Increasing representation of localized dung beetles across a chronosequence of regenerating vegetation and Natural Dune Forest in South Africa

Aim: Species assemblages with high proportions of localized taxa occur in regional islands with a history of strong ecoclimatic separation from adjacent systems. Current disturbance in such islands of relictualism or endemism disrupts the distinctive local character in favour of regionally distributed taxa with a wider range of tolerances. However, rehabilitation of the system should restore the localized biota. Thus, we used biogeographical composition to assess progress towards restoration of the dung beetle fauna associated with such an island of endemism following dredge-mining. Location: The study was conducted in natural coastal dune forest and a 23-year chronosequence of regenerating dune vegetation in the Maputaland centre of endemism, KwaZulu-Natal, South Africa. Methods: Dung beetles were trapped in eight stands of regenerating vegetation of different ages (< 1 year to $\sim$21 years) and in four stands of natural dune forest with differing ecological characteristics defined by measurements of vegetative physiognomy and microclimate. Species groups defined from multivariate analysis of biogeographical distribution patterns and vegetation associations were used to demonstrate quantitative compositional changes in the dung beetle assemblages across the chronosequence to natural forest. Results: Three biogeographical groups were defined. One group comprised species widespread in southern Africa or both southern and east Africa. The other two groups were endemic, one to the east coast and the other to Maputaland. There was a general trend from dominance by regionally distributed dung beetle taxa to dominance by locally distributed taxa across the chronosequence of regenerating vegetation from grassland, to open Acacia karroo thicket, to dense A. karroo-dominated woodland. However, this trend was linked closely to the relative physiognomic and microclimatic similarity between the regenerating vegetation and natural forest. Thus, proportions of locally distributed taxa were lower in older chronosequence woodland ($\sim$18-$\sim$21 years) with its low canopy cover and open understorey than in dense early chronosequence woodland ($\sim$9-$\sim$12 years), which is physiognomically and microclimatically closer to species-diverse natural forest with its dense canopy and understorey. Overall, the present dung beetle community comprises five species groups. Single widespread (21 spp.) and endemic groups (14 spp.) showed similar patterns of association with early chronosequence grassland and open thicket stands. A single widespread (3 spp.) and two endemic shade-associated groups (3 and 11 spp.) showed differing patterns of association centred, respectively, in late chronosequence woodland, natural forest, or all shaded stands. Main conclusions: At 23 years, vegetative regeneration is still at an early stage, but abundant activity of most, although not all species recorded in natural forest, is recovered with the closure of the woodland canopy at $\sim$9 years. Compositional differences with respect to natural forest vary closely with vegetative physiognomy and its effect on the microclimate. Therefore, full compositional recovery is dependent on the re-establishment of natural forest physiognomy and microclimate.