Bloemfontein extends from approximately 29°00’ to 29°15’ south and 26°07’ to 26°21’ east, with altitude ranging from 1350 m to 1450 m above sea level. According to the climate statistics from the South African Weather Service, the annual mean maximum and minimum temperatures are 24.6 °C and 7.6 °C, respectively. Rainfall mainly occurs in summer in the form of thunderstorms, and it averages 550 mm annually. The main geologic feature of the study area is the Karoo Supergroup, represented by the Tierberg Formation of the Ecca Group and the Adelaide Sub-Group of the Beaufort Group; there are also dolerite intrusions of the post-Karoo age (Johnson et al. 2006). Prominent soil groups are oxidic (Hutton form), plinthic (Bainsvlei form), duplex (Valsrivier, Swartland and Sterkspruit forms), cumulic (Oakleaf form), vertic (Arcadia form) and melanic (Milkwood form) (Fey 2010; Soil Classification Working Group 1991). Bloemfontein is situated in the Grassland Biome (Rutherford & Westfall 1994), and is part of the Central Variation of the Dry Cymbopogon-Themeda Veld (Acocks 1988). Other classifications describe Bloemfontein’s vegetation as Dry Sandy Highveld Grassland (Bredenkamp & Van Rooyen 1996) and Dry Highveld Grassland (Mucina et al. 2006).

The first step of the survey entailed the stratification of vegetation prior to sampling. Stratification of the area was done on 1:50 000 scale maps and 1:30 000 aerial photographs, based on the topography and relative homogeneity of physiognomic units. The topographic units recognised were the watercourses, flat plains as well as the hills and ridges. A total of 248 relevés were compiled; 160 were compiled for the first time, while 88 were from existing data (Muller 1970; Rossouw 1983). Sample plots ranging in size between 16 m² for the grassland vegetation and 100 m² for the woody vegetation were surveyed across the study area. All plant species present in each sample plot were recorded and each was given a cover-abundance value according to the Braun–Blanquet scale (Kent & Coker 1996; Mueller-Dombois & Ellenberg 1974). Plant taxonomy generally follows Germishuizen and Meyer (2003). For each relevé, habitat attributes were also noted, including rock type (geology), terrain type (topographical position) and an estimation of the percentage of rockiness of the soil surface. Soil characteristics such as soil depth, pH, organic matter and texture were used for the study. Other noted observations included the extent of soil erosion and forms of biotic influence such as utilisation by herbivores and management practices.

Phytosociological data were first captured and processed in the TURBOVEG database (Hennekens 1996a), and then exported to the MEGATAB computer program (Hennekens 1996b) for classification using TWINSPAN (Hill 1979a). The result was a synoptic table that shows a hierarchical classification of the syntaxa, with each synrelevé representing a plant community. The principle of synoptic tables is based on rating the presence of each species within a community on a constancy scale (Kent & Coker 1996; Mueller-Dombois & Ellenberg 1974). An ordination technique, Detrended Correspondence Analysis (DECORANA) (Hill 1979b), was applied to the data set to illustrate floristic relationships between the plant communities and to detect possible relationships between the communities and the environment. Canonical Correspondence Analysis (CANOCO) (Ter Braak & Šmilauer 2009), an extension of DECORANA, was also carried out to further illustrate the correlations between the vegetation data and the environmental variables.

Patterns of species diversity were analysed using two types of diversity, that is, α-diversity and β-diversity, and also evenness. Two aspects of α-diversity were analysed, the first being species richness (S) that is defined as the number of species per sample plot. Because S can be exaggerated by the presence of rare species, α-diversity was also measured with the Shannon–Wiener diversity index (H′). It is a weighted expression of species richness and the proportion in which each species is represented in a sample plot, which is calculated as:

H′=∑i=1Sρilnρi,[Eqn 1]

where p_i is the proportion of cover in the ith species. The cover values used were based on median values (except Category r & +) of the cover categories derived from the Braun–Blanquet cover-abundance scale: 1% for Category r & + (cf. Ma 2005), 3% for Category 1 (1%–5%), 9% for Category 2a (6%–12%), 19% for Category 2b (13%–25%), 38% for Category 3 (26%–50%), 63% for Category 4 (51%–75%) and 88% for Category 5 (76%–100%). Evenness, defined as the relative abundance of species in a unit area (Stirling & Wilsey 2001; Wilsey & Stirling 2007), was used to measure the similarity of relative abundances of species within sample plots (Sankaran 2009). It was calculated with the Pielou’s evenness index (J′) as

J′=H′/lnS.[Eqn 2]

A one-way analysis of variance (ANOVA) with Tukey’s honestly significant difference (HSD) test (using SPSS® software version 19) was then conducted to compare S, H′ and J′ between the different VTs.

Beta (β) diversity was calculated to determine species turnover or the extent to which species diversity differs within the VTs. Various measures of β-diversity have been proposed over the years, but in the present study, the Whittaker’s diversity index (β_w) was used because it is widely regarded as a simple but highly effective measure of β-diversity (Magurran 2004; Van der Maarel 2005). It was calculated as:

βw=Stotal/Save′[Eqn 3]

where S_total is the total number of species present in each VT (γ-diversity) and S_ave is the average species richness (α-diversity) for each sample plot in a community.

The vegetation of Bloemfontein is dominated by the red grass Themeda triandra with Eragrostis lehmanniana as a constant companion. Other prominent grasses are Aristida congesta, Digitaria eriantha, Sporobolus fimbriatus and Eragrostis curvula. Forbs such as Oxalis depressa, Hibiscus pusillus and Felicia muricata are widespread, but they never attain dominance within the communities. It is only in localised wetland habitats where hydrophytic sedges and other forbs dominate. The vegetation is represented by 77 families and 248 genera (Table 1). The families with the highest number of genera are Poaceae, Asteraceae and Fabaceae (Table 2), while the most diverse genera are Eragrostis, Cyperus, Aristida and Asparagus (Figure 1). A total of 376 plant species were identified for the study area: 82 play a diagnostic role, 66 are companion species and a further 228 are either localised or of very rare occurrence. Of these rare species, only 175 are presented in the synoptic table (see Appendix 1); the other 53 are excluded because of their extremely rare occurrence. Eight species are declared invasives (Department of Environmental Affairs 2016), namely Argemone ochroleuca subsp. ochroleuca, Verbena bonariensis, Gleditsia triacanthos, Cestrum laevigatum, Cuscuta campestris, Pennisetum villosum, Salsola kali and Convolvulus arvensis. None of the species recorded is threatened (http://redlist.sanbi.org/).

A synoptic classification of the vegetation is presented, showing only the major VTs and not the lower ranked syntaxa constituting each VT. The following five major vegetation units and four subdivisions were recognised from the study area, as summarised in Table 3:

The DECORANA ordination plot (Figure 2) shows a clear grouping of the relevés into the VTs as classified in the synoptic table (Appendix 1). Axis 1 and Axis 2 have eigenvalues of 0.863 and 0.664, respectively. Axis 1 is positively associated with the soil moisture gradient, beginning with the D. pottsii–C. orbiculata Grassland and the O. capense–E. nindensis Grassland on the left portion of the ordination plot, which are characteristics of relatively dry habitats. The R. lanceolatus–C. longus Streambed vegetation and V. karroo–A. laricinus Streambank vegetation are found towards the right end of the ordination plot because of their occurrence in wetter habitats. Axis 2 does not show any clear environmental trends.

The deepest soils were recorded in the R. lanceolatus–C. longus Streambed vegetation (433 mm ± 78 mm) and V. karroo–A. laricinus Streambank vegetation (475 mm ± 62 mm); these soils also have the highest pH of 7.2 ± 1.1 and 7.4 ± 0.9, respectively (Table 4). The O. europaea–B. saligna Shrubland has the highest clay content (30% ± 5%) and organic matter content (4.65% ± 2.02%). The soils of the A. congesta–T. triandra Grassland and the F. muricata–T. triandra Grassland have the highest sand content at 76% ± 12% and 66% ± 8%, respectively. These communities also occur on relatively shallower soils with the average depth of 285 mm ± 105 mm and 205 mm ± 68 mm, respectively. No soil samples were collected for the D. pottsii–C. orbiculata Grassland and the O. capense–E. nindensis Grassland. The collecting of samples was mainly restricted by the shallow nature of the soil, compounded by the extensive dolerite rocks and boulders in habitats where these communities occur. The CANOCO biplot (Figure 3) reveals community correlations with soil depth, texture (clay, sand and silt contents), pH and organic matter. Axis 1 (eigenvalue 0.618) shows correlations with soil depth and clay content. Axis 2 (eigenvalue 0.553), on the other hand, is correlated with silt, organic matter, pH and sand.

The D. pottsii–C. orbiculata Grassland, O. capense–E. nindensis Grassland, O. europaea–B. saligna Shrubland and V. karroo–A. laricinus Streambank vegetation have high α-diversity (both S and H′), with S of 20.7 ± 5.7, 21.2 ± 8.6, 23.9 ± 6.7 and 24.4 ± 6.2, respectively, and H′ of 2.51 ± 0.42, 2.44 ± 0.70, 2.17 ± 0.43 and 1.96 ± 0.35, respectively (Table 5). The four VTs also have high J′ (0.84 ± 0.14, 0.83 ± 0.18, 0.69 ± 0.09 and 0.62 ± 0.08, respectively) and γ-diversity (130, 115, 137 and 128, respectively). The R. lanceolatus–C. longus Streambed vegetation has the lowest γ-diversity (75), S (9.6 ± 4.9), H′ (1.14 ± 0.61) and J′ (0.50 ± 0.21). With regard to β-diversity, relatively low β_w (5.2–6.3) was measured for the D. pottsii–C. orbiculata Grassland, O. capense–E. nindensis Grassland, O. europaea–B. saligna Shrubland and V. karroo–A. laricinus Streambank vegetation. The highest β_w was recorded for F. muricata–T. triandra Grassland (9.3) and R. lanceolatus–C. longus Streambed vegetation (7.8), while the lowest β_w (5.1) was recorded for A. congesta–T. triandra Grassland.