Aim: To identify biogeographical boundaries which are obscured by faunal overlap and habitat modification. Location: KwaZulu-Natal in south-east, South Africa beyond the southern tip of the Mocambique Coastal Plain. Methods: Species abundance data for dung beetles (Scarabaeidae: Scarabacinae) were collected at six levels from the coastal escarpment (30 degrees 16'S - 500 m) to the top of the nearby Drakensberg (29 degrees 35'S - 2850 m). Cross-altitudinal boundaries were identified using clustering techniques, beta-diversity indices, and range edge analysis. Biogeographical data for the species were drawn from an extensive reference collection and used to classify the biogeographical affinities of the assemblages. Results: Three discrete communities are defined (<10% similarity) from species abundance distributions. These communities occur in coastal forest (500 m), coastal to highveld grassland (500-1500 m), and montane grassland (1900-2850 m). Two of these communities are biogeographically homogeneous comprising >89% east coast endemics (coastal forest) or >84% South African montane endemics (montane grassland) in terms of abundance. The third community in coastal to highveld grassland is biogeographically more heterogeneous. Predominant biota of this community comprise both South African highveld endemics and elements with distributions extending into the tropics. At highveld levels (1500 m), there are proportionately more highveld endemics whereas at lowland levels (500 m), there are proportionately more tropical elements. At 1000 m, there was a change in the balance between these two groups across an anthropogenic gradient due to a decline in the proportion of endemics in favour of temperate/tropical generalists. This gradient from a natural grassland fragment to improved pastures of Kikuyu grass also parallels a decline in species richness and abundance. Species turnover analyses showed three different cross-altitudinal patterns. Range-edge analysis showed a trimodal pattern of species turnover (peaks in forest and the Drakensberg foothills as in the community analysis but also at (1000 m). Five beta-diversity indices showed either a bimodal pattern of turnover (forest/grassland and foothills/middle Drakensberg slopes) or a trimodal turnover pattern (forest/grassland, highveld Drakensberg foothills, Drakensberg peaks). Main conclusions: Clear altitudinal zonation is revealed by community and biogeographical analysis but one natural biogeographical boundary may be obscured by the process of habitat modification. This boundary at 1000 m is revealed by range-edge analysis and is supported by findings for plant communities. Beta-diversity, species turnover patterns diverged slightly from those suggested by the community and range-edge analyses.
