The Omaheke [‘great sand desert’] forms the western fringe of the Kalahari basin (Mendelsohn & el Obeid 2002, 2003; Mendelsohn et al. 2002) (Figure 1). The central, major part of the landscape is formed by an extensive sand plateau, sloping from the southern, western and north-western fringe towards the central east. The topography of the Kalahari sand plains is flat to almost flat. These Kalahari sands, with weak to little development in the profiles, are classified as haplic or ferralic Arenosols.1 These soils are generally deep to very deep and well drained. They have a low agricultural potential as they are leached, have a low water-holding capacity and a low nutrient status (Coetzee 2003; Mendelsohn et al. 2002; Strohbach et al. 2004).

Within this sandy plain, occasional dune fields occur (De Pauw et al. 1998; Thomas 1984). The sand dunes reach a height of up to 20 m, forming an undulating to rolling landscape. The dune crest, dune slopes and foot of the dunes consist of ferralic Arenosols with haplic Arenosols mostly on the lower footslopes of the dunes. Soils on the interdunal depressions and isolated small pans of the Kalahari vary in depth from moderately deep to very deep sandy soils. The two major soil types that occur within this soil-mapping unit are petric Calcisols (Table 1) and haplic Arenosols. Isolated pans have shallow haplic Calcisols with calcrete horizons surfacing and some gleying was occasionally found (Strohbach et al. 2004) (Figure 2).

The sand plateau is drained by a number of deeply incised drainage channels, commonly referred to as omiramba (singular: omuramba) (King 1963; Strohbach 2008). These fossil drainage lines were formed during the wetter phase predominating the early to mid-Pleistocene age. The depths and widths of the omiramba depend on the prevailing winds and water flows experienced throughout the ages. Very often, calcrete outcrops are exposed at the edges of the major omiramba, especially the Otjozondjou, Epukiro, Eiseb and Rietfontein omiramba. The major soil types of the omiramba are arenic or calcic Fluvisols, occasionally associated with petric or haplic Calcisols (Strohbach et al. 2004) (Table 1) (Figure 2).

Along the south-western fringes of the Kalahari sand basin, but also south and east of Gam, the sand cover is shallow, often overlaying calcretes, with occasional calcrete pans and reworked whitish, loose, unstructured Kalahari sands. The topography is flat to almost flat. Typical here are haplic Calcisols, whilst in the far north-east some eutric Cambisols (Table 1) have also been found (Strohbach et al. 2004) (Figure 2).

Floodplains occur between the Klein Omatako Omuramba and the Omuramba Omatako and stretch up to Okanguindi village, south-east of Okakarara. They are subjected to repeated seasonal flooding as a result of the relatively low relief. The continuous erosion of the surface horizon by water, results in the formation of many pans and watercourses throughout the area. Owing to continuous reshaping and removal of the parent material and topsoil, the soils of this mapping unit are much younger in terms of development and age. The dominant soils of this mapping unit, arenic Fluvisols (Table 1), are weakly developed and poorly structured (Figure 2). Soils within the lower positions are subjected to flooding and show signs of mottling caused by poor drainage and water-logging.

In the far north-west of the study area (north-west of Otjituuo towards Grootfontein), but also north of Gam, is a plateau with limestone related to the Karstveld. Hard and soft calcrete and/or limestone outcrops are exposed at the soil surface. The dominant soils are lithic and molic Leptosols, with some skeletic Leptosols being associated (Strohbach et al. 2004) (Figure 2).

The climate is relatively uniform, with a rainfall gradient ranging, on average, from just under 400 mm in the south to just under 500 mm in the north, annually (Figure 1). The variation in rainfall (CV) is about 60% for the entire study area (Botha 1996; Mendelsohn et al. 2002; Namibian Agricultural Resources Information System [NARIS] 2001). This typical summer rainfall area has a dry winter and a dry and hot spring season (Figure 3). Frost is not unheard of in winter, with between 5 days of frost in the north and 30 days or more in the south (Mendelsohn et al. 2002; Southern African Science Service Centre for Climate Change and Adaptive Land Use [SASSCAL] WeatherNet 2012) (Figure 3a–3c). The average relative humidity drops to below 10% during the dry months, but rises to 70% – 80% during the rainfall months (Mendelsohn et al. 2002).

Regular veld fires are known to occur within the study area (Le Roux 2011; Mendelsohn & el Obeid 2002). The fire return period ranges between less than 2 years and 10 years and longer, with especially the north-eastern part of the study area being burned very frequently. As a combination of a variety of factors, including the fire return period, the seasonality of fires, et cetera, Le Roux (2011) determined a ‘fire regime’ for various areas in Namibia. Most of the study area is classed as having a ‘mild’ to ‘moderate’ fire regime.

At the same time as the vegetation survey in April 2004, a survey of the major landscape types and associated soil types was conducted. In addition to describing the landscapes based on field observations and topographic map interpretation, soil profiles were exposed by digging soil pits on representative sites throughout the study area (Strohbach et al. 2004). However, because of the fairly labour-intensive sampling procedure, only 71 sample pits were dug. Very few of these were at the same localities as the vegetation sample plots (Strohbach 2014). This means that for this particular study, whilst also relying on existing geographic information system (GIS) sources (Le Roux 2011; Mendelsohn & el Obeid 2002; Mendelsohn et al. 2002; NARIS 2001), only extrapolated environmental data were available. It thus made sense to use synoptic composition data of the associations and subassociations for all analyses, rather than the individual relevé data, for comparison to the available environmental data; see the summarised data used in Table 1. For all analyses, the Panicum gilvum–Marsilea vlei community was omitted as a potential outlier, also because only one relevé was collected in this community. Analytical soil data from the soil profile descriptions were augmented with standard soil parameters developed for the World Soil Reference Base (Batjes 2002).

Rainfall has, by far, the strongest and most consistent effect on the vegetation in arid to semi-arid climates. This also holds true for the vegetation within the present study area, which was proven to be the only significant driver in the change from microphyll to macrophyll vegetation (Table 3). Our present gradient is roughly 200 km long, with a rainfall gradient of less than 150 mm. Skarpe (1996) described a similar change in vegetation as it occurs across Botswana along a rainfall gradient over 1100 km on uniformly sandy soils; she unfortunately does not provide rainfall differences in her paper. From the geographical description, it is assumed to be a gradient from Bokspits in far south-west Botswana (177 mm mean annual precipitation – Bhattachan et al. 2012) to the four-country corner with Zimbabwe, Namibia and Zambia in north-east Botswana. For nearby Katima Mulilo in north-east Namibia, a long-term mean annual precipitation of 758 mm has been recorded (Namibia Meteorological Services 1997). From this, a rainfall gradient of approximately 580 mm is assumed for Skarpe's study. Her main finding is that broad-leafed trees and grasses are positively correlated with increasing precipitation, whilst fine-leafed phanerophytic species and chamaphytic species were positively correlated with increasing aridity along the gradient.

Whereas Skarpe (1996) describes a gradual change in vegetation over a rather lengthy gradient, the change from microphyll-dominant vegetation (associations A1 – A11) to macrophyll-dominant vegetation (associations A12 – A14) within the study area is rather abrupt over a very short geographical distance (cf. Hüttich et al. 2009; Strohbach 2014). These dramatic changes occur over virtually no rainfall gradient at all and are related to the change in soil type, from a fine-textured, loamy soil to the coarse-textured, deep sands of the Kalahari basin, as suggested by Leser (1972). Tinley (1982) describes similar dramatic changes resulting from changes in soil texture for Mozambique. These changes are caused by the difference in water availability as a result of differences in soil moisture relationships (in particular, a higher soil matric potential and higher porosity, but also higher adhesion, in loamy soils versus a high hydraulic conductivity in sandy soils), causing the inverse texture effect (Figure 4) (Austin et al. 2004; Noy-Meir 1973; Porporato et al. 2001). The gradual change between the various subassociations of the Combreto collini–Terminalietum sericeae (A12) and the Burekeo africanae–Pterocarpetum angolensis (A14) from south to north happens on similar sandy soils and can be compared to the changes in PFAs across a climatic gradient described by Skarpe (1996).

Owing to the higher hydraulic conductivity of sandy soils, rainfall infiltrates faster, but, at the same time, water percolates to deep soil layers. This means that water is available in the upper soil layers for a comparatively short period during the rainy season, whereas deeper soil layers (below 2 m) act as a reservoir for deeper-rooted plants (B. Strohbach, unpublished data). Therefore, deep-rootedness is a prerequisite for the survival of phanerophytes growing on such sandy soils. This holds true for most of these species (Hipondoka & Versfeld 2006; Hipondopka et al. 2003; Mwita et al. 2008; Savory 1963; see also Timberlake & Calvert 1993). Deep-rooted geoxylic suffrutices, for example Entada arenaria and Dichapetalum cymosum, better known from moister savannas and woodlands of the Sudano-Zambezian Ecoregions (500 mm – 700 mm mean annual rainfall) (White 1976), thus also succeed in growing in this environment. The latter species (gifblaar or magou) is problematic throughout the study area, where it can increase rapidly in density by taking advantage of disturbances caused by high numbers of cattle (Bester 1989).

Another adaptation to the temporary high availability of water in sandy soils during the rainy season (because of high infiltration rate) can be seen in stem succulents, specifically various Commiphora species (Figure 4b). These are able to photosynthesise, also during the dry season, by means of a photosynthetic layer under the translucent bark – an adaptation to xeric environments (referred to as ‘cortic photosynthesis’ by Gillison & Carpenter 1997). Whether pinnately compound leaves (‘filicoid leaves’) point to any specific adaptation, or are rather a coincidence related to the species composition, could not be established.

The grass sward of the sandy soils is dominated by the annual species Aristida stipitata and Megaloprotachne albescens. Most grass roots tend to be limited to the upper 40 cm to 60 cm of the soil profile; they are therefore unable to access water from deeper water resources (Hipondoka et al. 2003). With this annual character, the grass sward utilises a strategy of growing rapidly during the rainy season to exploit available resources to the fullest in the shortest possible time, but does not store water in underground organs or access the water that is available in deeper soil layers during adverse seasons. The few perennial grass tufts either have hairy leaves (e.g. Digitaria seriata), or are extremely fine-leaved (e.g. Stipagrostis uniplumis var. uniplumis) as an adaptation to this harsh environment. Digitaria seriata is also often found within bush clumps, making use of both shade and increased water infiltration as a result of stem flow under these canopies. Similar patterns were found in other arid environments (Martinez-Meza & Whitford 1996; Pressland 1973). Perennial grasses within the Kalahari are also known to utilise inverse hydraulic lift to channel water down into the soil profile (to a depth of between 50 cm and 100 cm). This is postulated as a potential mechanism (not only by grasses, but also by trees) to create an own ‘reservoir’ of soil water for use during the dry season (Schulze et al. 1998).

Most of the phanerophytes growing on sandy soils are deciduous, shedding their leaves during the dry winter and hot, dry spring months (Table 2, own observations). This is a typical adaptation to avoid water loss (Fensham & Fairfax 2009; Seghieri, Floret & Pontanier 1995). Many phanerophytes, especially Terminalia sericea and various Combretum spp., have hairy leaves, an adaptation typically associated with psammophil plants (Warming 1909).

Within this study, deciduous plants were not limited to the sandy soils. Many species growing on loamy soils exhibit the same adaptation; however, several species are distinctly evergreen (e.g. Boscia albitrunca) or at least semi-deciduous, that is, shedding their leaves just before the next leaf flush (e.g. Acacia erioloba, Acacia hebeclada subsp. hebeclada). These evergreen to semi-deciduous species are also deep-rooted (Canadell et al. 1996; Moustakas et al. 2006; Schulze et al. 1998), enabling them to obtain water from deep soil layers during the dry season. Overall, however, the tendency is for the leaves to become distinctly smaller on the more loamy soils, an indication that the soil moisture availability is far lower on loamy soils than on sandy soils.

An extreme case of edaphic aridity is presented by the shallow calcrete pans of the Eragrostio echinochloideae–Eriocephaloetum luederitziani (A4). In addition to the very shallow profile, the soil is characterised by a high skeletal content, a high pH (> 8) and a relatively high electric conductivity (> 100 µS/cm). Moreover, the low soil volume can only store limited quantities of water, resulting in this water being saturated with ions, which means that it is not readily available to plants. The nature of these soils is manifested in the vegetation structure; the association being dominated by dwarf shrubs and, furthermore, within this association, 14% of all species had sclerophyll leaves (compared to an average of 6% – 9% for all other associations) and a relatively high microphyll leaf component (89%) (Table 2).

Although fire is known to have a severe influence on the vegetation composition in the Kavango and Zambezi regions (Geldenhuys 1977; Le Roux 2011; Mendelsohn & el Obeid 2003; Mendelsohn & Roberts 1997; Strohbach & Petersen 2007), only structural alterations were evident during the present study (Figure 6). It is thus unlikely that fire is a driving factor in the discrimination of vegetation associations within the study area.

A major shortcoming of this study is the fact that only synoptical data (both for vegetation composition as well as for environmental variables) could be employed. Because of this, the clear soil gradients illustrated by the ordination techniques were not conclusive in the multiple linear regression. Future studies concerning such moisture relations should take this problem into consideration.