Vegetation survey of the Khomas Hochland in central-western Namibia: syntaxonomical descriptions

, the Khomas Hochland proper, riverine habitats as well as surrounding lowlands. The classification was further refined using Cocktail procedures to produce 30 associations, one with four sub-associations. These are described in this paper. Conclusion: A classification of synoptic data grouped the associations into five orders and one undefined cluster of associations on specialised desert habitats. Four of these orders correspond to the habitat types identified in the first classification. The fifth order, the Senegalio hereroensis–Tarchonanthoetalia camphor­ athi , represents high mountains of the central Khomas Hochland, which link bio-geographically to the grassland biome in South Africa.


Introduction
The Great Escarpment is a ± 5000 km long geomorphological feature along the rim of the southern African subcontinent. It is regarded as a zone of high biological diversity, containing numerous Centres of Endemism (Clark et al. 2011). Whereas most of the Great Escarpment forms a narrow divide between the coastal lowlands and the inland plateaux, in central Namibia the Damara Orogen created a mountainous landscape nearly 200 km wide from west to east. This mountainous landscape is commonly referred to as the Khomas Hochland (or Khomas highlands) (Schneider 2004;Swart & Marais 2009;Goudie & Viles 2014). Giess (1998) broadly describes the vegetation as 'highland savanna', without providing any details on composition or diversity.
Although the Khomas Hochland has been identified as an area of high botanical diversity (Hofmeyr 2004;Craven & Vorster 2006), only a few localised descriptions of the vegetation of this landscape exists. Of note here are a study by Volk and Leippert (1971) on a few farms southeast of Windhoek, using data from the 1950s and 1960s; an unpublished study by Kellner (1986) focusing on the Daan Viljoen Game Reserve west of Windhoek as well as portions of two farms southwest and east of Windhoek; a preliminary description of the vegetation of the Auas Mountain Range south of Windhoek (Burke & Wittneben 2007); as well as an account of the vegetation of the Auas-Oanob Conservancy southwest of Windhoek (Strohbach 2017). None of these studies provide a comprehensive overview of the entire landscape. In contrast, a fairly comprehensive description of the adjacent central Namib desert is available (Jürgens et al. 2013).
The Vegetation Survey of Namibia project has been initiated to fill this, and similar data gaps, at a national level (Burke & Strohbach 2000;Strohbach 2001). The project is aimed at providing data about the resource 'natural vegetation' to allow for sustainable planning and management of this renewable resource (Strohbach 2018). Due to the sheer size of the country, however, a strong emphasis is placed on utilising existing reliable data sources in addition to collecting, over a number of seasons, additional, gap-filling data. At the same time, due to the lack of necessary data density for a detailed study, the result will be at a regional overview level (often referred to as 'reconnaissance level') (Küchler & Zonneveld 1988;Strohbach 2001). This paper aims to contribute a first formal classification and description at this level of the vegetation found in the Khomas Hochland in central-western Namibia.

Study area
The study area is a block of roughly 31 000 km 2 in central western Namibia, between the coastal Namib desert lowlands and the inland plateaux ( Figure 1). It stretches from the border of the Namib-Naukluft Park in the west to about 17° 30'E, and from the B6 trunk road between Okahandja and Karibib in the north to about the Gaub Valley in the south. Administratively, it covers the western half of the Khomas Region, but also reaches into the Erongo and Otjozondjupa regions of Namibia.
The Khomas Hochland had its origin some 900 Ma ago as a sea between two tectonic plates, the Kalahari-and Congo cratons. Sediments deposited in this sea solidified and were metamorphised during the formation of the Gondwana supercontinent ± 650 to 450 Ma ago. At that stage, uplift also happened, resulting in the Damara Orogen (Schneider 2004;Swart & Marais  The escarpment of the Khomas Hochland raises from the Namib desert plains at about 900 m above sea level (asl) to well over 1 400 m asl. The central Khomas Hochland forms a deeply dissected, steep mountainous highland, raising to over 2 000 m asl in places ( Figure  2B) (Swart & Marais 2009). This is only topped by the mountain ranges Auas, Lichtenstein, Hakos and Gamsberg, which reach altitudes of over 2 400 m asl (Schalk 1983;Swart & Marais 2009). The study area is drained through a dense system of ephemeral rivers, forming tributaries to the Swakop and Kuiseb rivers to the west, Oanaob and Skaap rivers to the southeast, and White Nossob as well as Olifants rivers to the east (Figure 2b) (Strohbach 2008).
The climate can be described as a hot desert in the western half to a hot steppe in the eastern half, following Köppen (1936). The mean annual precipitation (MAP) ranges from ± 50 mm in the west to about 350 mm in the east (Figures 1 and 3), with a high degree of variation between seasons. Along the western edge of the study area, the coefficient of variation (CV) of MAP is as high as 90%, in the east around Windhoek the CV of MAP is about 40% (Mendelsohn et al. 2002). The orographic effect of the escarpment and high mountains in the study area are not known. The desert margins below the escarpment become relatively hot, with maximum temperatures measured well over 40°C. Frost is also rare or absent here. In contrast, the central highlands are cooler, with maximum temperatures around 36°C. Frost regularly occurs in the highlands in the winter months between May and August, at places as late as October (Figure 3).

Data sources
A data set of 1 151 relevés with 914 species was selected from the phytosociological database of Namibia (GVID ID AF-NA-001) (Strohbach & Kangombe 2012). Details of the selected data subsets are listed in Table 1.
The quality of the rainy season, which has an influence on the growth of the vegetation, was derived according to the criteria of Botha (1998). Normal years had an annual precipitation of between the 40th and 70th percentile of long-term precipitation records, whilst extreme years had below the 10th (extreme dry) or above the 90th percentile (extreme wet) annual precipitation.
Data collected by Kellner (1986) was captured from tables in his thesis as the original relevé data (field sheets) were no longer available. These relevés were also collected from 25 × 25 m (i.e. 625 m 2 ) plots. No accurate position data, nor habitat data, are available for these relevés. Data collected by Burke, Wittneben and Mannheimer (Auas Mountains and Windhoek Townlands) were in the form of species lists on specific sites rather than regular survey plots. Position data are available in most cases, but the habitat data are incomplete. These data were however included, as the sites were limited to specific habitats and limited in sizes (no longer than 100 m, no wider than 10 m), i.e. comparable to regular survey sites used for the Vegetation Survey of Namibia project. Especially the data by Burke and Wittneben  All other surveys followed the guidelines of the Vegetation Survey of Namibia project, i.e. were surveyed on a 20 × 50 m (1 000 m 2 ) plot, whilst plot layout was restricted to a specific habitat. In cases where the nature of the habitat did not allow a 20 m × 50 m sized plot (e.g. riverine habitats, rock outcrops), the plot shape was adapted to fit the habitat, without reducing the size, nor moving into a different habitat. The size of 20 × 50 m was chosen as suitable for an arid savanna and conforms to size criteria proposed by Brown et al. (2013). For these relevés, the position as determined by GPS, as well as habitat descriptors related to landscape, topography, lithology and stone cover have been noted.
Unknown species were collected for identification in the National Herbarium of Namibia (WIND). Species' nomenclature follows Klaassen and Kwembeya (2013), with the exception of the genus Acacia s.l., for which Kyalangalilwa et al. (2013) was followed.

Classification procedures
An initial classification was done using modified TWIN-SPAN (Roleček et al. 2009), with average Sørensen as distance measure, and utilising pseudospecies cut levels at 0 and 5% cover to differentiate between low-cover desert margin vegetation and inland vegetation at higher cover values. This classification was based on synusial data (i.e. trees, shrubs, dwarf shrubs and grasses  (Gillet & Julve 2018), and resulted in three clusters. These clusters were interpreted as representing the Pre-Namib and Escarpment zone, the central Khomas Hochland as well as a third group comprising riverine habitats and lowlands surrounding the Khomas Hochland. This classification result was used to split the data set into three subsets for further analysis.
The three clusters were further classified using the modified TWINSPAN classification algorithm, always using average Sørensen as distance measure, and utilising pseudospecies. For the Pre-Namib/Escarpment zone cluster, these were kept at 0 and 5%; for the other clusters cut levels were set at 0 and 10%. This difference in pseudospecies cut levels was necessary because the Pre-Namib has an inherent lower vegetation cover than the inland vegetation. In the case of Cluster 3 (riverine and lowland habitats), the data set was again split into two subsets after an initial classification using synusial data. The level of splitting was determined using peaks in crispness values (Botta-Dukát et al. 2005).
The classification of the clusters using synusial data was transferred to the full data set. Analysis of the resulting clusters revealed partial mixing of relevés between associations, or in several cases, known units to be included in other associations. This prompted a refinement of the classification results using Cocktail procedures (Bruelheide & Flintrop 1994;Bruelheide 1997), based on existing descriptions (e.g. Kellner 1986;Strohbach 2017), or on field observations. A detailed account of these Cocktail refinements is presented in the Results section of this paper.
Once an ecologically interpretable result was achieved, phytosociological tables were compiled and the synopsis for various associations extracted. Diagnostic species were determined using the phi coefficient of association (Chytrý et al. 2002). For this calculation the numbers of relevés 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 with at least a 60% frequency were regarded as constant and with at least an 80% frequency as highly constant.

Further descriptors of the associations
The average structure for each grouping (i.e. average tree, shrub, dwarf shrub, perennial grass, annual grass and herb cover) was calculated using the available growth form data. Descriptions follow Edwards (1983). For the species density (number of species per 1 000 m 2 ), the relevé data from Kellner (1986) were excluded, as these were sampled on 625 m 2 plots (25 × 25 m), not 1 000 m 2 plots as for all other relevés. In addition, an estimate of potential species richness for the association has been calculated with a first order Jackknife as proposed by Heltshe and Forrester (1983) as well as Palmer (1990).

Higher syntaxonomy
Due to the extensive refining of clusters with Cocktail, partially resulting in splitting of clusters or the definition of new clusters, the initial classification dendrograms could not be used as an indication of higher-order syntaxonomy as is customary (e.g. Luther-Mosebach et al. 2012). Instead, synoptic tables of the associations were prepared using percentage frequency for all four classifications. The synopsis of each association was taken as a pseudo-relevé, and these were combined into a single data set for classification with modified TWINSPAN (Roleček et al. 2009). No pseudospecies were used for this classification. This approach follows broadly the approached used by Winterbach et al. (2000).

First classification results and Cocktail refinements
The first classification resulted in three major vegetation zones, namely the Pre-Namib and Escarpment zone, the Khomas Hochland proper and the riverine habitats and lowlands vegetation surrounding the Khomas Hochland. The latter was further subdivided into two subsets for further classification. Cocktail refinements were necessary on all four clusters. Details of these refinements are provided in Table 2.

Pre-Namib and Escarpment zone
The vegetation of the Pre-Namib (Vornamib sensu Giess (1962Giess ( , 1998 and Escarpment zone is dominated by the plant families Poaceae, Fabaceae, Bignoniaceae and Burseraceae, in descending order of importance. The classification resulted in seven associations, which are formally described according to the International Code of Phytosociological Nomenclature (Weber et al. 2000 Within the synopsis, basic statistics on the association (or other syntaxa) are given, as derived from the analysis of clusters in Juice. Highly diagnostic species (with phi coefficient >60) and highly constant species (occurring in more than 80% of relevés) are indicated in bold. All structural descriptions follow Edwards (1983  Geographical location and habitat Several relevés grouped into Cluster 8 (big inland rivers) were found to belong to the Omeya plains according to their location. These were manually moved to the data subset of Cluster 4, Association 4.7

Cluster 4: Lowlands surrounding the Khomas Hochland:
This data cluster consisted of 462 relevés and 688 species. The synusial data set contained only 227 species. Classification of this data set resulted in ten clusters. On closer inspection, these were found to be fairly mixed due to the synusial data reduction, requiring extensive refinement through Cocktail procedures as follows: The recombined Clusters 7 and 10, as well as the original Clusters 8 and 9 were recognised as sub-associations of association 4.8. Soil and Terrain Database (SOTER) of FAO (FAO 1993). Specifically, the following slope classes apply: • Flat: 0-1° / 0-2% • Gently undulating: 1-3° / 2-5% • Undulating: 3-6° / 5-10% • Rolling: 6-9° / 10-15% • Moderately steep: 9-17° / 15-30% This format is also followed for all further descriptions in this paper.  (Ward 1987) along the upper reaches of the Kuiseb and Gaub canyons, with altitudinal ranges between 800 and 1 100 m asl. The substrate consists of carbonate-cemented conglomerates, which are between 40 and 60 m deep. The erosion of these conglomerates results in a steeply dissected hill landscape. Low rainfall (between 100 and 150 mm Mean Annual Precipitation -MAP), high runoff and a compact substrate allowing little infiltration result in this depauperated form (diversity-and cover-wise) of the Enneapogo no desvauxii-Eragrostietum nindensis. Because of its unique habitat, it is recognised as an association.  Figures 4B and 5B). This association occurs on the expansive calcrete gravel plains of the Pre-Namib and has been described by Jürgens et al. (2013) as 'eastern calcrete plains grasslands'. The altitude ranges between 780 and 1 300 m asl, and the landscape is generally flat. MAP ranges between 50 and 150 mm, but because of the reduced run-off and looser substrate, this association is far more species rich compared to the Enneap ogono desvauxii-Adenoloboetum pechuelii (depending on the quality of the rainfall season). These low open bushlands are characterised by a variety of stem succulents typical of the escarpment zone ( Figures 4F and 5F). They occur on the inselberg ranges along the Pre-Namib fringes as well as the lower escarpment zone, at an altitudinal range between 1 050 and 1 500 m asl. The topography consists of steep mountain slopes generally with a gradient of well above 30%, as well as with considerable stone cover (up to 80%), mostly small, medium and large stones. MAP ranges between 100 and 200 mm.  5G). They occur on rolling to moderately steep (10-30%) plains and footslopes of the escarpment, often occurring far inland in valleys of the escarpment. Stone cover is not as high, roughly 40%, again mostly small, medium and large stones. The altitude ranges between 950 and 1 400 m asl, whilst MAP ranges between 100 and 250 mm.

Khomas Hochland proper
The vegetation of the Khomas Hochland is dominated by the plant families Poaceae, Asteraceae, Fabaceae and Scrophulariaceae. The classification resulted in eight associations, which are formally described according to the International Code of Phytosociological Nomenclature (Weber et al. 2000) below (with one exception). The synoptic table for these is presented in online Appendix 1, and the phytosociological  Figures 6A and 7A). The occurrence of Euryops walterorum is restricted to the Gamsberg Plateau (i.e. a limited-range endemic) (Nordenstam 1966;Loots 2005), which is 210 ha in size. The plateau consists of a fine-grained quartzite layer ± 30 m thick, at an altitude 2 347 m asl (Wittig 1976;Schalk 1983). The topography is near flat but displays a conspicuous cover by large stones of roughly 40%. The plateau receives an estimated 150 mm MAP, based on general rainfall maps (Mendelsohn et al. 2002). This, however, does not take any possible orographic effects into account.  Shrubs also occur only on the lower reaches, whilst the mountain tops are distinctly grass covered ( Figures  6D and 7C). It occurs on the northern face of the Auas Mountain range at an altitudinal range of 2 200 m asl and above (Burke & Wittneben 2007;Strohbach 2017 Figures 6E and 7D). This association occurs on the Auas Mountain and Lichtenstein ranges at mid-altitude, between 2 000 and 2 300 m asl. No evidence could be found for a differentiation between north-and south-facing sides of the mountains, unlike reported by Burke and Wittneben (2007) or Strohbach (2017). These slopes are steep (30-60%) with considerable quartzite stone cover, especially medium to large stones, even rocks (± 80% stone and rock cover), of the Auas formation.

Pennisetum foermerianum-Ficus ilicina association
2.6 Triraphio ramosissimae-Manuleopsietum dinteri ass. nov.  Figures 6F and 7E). This association occurs on smaller mica-schist rock outcrops common within the central Khomas Hochland, as part of the Kuiseb formation. These are normally between 10 × 10 to over 50 × 50 m in surface area, but always broken, never solid rock faces. Due to the nature of the rock beds of the Khomas Hochland, many of these rock outcrops face south (but not exclusively). Altitude ranges between 1 600 and 1 800 m asl, whilst MAP ranges between 250 and 400 mm. Kellner (1986) distinguishes two forms of this association, being a Pennisetum foermerianum-Dombeya rotundifolia community on broken rock outcrops, as well as a Triraphis ramosissima-Combretum apicula tum community on more solid, platy rock outcrops, generally also with more gentle slopes. This subdivision was neither confirmed in this study, nor by Strohbach (2017). Volk and Leippert (1971) also describe a Com bretum apiculatum-Eragrostis scopelophila association, unfortunately without proper synopsis.  (Geological Survey 1980, Schneider 2004. The stone cover is dominated by pebbles and medium-sized quartz stones up to 40%, with only occasional sub-outcropping mica schists. This association is the most extensive association within the central Khomas Hochland and occurs at an altitude of between 1 600 and 2 000 m asl. MAP ranges between 200 and nearly 400 mm. As the rivers widen, a distinct near-flat, coarse sandy bed is formed, on which this sparse short grassland establishes ( Figures 8B and 9B). Depending on the flow regime (fast-or slow flowing, or even standing or seepage areas), a variety of species establish, often of ephemeral nature. Generally, Chloris virgata, Sporobolus fimbria tus, Eragrostis rotifer, Eragrostis omahekensis, Eragrostis echinochloidea and Cynodon dactylon dominate this association. This association occurs widespread in the study area as important part of the drainage system.

Stipagrostioetum namaquensis ass. nov.
Larger rivers in relatively low-gradient environments (e.g. the Oanob) form deep sandbanks adjacent to the main flow channels, on which this tall, moderately closed grassland occurs ( Figures 8C and 9C). Unlike the Cynodo dactylonis-Eragrostioetum rotiferi, this association is of more permanent nature, with Stipagrostis namaquensis stabilising the sand bank, and allowing sedimentation on this sand bank. This eco-engineer also forms a suitable habitat for phanerophytic species (commonly Vachellia karroo, but also Vachellia eriolo ba, Faidherbia albida and Euclea pseudebenus) to establish. The sandbanks found in the Oanob River form the western-most occurrence of this association, with extensive sandbanks found further east in the Seeis and Nossob river systems. These sand banks occur within the study area at an altitude of between 1 850 and 1 950 m asl, and a MAP of between 300 and 350 mm.

Themedio triandrae-Chloroetum virgatae ass. nov.
In the low-gradient landscape of the Oanob Plateau, but also other low-gradient areas like the Regenstein Valley and at Neudamm, valleys have been filled with sediments, forming a low-gradient, overgrown emphemeral wetland locally referred to as an omuramba (plural -omirimbi) (King 1963, Strohbach 2008). This forms the habitat for the Themedio triandrae-Chloro etum virgatae, a short, moderately-closed grassland ( Figures 8D and 9D), dominated by ephemeral species like Geigeria pectidea, Tribulus terrestris, Chloris virgata, Aristida hordeacea and Aristida adscensionis. Only Aris tida congesta subsp. congesta is a prominent perennial species, whilst the name-giving Themeda triandra is a rare species, often only occurring as remnant of the original lush grassland described by Volk and Leippert (1971). These authors already noted the generally poor state of this vegetation, which is prone to degradation and erosion. Once erosion sets in, the sediments are soon washed away to form a low-productive riverbed similar to the Cynodo dactylonis-Eragrostioetum rotiferi, with shrubs and trees establishing on the sides to form the start of the Chloro virgatae-Vachellietum karroo.
The Themedio triandrae-Chloroetum virgatae occurs at an altitudinal range of between 1 740 and 2 200 m asl, and at a MAP of between 300 and 400 mm.  (1986) and Strohbach (2017) refer to the Chloro vir gatae-Vachellietum karroo as Cynodon dactylon-Acacia karroo association, a name which was rejected as the name-giving Cynodon dactylon was neither a diagnostic nor constant species in this association.

Chloro virgatae-
The Chloro virgatae-Vachellietum karroo generally have a steep bank towards the riverbed but are fairly flat beyond the bank. Stone cover is low, rather occasional, mostly deposited by extreme flood events. This association occurs at altitudes of between 1 540 and 1 950 m asl, and MAP of between 250 and 400 mm.  Figures 8F and 9F). This association form the riparian forests along the bigger rivers in the western (Pre-Namib) part of the study area, i.e. the middle reaches of the Swakop, Kuiseb and Gaub rivers. The relationship to the lower Kuiseb vegetation (Theron et al. 1980) needs to be investigated further.
The Salvadoro persicae-Eucleetum pseudebeno occurs in the lower rainfall areas, with a MAP of less than 200 mm, and at altitudes less than 1 100 m asl. Similar to its inland equivalents (the Chloro virgatae-Vach ellietum karroo and the Setario finitae-Vachellietum eriolobae), this association displays low slope gradients (flat to gently undulating) with little stone cover.
This tall, moderately closed woodland is found along the banks of the big inland rivers of the Khomas Hochland, notably the upper reaches of the Swakop River, but also its tributaries like the Gammams and Otjiseva rivers ( Figures 8G and 9G). It is dominated by Ziziphus mucronata, Vachellia karroo and Vachellia erioloba trees, with Faidherbia albida also occurring occasionally.
The Setario finitae-Vachellietum eriolobae often forms islands within broad riverbeds, indicating a relationship,

Lowlands surrounding the Khomas Hochland
The vegetation of the surrounding lowlands is dominated by the plant families Poaceae, Fabaceae, Asteraceae and Amaranthaceae. Eight associations, one with four sub-associations, were identified and are formally described following guidelines of the ICPN below (with one exception). Their composition is presented in Appendices 1 and 2 as synoptic and phytosociological tables.  Figures 10A and 11A). This association is limited to the Narais plains towards Rehoboth, occurring on shallow loamy soils on calcretes, which have formed over the quarzites of the Duruchaus formation. Stone cover is about 20%, mostly calcrete gravel and small stones. Occasional deep pockets in the subsurfacing calcretes allow Vachellia erioloba trees to establish, adding a very sparse tree layer to this association's structure (Jürgens et al. 2010). This association receives between 250 and 300 mm MAP and is at an altitudinal range of between 1 580 and 1 640 m asl. The topography is generally flat.  Figures 10B and 11B). This association is however prone to encroachment by specifically Senegalia mel lifera subsp. detinens, Catophractes alexandri and/or Vachellia reficiens, which causes the structure to change to a tall, denser shrubland. It occurs specifically on the rolling to moderately steep Oanob plateau, whilst similar vegetation has been also observed on the Hoffnung plateau east of Windhoek as well as at Neu Heusis west of Windhoek. As with the surrounding Brachiario ni gropedatae-Senegalietum hereroensis, the subsurface geology are mica schists of the Kuiseb formation. However, the landscape is a gentler, undulating to rolling plateau at an altitudinal range of between 1 800 and 1 900 m asl. Stone cover is about 40%, mainly by small and medium-sized stones. MAP is between 300 and 350 mm.  Figures 10D and 11D). The presence of Rhigozum trichotomum and Panicum arbusculum within this southern Khomas Hochland is indicative of the transitional nature of these shrublands to the Dwarf Shrub Savanna sensu Giess (1998) or Nama Karoo. This association occurs on the rolling to moderately steep hilly landscape of the southern Khomas Hochland, and is especially well-represented along the C26. The underlying geology is formed by the mixtites of the Chuos Formation. Stone cover is about 40%, mainly small and medium-sized stones. The altitude ranges between 1 650 and 1 850 m asl, whilst the MAP ranges between 200 and 300 mm.

Elephantorrhizo suffruticosae-Euphorbietum guerichianae ass. nov.
These short, moderately closed bushlands occur in the upper reaches of the escarpment zone, between 1 450 and 1 780 m asl. The slopes are generally very steep ( Figures 10E and 11E). This association is dominated by Vachellia reficiens, Elephantorrhiza suffruticosa, Senegalia erubescens, as well as the grasses Stipagrostis uniplumis var. uniplumis, Eragrostis nindensis and Setaria appen diculata. The latter again is an indication of a transition to mountainous habitats in the Nama Karoo biome. Geology is varied, however mostly mica schists of the Kuiseb Formation and quarzites of the Auas formation (Geological Survey 1980; South African Committee for Stratigraphy 1980). Especially rock cover is high, with up to 80% of the surface covered by large stones and/ or rocks. MAP ranges between 200 and 300 mm, not taking any orographic effects into account.

4.6
Dichrostachyo cinereae-Senegalietum erubescentis ass. nov.   Figures 10G and 11G). This association occurs on the alluvial plains around Aris and at Omeya south of Windhoek but has also been observed in the Brakwater/Döbra area in the valley north of Windhoek. The slopes are flat to gently undulating, with hardly any stone cover (mostly gravel and small stones, less than 2%). The altitude ranges between 1 600 and 1 770 m asl, with MAP of between 300 and 350 mm.  Figure  10H). This association is often associated with degradation within various units in the Khomas Hochland, but also the regular Thornbush savanna elements to the east and north of the Khomas Hochland. Although four distinct forms have been recognised (and described as sub-associations below), there is still sufficient uncertainty regarding the classification not to allow a formal description of these communities yet.

Stipagrostis uniplumis-Senegalia detinens typical sub-association
These tall, semi-open shrublands are the typical form of the Stipagrostis uniplumis-Senegalia detinens association ( Figures 10I and 11H). Strohbach (2017) referred to this as the 'Pupalia lappacea-Acacia mellifera bush encroached lowlands', indicating that this association is often a degradation state of other associations. As Pupa lia lappacea however was not found to be a diagnostic nor constant species, the proposed name was rejected. This sub-association is found as patches in various components and associations within the Khomas Hochland. Especially the Brachiario nigropedatae-Senegalietum hereroensis, the Panico arbusculi-Senegalietum det inentis and the Panico lanipedis-Pentzietum incanae seem prone to degradation to this state.
The Stipagrostis uniplumis-Senegalia detinens typical sub-association occurs on undulating to rolling slopes, often with up to 40% gravel and small stone cover, at altitudinal ranges between 1 650 and 1 800 m asl.

Stipagrostis uniplumis-Senegalia detinens tarchonanthus camphoratus sub-association
These high, semi-open shrublands are also dominated by Senegalia mellifera subsp. detinens and Catophractes alexandri ( Figures 10J and 11I). The occurrence of Tar chonanthus camphoratus indicates slightly moister conditions, reflected in deeper, sandier soils, but also higher rainfall regimes (between 350 and 400 mm MAP). This sub-association is found in the eastern parts of the study area, linking into the thornbush savanna types to the east of Windhoek. The altitude ranges between 1 550 and 1 830 m asl, with rolling to moderately steep slopes. The subsurface geology are granites and gneiss from the Hohewarte complex and Gamsberg Suite. Stone cover is low (< 15%), mostly gravel and small stones.  Figure 10K). It has a limited occurrence in patches in the Khomas Hochland and has mainly been observed west of Windhoek in the Daan Viljoen Game Reserve. This sub-association occurs on moderately steep to steep slopes at altitudinal ranges between 1 500 and 1 800 m asl. Stone cover is fairly high, above 40%, mostly small and medium stones. Larger stones, and occasional rock outcrops, are also present. The subsurface geology is also mica-schists of the Kuiseb Formation.

Higher syntaxonomy
The classification of the synoptic relevés of 30 associations yielded a classification with crispness peaks at 4, 8, and 14/15 divisions. Accordingly, a classification with 15 clusters was accepted and interpreted. The 15 clusters were interpreted as alliances, whilst these alliances were grouped into orders between level 4 and level 8 divisions ( Figure 12). An overview of associations, associated with their relevant alliances, orders and broad habitats are given in Table 3.
The Senegalio hereroensis-Tarchonanthoetalia cam phorati is typified by the Senegalio hereroensis-Tar chonanthion camphorati. This order represents the high mountains of central Namibia. Based on the present data, it can be subdivided into two alliances, being the Senegalio hereroensis-Tarchonanthion camphorati, representing the Auas Mountains, as well as the Digitario erianthae-Euryion walterorum, representing the Gamsberg Mountain. The Senegalio hereroensis-Tarchon anthion camphorati is typified by the Senegalio her eroensis-Tarchonanthoetum camphorati and contains also the two other associations occurring on the Auas Mountains. The Digitario erianthae-Euryion waltero rum is typified by the Digitario erianthae-Euryopietum walterorum and contains only this association.
The Cynodo dactylonis-Eragrostioetalia rotiferi is recognised as a provisional order representing the riverine habitats. According to the present classification, it breaks up into five alliances, each representing a single association. Due to this, and as other, similar habitats exist in central and southern Namibia, this provisional order is not yet formally described pending a thorough revision.
The Brachiario nigropedatae-Senegalietalia hereroensis is typified by the Brachiario nigropedatae-Senegalion her eroensis, which is also the only alliance to this order. This order represents the vegetation of the central Khomas Hochland. The Brachiario nigropedatae-Senegalion her eroensis in turn is typified by the Brachiario nigropeda tae-Senegalietum hereroensis. In addition to the typical association, three other associations have been grouped into this alliance. Two of these represent rocky habitats typical for the central Khomas Hochland, the third represents smaller rivers within the Khomas Hochland.
The sandy and/or specialised habitats in desert environment form a cluster of associations, each being represented by an own alliance, and possibly (depending on interpretation), even own orders. One of these associations, the Salvadoro persicae-Eucleetum pseudebeno, could possibly also be grouped with the Cynodo dac tylonis-Eragrostioetalia rotiferi due to its habitat. This is speculative, and subject to review in future. Because of this, no formal higher syntaxonomic groupings are described here.
The Eragrostio nindensis-Vachellietalia reficientis is typified by the Eragrostio nindensis-Vachellion reficientis. This order represents the Pre-Namib as well as the escarpment zone. The classification does not present any evidence of this order to be split into two alliances, yet, based on the habitat and vegetation of the associations in this order, a split was made between the name-giving alliance and the Enneapogono desvauxii-Eragrostion nindensis. The Eragrostio nindensis-Vachellion reficien tis is typified by the Eragrostio nindensis-Vachellietum reficientis and includes two other association within the escarpment zone. The Enneapogono desvauxii-Eragros tion nindensis is typified by the Enneapogono desvauxii-Eragrostietum nindensis and includes the closely related Enneapogono desvauxii-Adenoloboetum pechuelii, as part of the Pro-Namib landscape.
The Dichrostachyo cinereae-Senegalion erubescentis is typified by the Dichrostachyo cinereae-Senegalietum erubescentis and includes the as yet not formally described Stipagrostis uniplumis-Senegalia detinens association. This alliance represents the northern and eastern lowlands associated with the Khomas Hochland.

Discussion and conclusion
The higher syntaxonomy confirmed several preliminary observations. The initial classification into Pre-Namib and Escarpment, Khomas Hochland proper, Riverine habitats as well as surrounding lowlands are largely reflected in the orders Cynodo dactylonis-Eragrostioetalia rotiferi, the Brachiario nigropedatae-Senegalietalia her eroensis, the Eragrostio nindensis-Vachellietalia reficientis and the Panico arbusculi-Senegalietalia detinentis. This is also in line with the groupings proposed by Giess (1998), i.e. the Central Namib, the Desert transition and escarpment Zone as well as the Highland savanna.
A notable exception is the Senegalio hereroensis-Tar chonanthoetalia camphorati (high mountains), which is highlighted as a grouping on its own. This also confirms the findings of Strohbach (2017), who has highlighted the high Auas Mountain vegetation as a separate grouping. The Senegalio hereroensis-Tarchonanthoetalia cam phorati has close affinities to the grassland biome of South Africa, with its dominant cover of grasses, in particular Digitaria eriantha (Mucina & Rutherford 2006). Prominent representatives of the Asteraceae in this order, amongst others Eriocephalus spp., Stoebe plumosa and Euryops walterorum, also have close relatives in the grasslands and adjacent Karoo vegetation in South Africa (cf. Nordenstam 1966;Müller et al. 2001;Njenga 2005), indicating a biogeographic relationship to these biomes, rather than the surrounding savanna biome.
The other notable exception is the formation of a cluster of specialised habitats within the desert biome environment. Although the Salvadoro persicae-Eucleetum pseudebeno has distinct affinities to the inland rivers, it still groups into the desert environment rather than the inland savanna environment. Of the other two associations in this cluster, the Crotalario podocarpae-Stipa grostioetum obtusae has affinities to the arid grasslands of the Nama Karoo biome (Mucina & Rutherford 2006), whilst the Tribulocarpo dimorphantho-Vachellietum eri olobae has affinities to the Kalahari duneveld sensu Mucina and Rutherford (2006) or southern Kalahari sensu Giess (1998).
The vegetation classification and description presented in this paper represents a broad overview, as dictated by the 'reconnaissance level' scale. Intensified sampling in several specialist habitats, e.g. the riverine habitats, Northern and eastern fringe lowlands Table 3. Overview of the higher syntaxonomy of associations described in this study (continued) the high mountains (in particular the Hakos mountains, as well as ranges like the Bismark Mountains east of the Auas and various ranges of the Kamtsas and Sinclair formations south of Windhoek) and the Matchless Member dissecting the central Khomas Hochland, will likely yield further associations. The higher syntaxonomy will likely also be clarified and strengthened once more such studies are available and included in the analysis.