The response of bird assemblages to landscape transformation along the northeast coastal plains of southern Africa

The fragmentation and loss of natural habitats brings about species losses. These losses may be amplified or reduced by the composition of the habitats surrounding remnant habitat fragments, i.e the matrix. The influence of the matrix on biological assemblages is, however, complex and poorly understood. Therefore, in this dissertation, I aim to determine whether matrix transformation influences bird assemblages within remnant natural habitat fragments and how patterns of bird diversity and the processes which structure those patterns are influenced at the landscape scale. In the first study component of this dissertation, bird assemblages were surveyed within 29 and 30 forest fragments embedded within contrasting matrix types: natural (grasslands and woodlands) and transformed (sugarcane, agroforestry, human settlements), and categorised as either forest-specialist or habitat-generalist species. I first fitted eight species-area models to evaluate species-area relationships (SARs) in both matrix types. Second, I calculated Pielou’s evenness index to determine if the relationship between evenness and area varied among matrix types. Third, I used a nestedness metric based on overlap and decreasing fill to determine if matrix type influences nestedness amongst forest bird assemblages. Finally, I used binomial logistic regressions to determine the relationship between the presence of each species and forest fragment area in each matrix type. I found that SARs for habitat-generalists were weak within transformed and natural matrices. For forest-specialists, the SAR was weak when the matrix was natural, however, when the matrix was transformed the SAR was strong and resembled that of real islands systems. Forest-specialist assemblages were also significantly nested by area within transformed, but not natural matrices. Moreover, fragment area significantly affected the occurrence of 11 forest-specialist species in transformed matrices, compared to only one species in natural matrices. In the second study component of this dissertation, I made use of a fractal-based design to survey bird assemblages in seven different habitat types (four human-modified and three natural habitats). I calculated and compared estimated species richness (Chao1) for each habitat. I then investigated and compared patterns of alpha and beta-diversity within each habitat at three different spatial sampling grains, using generalised linear models and multiple comparison tests. I also made use of the Raup-Crick null modelling approach to determine if processes structuring bird assemblages were more deterministic (niche-based) or stochastic (dispersal-based) than expected by chance. My findings suggested that natural habitats did not always support higher species richness. Rural and urban areas were also capable of supporting high species richness comparable with that of natural forests and woodlands. In contrast, grasslands supported reduced species richness comparable to sugarcane and Eucalyptus plantations. Patterns of beta-diversity were found to be inconsistent and to decrease with an increase in spatial grain. Therefore, inferences based on beta-diversity were dependent on the scale at which the analysis was conducted. In contrast, patterns of alpha-diversity were not influenced by spatial grain. The processes driving community assembly were found to not be influenced by anthropogenic transformation and were non-random, and influenced by deterministic niche-based processes. In this dissertation, I therefore show that forest-specialist bird assemblages may conform to island biogeographic predictions of species loss when embedded within transformed, but not natural matrices. Species extinctions from small forest fragments may, therefore, be prevented by conserving natural- or restoring transformed matrices, and not only by increasing forest area. Moreover, I show that while the conservation of natural habitats is important all habitat types in a landscape are important for regional diversity. Therefore, the inclusion of human-modified habitats into conservation planning and the maintenance of structural heterogeneity within these landscapes may be important for maintaining landscape biological diversity in the future. Finally, I show that a multi-scale sampling approach is important when making inferences about patterns of diversity at the landscape scale.