perennial grasses are Stipagrostis uniplumis, Aristida congesta, Eragrostis rigidior, Eragrostis trichophora and Cenchrus cilliaris. These are all known to be neither highly palatable, nor indicators of good veld (Müller 2007). Most grass species are annuals and also typical pioneer or subclimax species. It is, however, commonly believed that the grazing on these calcareous soils is sweeter and more nutritional, thus the quality of grazing is considered moderate.
6. Acacio melliferae-Hyphaenetum petersianae ass. nov. The typical structure is a tall, moderately closed bushland, with a tree layer varying between 7 m and 12 m in height, reaching a cover of 5%, with shrubs covering about 45% of the landscape (Figure 3f and Figure 4f). The soils are generally shallow to very shallow on calcrete and/or limestone, with a calcrete fragment cover of up to 40% on the soil surface. The surface is also moderately to strongly crusted, occasionally with signs of slight sheet erosion. Bush encroachment, often severe, is a common problem here, as it is in the Karstveld in general.
The grazing condition is low, with only two perennial grasses (Eragrostis trichophora and Stipagrostis uniplumis) occurring consistently, but at a very low abundance, at the sample sites. Some of the annual grasses found here are known to be palatable, but all are low biomass producers, whilst many of them are also shade grasses and difficult to reach under the trees and shrubs where they grow.

The Panicum gilvum-Marsilea vlei community
The vleis are represented only by one relevé, which is regarded as inadequate to warrant a formal description of these wetland grasslands as an association. However, for the record, the following is presented: characteristic for this association are the species Tree cover is virtually absent, whilst the shrub cover is generally absent or very low. If shrubs are present, their cover is below 20%. Small dense patches of Tarchonanthus camphoratus or Acacia hebeclada subsp. hebeclada in particular may occasionally be found, usually where water accumulates during the rainy season. Accordingly, this association is best described as a tall, semi-open shrubland with a strong grassland character (Figure 3g and Figure 4h). The characteristic Acacia erioloba occurs here mainly as tall shrub and occasionally as short to tall trees. Thus the typical structure is a short, moderately closed shrubland within the study area (Figure 3h and Figure 4i). Leser (1972) describes an Acacia giraffae-Terminalia sericea-Grewia flava-Tarchonanthus camphoratus association for the south-western part of the sandveld -that is, the area south and west of Otjinene towards Summerdown and Steinhausen. Here, Acacia erioloba (syn. = Acacia giraffae) occurs as a dominant tree, typical for the Camelthorn savanna sensu Giess (1998). It is clear that the relevés sampled from this association within the present study are along the northern fringe, thus not displaying the characteristic structure.
The grazing is dominated by the two relatively hard, bulkproducing species, Stipagrostis uniplumis var. uniplumis and Eragrostis rigidior. Schmidtia pappophoroides, Digitaria seriata, Anthephora pubescens and Brachiaria nigropedata are occasionally present as palatable species. Grazing trails at the Sandveld Research Station near Drimiopsis indicate that the latter three species could contribute a far higher percentage to the grazing production (Bester & Reed 2003;Bester et al. 2003aBester et al. , 2003bBester et al. , 2003c. The poisonous species Elephantorrhiza elephantina and Gnidia polycephala are commonly found, indicating, together with the relatively low occurrence of palatable species, that the grazing within the study area has been degraded and, at present, produces only moderately as subclimax.

Stipagrostio uniplumis-Acacietum melliferae ass. nov.
The type for this association is a subassociation, the Stipagrostio uniplumis-Acacietum melliferae typicum. Fifty-nine relevés were classified within this association. This association has no diagnostic species, but is typified by the constant occurrence of Acacia mellifera subsp. Trees do occur in this subassociation, but generally at very low cover. With a high shrub cover, this subassociation is best described as a short, moderately closed bushland (Figure 3i and Figure 4j). With a highly variable grass cover, this subassociation has a moderate grazing capacity.
Stipagrostio uniplumis-Acacietum melliferae rhigozetosum brevispinosi is a typical ecotonal vegetation type towards sandy soils, fringing the Stipagrostio uniplumis-Acacietum melliferae typicum on more loamy soils. This is also evident from the species composition, with its numerous psammophyll species. This subassociation occurs on sand drift or 'hardeveld' plains (i.e. plains with more loamy soils), with moderately deep to very deep soils and weak surface crusting. Bush encroachment is a common disturbance. The altitude of these plains ranges between 1030 m.a.s.l. and 1470 m.a.s.l. Slopes vary between flat and undulating, whilst the average annual rainfall varies between 400 mm and 500 mm. Although some trees are present, their cover was found to be very variable. However, the shrub component is very consistent, with a cover of 40% -50% and a height of 3 m -5 m, thus forming a high, moderately closed bushland sensu Edwards (1983) (Figure 3j and Figure 4k).

Stipagrostio uniplumis-Acacietum
The Stipagrostio uniplumis-Acacietum melliferae typicum occurs on plains, interdunal valleys ('streets') and sand drift plains throughout the study area. Slopes range from flat to undulating. The altitude of these plains ranges from 1120 m.a.s.l. to 1460 m.a.s.l. The soils are mostly sandy and deep, only occasionally with signs of calcretes. The soil surface is normally crusted, ranging from moderate to strong. As in Stipagrostio uniplumis-Acacietum melliferae rhigozetosum brevispinosi, bush encroachment is also a widespread problem. With Acacia erioloba trees being the most prominent species, this association can best be described as a short, moderately closed bushland (Figure 3k and Figure 4l). The Terminalio sericeae-Acacietum eriolobae resembles the Terminalia sericea-Acacia giraffae-Grewia flava association described for the north-eastern sandveld area (Leser 1972) and is thus a typical example of the Camelthorn savanna of the Central Kalahari sensu Giess (1998). The dominance of Aristida stipitata and Terminalia sericea demonstrates the nearness of this association to that of Combreto collini-Terminalietum sericeo. These bushlands are to be seen as an ecotone between the broad-leaved savannas to the north and east and the microphyll Thornbush savannas to the south and west.
This bushland association occurs on gently undulating to undulating plains and sand drift plains and even on interdunal streets at an altitude of between 1020 m.a.s.l. and 1490 m.a.s.l. Soil depth varies, but is mostly moderately deep to very deep. Surface crusting is variable -from uncrusted to moderately and even strongly crusted. Likewise, erosion varies between slight to moderate wind erosion and slight to moderate sheet erosion.
The typical poisonous plants for this association are Elephantorrhiza elephantina and Gnidia polycephala. Both are more or less equally frequent. Occasionally, both Dichapetalum cymosum and Geigeria ornativa also occurthe former as remnant from the more mesic broad-leaved savanna, the latter as an indicator of degradation. The medicinal plant Harpagophytum procumbens occurs in small patches throughout this vegetation type, but densities are too low to warrant economic exploitation.

Combreto collini-Terminalietum sericeae ass. nov.
The type for this association is a subassociation, Combreto collini-Terminalietum sericeae typicum. A total of 243 relevés were classified into this association. This association covers most of the landscape of the eastern communal areas. Owing to slight differences in habitat (mainly rainfall variability but also, to a lesser extent, soil depth and loam content of the sand), four subassociations can be recognised.
Livestock is equally dependant on grazing as on browse.
A widespread threat to livestock farming in this area is the poisonous plant Dichapetalum cymosum ('magou', 'gifblaar'), which is known to increase under conditions of overgrazing and disturbance (Bester 1989 The structure is best described as a short, moderately closed bushland (Figure 3l and Figures 4m and 4n). The variable grass cover often observed in this subassociation is most likely the result of the severe overgrazing, often resulting in bush encroachment. The grass layer is dominated by the annual Megaloprotachne albescens and Urochloa brachyura, with Stipagrostis uniplumis var. uniplumis and Eragrostis trichophora being the main perennial species. At best, the grazing capacity is poor to mediocre.
This subassociation forms a patchy ecotone to the Thornbush savannas to the south and west, probably because of localised, less favourable soil-moisture regimes (either from higher runoff on the slopes of omiramba, or more xeric soils). The altitude ranges between 1200 m.a.s.l. in the east and north to 1500 m.a.s.l. in the south-west, as this vegetation type gradually changes to Stipagrostio uniplumis-Acacietum melliferae.

Combreto collini-Terminalietum sericeae grewietosum flavae subass. nov.
The type for this subassociation is relevé 1439, sampled on 29 April 2004 at 20°05'05" S, 20°29'08" E. Forty-five relevés were classified into this subassociation, which is characterised by the diagnostic species Tree cover is virtually absent, making this subassociation a high, moderately closed shrubland (Figure 3m and Figure 4o). It occurs on sand drift plains and occasionally in sandy omiramba at an altitude of between 1040 m.a.s.l. and 1450 m.a.s.l. The sands are generally deep to very deep. The most common disturbances observed are bush encroachment and sometimes severe overgrazing.
Next to Dichapetalum cymosum, Elephantorrhiza elephantina also occurs occasionally. Although this suffrutex is less poisonous to cattle, it is still a threat to grazers. Grass cover in this variation is in general low. The annual species Megaloprotachne albescens, Urochloa brachyura and Melinis repens subsp. grandiflora dominate the grass sward. Important perennial species are Stipagrostis uniplumis var. uniplumis, Eragrostis rigidior, Eragrostis lehmanniana and Eragrostis trichophora. In general, the grazing is regarded as poor to mediocre.

Combreto collini-Terminalietum sericeae typicum
subass. nov. Combreto collini-Terminalietum sericeae typicum only occasionally has trees, making it a short, moderately closed bushland with a strong shrubland character, dominated by Terminalia sericea and various Combretum species -of which Combretum collinum is the most prominent (Figure 3n and Figure 4p). The sandy habitat occurs on extensive plains ranging between 1020 m.a.s.l. and 1500 m.a.s.l. altitude.
The grazing is dominated by the annual Megaloprotachne albescens, followed by Aristida stipitata. Characteristic of this subassociation are Burkea africana trees amidst a shrubland dominated by Terminalia sericea and various Combretum species. The structure is thus a typically short, moderately closed bushland (Figure 3o and Figure 4q), found mostly on sandy plains, occasionally also on dunes, between 1030 m.a.s.l. and 1460 m.a.s.l. altitude. Grazing conditions approximate those found in Combreto collini-Terminalietum sericeae typicum.
13. Terminalio sericeae-Schinziophytetum rautanenii ass. nov. Terminalio sericeae-Schinziophytetum rautanenii has been found on the crests and upper slopes of the high longitudinal dunes just north of the Rooiboklaagte omuramba (21°00' S). Additional Schinziophyton rautanenii (Manketti tree) groves could be identified from a distance at a few locations in the Gam and Eiseb area. The dunes on which the Manketti groves occur vary in altitude between 1240 m.a.s.l. in the south and only 1110 m.a.s.l. further north. Long-term average annual rainfall ranges between 400 mm and 500 mm. The deep sands found here (> 150 cm) show only few signs of wind erosion and weak surface crusting, with almost no disturbances. The tall Manketti trees associated with a dense understorey shrub layer result in a tall, moderately closed thicket (Figure 3p and Figure 4r). This association occurs mostly in the northern sandy plains of the Otjituuo and Gam Conservancies. Patches of this vegetation type can also be found in the northern Okamatapati and Otjinene Conservancies. It is fairly widespread further north and has been described as 'Burkea africana tall closed woodland' by Hines (1992). Typically, the habitat consists of extensive sandy plains ranging at altitudes of between 1050 m.a.s.l. and 1340 m.a.s.l. The depth of the sands is generally more than 150 cm, with weak or no surface crusting and signs of only slight wind erosion. Typical disturbance factors in these bushlands are frequent fires and bush encroachment.

Burkeo africanae-Pterocarpetum angolensis
Although Pterocarpus angolensis (teak tree) is a very prominent species, reaching a height of 12 m -15 m, neither its density nor its trunk diameter warrants commercial exploitation of this timber resource. Few of the other woody species reach tree height, the only ones being Burkea africana and, occasionally, Terminalia sericea and Combretum collinum.
Owing to the relatively high cover of understorey shrubs, the structure is typically a tall, moderately closed thicket (Figure 3q and Figure 4s).

Higher syntaxonomic groupings and relations to similar vegetation in southern Africa
The initial splitting of the dataset between the sandveld and the hardeveld -that is, between the deep sands of the central Kalahari basin and the surrounding transitions to the Central Plateau (De Pauw et al. 1998), with more mesic conditions on shallower soils and/or soils with higher loam content -is reflected as between mostly the broad-leaved savannas (associations 12-14) and the microphyll savannas (associations 1-11).
The broad-leafed savannas in turn can be split into two major groupings: Strohbach and Petersen (2007) suggest the recognition of a vegetation class Burkeo-Pterocarpetea, which is considered representative of the Zambesian-Baikiaea Woodlands ecoregion of the World Wildlife Fund (WWF) (Vetter 2001    Caesalpinioideae, compared to the communities described for the central Kavango (Strohbach & Petersen 2007). These two associations thus form the southernmost, most xeric extent of Burkeo-Pterocarpetea and the Zambesian-Baikiaea Woodlands ecoregion of the WWF (Vetter 2001), extending into north-western Botswana. Bekker and De Wit (1991) make no mention of such species combinations in their vegetation description, though.
In contrast, the Combreto collini-Terminalietum sericeae has no woodland character and generally forms a shrubland, at best a bushland. The Caesalpinioidae are represented only by Bauhinia petersiana subsp. macrantha (a shrub) and occasionally Burkea africana (as a short tree). Owing to, especially, these structural, but also to compositional differences, this association is to be recognised as a separate higher syntaxon, possibly a class Combreto-Terminalietea sericeae. The fact that the subdivisions of Combreto collini-Terminalietum sericeae are recognised only as subassociations is because of the very close compositional relationship between them and this needs to be confirmed through further studies. This association covers most of the landscape of the eastern communal areas and is expected to make a significant contribution to the landscape of the Kalahari Acacia-Baikiaea woodlands as defined by the WWF (Spriggs 2001a (Scholes & Walker 1993).
The subdivision of the hardeveld is not as clear-cut. It is speculated that the Acacio fleckii-Terminalietum prunioidis and the Acacio melliferae-Hyphaenetum petersianae could potentially form a separate class. This is also suggested by Giess (1998), who recognised the Karsteveld as separate from the Thornbush savanna. Similar vegetation, especially to Acacio fleckii-Terminalietum prunioidis, has been described by Cole and Brown (1976) for the Ghanzi Ridge area north-east of Ghanzi towards the Ngamiland border.
The Eragrostis echinocloideae-Eriocephaloetum luederitziani and the Acacio melliferae-Catophractetum alexandri form typical niche vegetation related to shallow carbonate-rich soils. There are obvious resemblances between these two associations, warranting their own higher syntaxonomic unit. Acacio melliferae-Catophractetum alexandri, especially, occurs widespread throughout eastern and north-eastern Namibia (Hines 1992;Leser 1972), but also in north-western Botswana (Cole & Brown 1976). It could not be established whether similar vegetation exists within the Kalahari Bushveld types of South Africa (Mucina & Rutherford 2006).
The remainder of the microphyll savannas are typical for the Acacietea (Volk & Leippert 1971); that is, they are dominated by various species of the genus Acacia. The Acacietea are divided by both Giess (1998) and the WWF into two larger units, these being the Thornbush savanna (roughly equivalent to the Namibian savanna woodlands -Spriggs 2001b) and the Camelthorn savanna, which forms the northern part of the Kalahari xeric savanna (Seymour 2001). Typical Camelthorn savanna elements are the Terminalio sericeae-Acacietum eriolobae and the Tarchonantho camphorate-Acacietum eriolobae. Whereas the former association seems to occur widespread in western Botswana (Cole & Brown 1976) and has been described also from the Sandveld Research Station north of Gobabis in Namibia September 2006), Tarchonantho camphorate-Acacietum eriolobae has resemblances to the Kimberley Thornveld (SVk 4) or Kuruman Thornveld (SVk 9) of South Africa (Mucina & Rutherford 2006).

Vegetation mapping
The initial vegetation map prepared by Strohbach et al. (2004) was refined by Hüttich et al. (2009) and is presented in that publication. For the mapping of the vegetation types, a number of associations had to be combined, in a similar way as was undertaken for the SOTER methodology (FAO 1995) or the landscapes of the Kalahari Gemsbok National Park (Van Rooyen et al. 2008). For purposes of comparison, the mapping units described by Hüttich et al. (2009) are related to the vegetation associations described in Table 1 in this article.

Land use and land degradation
Large parts of the vegetation have been found to be either over-utilised, subject to regular heavy fires, or bush encroached. In general, bush encroachment in these areas is regarded as high (De Klerk 2004), thus further threatening the grazing resources. With this in mind, many farmers resort to the harvesting of droppers (fence posts) from Terminalia sericea as an extra source of income. Overgrazing, especially if combined with overutilisation of this woody resource and/ or frequent fires in the area, will lead to an increased exposure of the soil surface to both wind erosion and denitrification, the latter as a result of increased heating of the surface during the day (Schlesinger & Peterjohn 1991;Schlesinger et al. 1996). The Kalahari sands are, in any case, extremely nutrient-poor (Dougill & Thomas 2004;Wang et al. 2007)such a denitrification cycle will lead to further reduction in the soil nutrient status and thus to further reduction in the vegetation cover. This is a classic example of 'biological feedback' into desertification (Schlesinger et al. 1990), in an ecosystem which is, in any case, predicted to aridify and turn into a mobile sand sea over the next half century as a result of global climate change (Midgley et al. 2005;Thomas, Knight & Wiggs 2005). This will mean an even more rapid resource loss over the entire area, resulting in major socioeconomic problems for the country. To ensure sustainable land use in the area, it is essential that good vegetative cover be maintained, whether dominated by grass, shrub or tree.

Conclusion
The vegetation within the eastern communal conservancies of Namibia forms a large expanse of transitional vegetation between the xeric southern Kalahari and the woodlands of the northern Kalahari, with adjacent transitions to the Central Plateau with its stonier and/or loamier soils. As the soils are generally nutrient-poor (Dougill & Thomas 2004;Wang et al. 2007), the entire ecosystem is extremely fragile and susceptible to desertification (Thomas et al. 2005). In order to prevent this, proper land-use planning is essential for the emerging and established communal conservancies. This article provides important baseline information needed for proper integrated land-use planning, also in line with the National Land Reform Programme of the Government of Namibia (Haub 2009). This article also provides an important contribution towards the inventorying of vegetation in Namibia under the vegetation survey of Namibia project (Strohbach 2001;Strohbach & Jürgens 2010). The newly described vegetation associations represent four important ecoregions in Namibia, being the Forest Savannas and Woodlands, the Karstveld, the Thornbush Savanna and the Camelthorn savannah sensu Giess (1998). The higher order syntaxonomy of these associations is still unclear, mainly because only limited similar studies exist within Namibia. This makes an assignment to a specific syntaxonomic rank impossible.
Future studies should concentrate on the description of adjacent areas especially within the Karstveld, Thornbush and Camelthorn savannas, to confirm and expand the classification results of the hardeveld. As more such studies become available, a better understanding of the higher order syntaxa will become available. A phytosociological study to the seasonal wetlands of the Kalahari, building on pioneer work by Hines (1993), will assist to deepen the understanding of these azonal features, which also form a major habitat in this ecoregion.
There is an urgent need for an online register of vegetation associations for southern Africa, as has been suggested by Brown et al. (2013). This will facilitate the proper description and comparison of vegetation types in relation to similar types and ecozones in southern Africa.