Habitat loss and fragmentation drives the current extinction crisis. The processes through which it affects biodiversity, however, are complex and poorly understood. This is especially true for spatially complex regions that comprise a mosaic of land-use types, which often range from protected areas to dense human settlements. In such human-modified landscapes, it is important to determine the extent and impact of changing land-use patterns on biodiversity if we are to meet conservation targets or regain ecosystem services. My analyses of coastal forests in KwaZulu-Natal suggest that extensive loss of forests (82%) incurred an extinction debt, modelled to match the 11 bird species now listed as threatened locally. Forest fragments are now also smaller, fewer, further apart and more encroached by human land uses than in the past. Yet, species interactions with the gradient of habitat conditions that now surround forest fragments may have forestalled the realisation of predicted extinctions. I found that natural matrix habitats adjacent forest fragments (e.g. grasslands and woodlands) may facilitate dispersal, enable species spillover from forest fragments, and buffer forest interiors from changes in abiotic conditions associated with high contrast matrix habitats (e.g. agricultural plantations). However, when natural matrix habitats were transformed, these processes were disrupted, which suggest that the effect of landscape change on coastal forest diversity may stretch beyond forest loss per se and the deterministic extinctions predicted by conventional species-area relationships.
Next, I determined that the response of different bird species to habitat fragmentation parameters (i.e. area, connectivity and matrix habitats), depended on life-history traits such as body size, feeding guild and habitat specialization. Extinction risk was, however, not a function of species traits or the fragmentation parameter species responded to. This means that a conservation approach that only focuses on restoring a single fragmentation parameter (e.g. area) may not be successful in halting predicted extinctions, simply because multiple factors may determine extinction risk in coastal forests.
The interpretation of biodiversity patterns in fragmented landscapes may, however, also be influenced by spatial scale. I therefore used a fractal-based sampling design to test how sampling at fine, intermediate and coarse scales influences (1) beta diversity of and (2) inferences from the modelled contribution of niche- versus dispersal-based assembly processes in structuring tree and bird assemblages. I showed that inferences from beta diversity are scale dependent. As a result, studies with similar sampling effort and temporal sampling protocol, but with different sampling grains are likely to report dissimilar ecological patterns, which may ultimately lead to inappropriate conservation strategies.
This thesis provides information of how land-use changes impact on biodiversity patterns and derived processes in a human-modified landscape. It also highlights some conservation opportunities in the coastal forest landscape mosaic, where conservation and restoration actions should focus on both forest fragments and on the surrounding matrices. The conservation of natural matrices may buffer forest communities from impacts associated with high contrast habitat edges, enhance natural plant regeneration through species spillover, provide important linkages between forest fragments, boost regional diversity and allow coastal forests to track environmental change under changing climatic conditions.