Study area

The study area is the central Namib Desert and semi-desert and savanna transition zone in Namibia (Giess 1971). It includes granite inselbergs in the Tumas, Spitzkoppe, and Brandberg area, as well as the Erongo Mountains (Figure 1; center approximately at 21°30′ latitude South and 15°00′ longitude East). It stretches some 320 km from north to south. The Brandberg rises to 2573 m, with an elevation of approximately 1800 m above the surrounding plains. The Erongo Mountains' highest peak is 2319 m, rising some 1200 m above the plains at its western edge. The inselbergs range from 50 to 470 m in height above the surrounding plains (Table 1).

The study area lies in a transitional zone between the desert, a narrow strip of the Nama Karoo, and the savanna biome in Namibia (Irish 1994). Rainfall decreases from east to west, with a long-term annual average of 290 mm at Omaruru, east of the Erongo Mountains, 134 mm at Usakos at the edge of the desert, and decreasing to as little as 15 mm near the coast (Weather Bureau, Windhoek). Most rains fall in the late summer months (January–March) and rains are unpredictable, highly variable between years, and extremely patchy.

The inselbergs studied are granite domes of Cretaceous origin (Geological Survey 1997) characterized by steep slopes with an overall rounded appearance and many boulders, particularly around the bases of these mountains. The Brandberg is largely composed of granite, while the Erongo Mountains are composed of a mix of granite and volcanic material, largely rhyolite and undifferentiated basalt lava flows (Geological Survey 1997).

The zonal vegetation in this area has been described as Calicorema capitata and Euphorbia damarana zone (Hachfeld 2000), with savanna species such as Acacia reficiens becoming more prominent towards the east.

Data analyses

Plant species lists were compiled for 11 sites using field data of inselbergs and the Erongo Mountains that were surveyed by the author during the period 1998 to 2004, as well as published data, and data from the Specimens Database of the National Botanical Research Institute of Namibia. The survey methodology has been described in detail by Burke (2002) and is here summarized briefly. To ensure the best coverage, field surveys were undertaken during an above-average season. Plant species abundance was recorded along transects from the bottom to the top of the inselberg or mountain under study, including species occurring nearby. The detailed information on plant species abundance was then summarized to obtain species lists per inselberg. A species list for the Brandberg was based on published information (Craven and Craven 2000), while a species list for the Erongo Mountains was generated using the author's own surveys and data from the Specimens Database of the National Botanical Research Institute. The nomenclature follows Craven (1999).

For the purpose of this study, perennial plants were defined as those that were always visible, even during a poor rainy season. This included all shrubs, trees and perennial succulents, but excluded herbs, geophytes, hemicryptophytes and grasses, some of which are strictly speaking perennials. Two plant species lists were then compiled for each site: one comprising all plant species, the second with perennials only. All analyses were carried out twice, once for all species and then for perennials only. Unique species were defined as those which were only reported once during this study (these are not necessarily endemic to these sites, but were not reported from other sites included in this study).

To investigate the floristic relationship between sites, similarity matrices of the flora data were prepared with the PRIMER computer program (Carr 1997) using the Bray-Curtis similarity index and 4th root transformation of the original data. In order to obtain some indication of surface area, perimeters of all inselbergs and mountains were determined based on topographic maps and then used to determine the (two-dimensional) base area of each inselberg and mountain. Although this is not equivalent to the surface area of an inselberg, it provides a comparative measure of size.

After testing the data for normality, the influence of distance between the sites was explored using linear regression analysis (Zar 1984). Two regression analyses were undertaken: (1) at the regional scale (all data) and (2) at the local scale (data <60 km distance, which excluded all distances beyond geographic areas). A 5% significance level was used in all statistical analyses.

Floristic affinities

When all plant species were included in the analysis, the two large mountains, Brandberg and Erongo, were clearly separated from the granite inselberg floras, sharing just under 30% of their plant species with the inselbergs (Figure 2). With the exception of Bloedkoppie, the remaining inselbergs formed 3 distinct groups which clearly reflected their spatial relationship to each other. These patterns emerged even more clearly when only perennial plants were included in the analysis (Figure 3). Now there were 3 distinct inselberg groups based on geographic area, and one group comprising the large mountains. However, the hierarchy of these groups differed from that with the complete data. Most separated from other groups were the northernmost inselbergs (Duine, Blaauwpoort and Granietkop). The two large mountains (Brandberg and Erongo) were now only second in separation from the other groups. The hierarchies within groups remained the same, except for Bloedkoppie, which joined the group of the southernmost inselbergs (Amichab and Tumas). Another difference was that similarities were generally slightly higher, eg the most separate group shared more than 30% of its species with other groups.

Floristic affinities and distance between sites

Based on 55 observations, distances between the sites ranged from 9 to 318 km (mean 139.7, standard deviation 87.6) (Table 2). The Bray-Curtis similarity measures for all species ranged between 12 and 64% (mean 34.3, standard deviation 12.3), and for perennials from 19 to 66% (mean 35.6, standard deviation 11.2). The data were normally distributed. Comparing individual observations for all flora on the one hand, and for the perennial flora on the other, indicated that the slightly higher similarities for perennials were generally maintained. However, in some instances similarities of perennials were lower than for all plants (eg more than 1% lower: Amichab–Granietkop, Amichab–Klein Spitzkoppe, Amichab–Klipkop, Blaauwpoort–Klein Spitzkoppe, Duine–Klipkop, Gross Spitzkoppe–Granietkop, Granietkop–Tumasberg, Klein Spitzkoppe–Klipkop, Klein Spitzkoppe–Tumasberg). With the exception of Klein Spitzkoppe–Klipkop, these were all at distances >60 km, ie observations at the regional scale.

At the regional scale (ie when all distance data were included), although positive and significant (p <0.001), the linear correlation between distances and the similarities of floras was very weak (R² = 0.17, n = 55). When only perennial plants were regressed against distances, a highly significant (p <0.0001), positive, but still weak (Fowler and Cohen 1992) correlation was found (R² = 0.36, n = 55). At the local scale (<60 km distance, ie within geographic groups), the linear correlation was positive, significant (p <0.0001), and modestly strong (R² = 0.66, n = 15) (Figure 4). This correlation remained positive and significant (p <0.001), but was slightly lower when only perennials were investigated (R² = 0.52, n = 15).

Floristic affinities

Comparing the outcome of the complete and the perennial data indicated that the general patterns of floristic groupings were largely maintained and only their hierarchy was affected (Figures 2 and 3). At the regional scale of this analysis, floristic affinities based on perennial plants in this arid area hence provided a good indicator of biogeographic patterns. The change in hierarchy, however, may reflect a change in the relative contribution of explanatory variables, which are complex and likely represented by multiple interactive, abiotic and biotic variables (Huston 1997). For example, when all plants were included, the two large mountains, which provided the largest surface area and the highest altitudes, were clearly separated from the other groups, pointing towards surface area and altitude as critical, but likely not the only aspects controlling floristic relationships (Table 1). Using perennial plants only, the group of high altitude mountains became secondary in importance, and spatial relationships appeared to become more important as explanatory variables for these floristic patterns.

Not surprisingly, the most species-rich sites, Brandberg and Erongo, also showed the highest number of unique plants (ie species recorded only once during this survey). At the same time, these mountains also showed the highest altitude, greatest height above surroundings, and the greatest area (Table 1), indicating the importance of these abiotic variables as well as the conservation importance of these high mountains.

Floristic affinities and distance between inselbergs

Trends in floristic similarities were maintained between complete data and the perennial plant subset at the local level, but 8 sites in the regional distance category (>60 km) showed a reverse pattern. At regional scales there was a low statistical correlation between floristic similarities and distances, for both the complete data and the perennial subset. However, at local scales there was a modest, positive correlation and this was supported by both data sets—although there were slight differences in the strength of the correlation. The influence of scale on biogeographic questions has been demonstrated repeatedly (eg Clark et al 1999; Gaston 2000; Green and Ostling 2003) and was manifested in this study in the results of investigating the influence of distances on floristic relationships at different scales.

In contrast to expectations based on island biogeographic principles, which predict decreasing affinities with increasing distance (MacArthur and Wilson 1967), relationships became stronger with increasing distance between sites at the local level. This likely indicates the complex nature of the influence of abiotic and biotic variables on floristic affinities. There could be several reasons for this pattern: 1) the study area is heterogeneous and environmental variables not considered in this study, which could be called ‘hidden treatments’ (Huston 1997), were more important than distance; 2) floras within regions are fairly homogenous and the relationship between distance and floristic affinities within a geographic area is random, providing this seeming correlation by chance; or 3) the number of observations (15) may also have been too few to provide a conclusive answer. In fragmented landscapes, connectivity and appropriate corridors between habitat patches are important (eg Hanski et al 1996; Gonzalez et al 1998), more likely than distance between habitat patches per se.