Self-thinning Concepts Applied to Savannas

Most investigations into savanna vegetation dynamics have focused either on tree-grass partitioning of water resources or on the effects of disturbances such as fire and herbivory on vegetation structure. Few studies have focused exclusively on tree-tree competition as a mechanism structuring savanna vegetation. The studies that have considered tree-tree competition have used nearest neighborhood methods to infer competition from the spatial pattern of trees, and the results of these studies for savannas have been mixed. However, no substantive work has studied tree-tree competition in savannas using the self-thinning concept, which is surprising since the concept is so heavily used in forest ecology. The self-thinning concept is a power law scaling relationship between mean size and density, with the intercept characterizing the carrying capacity of the system and the slope relating size-dependent resource use. Sankaran et al. (2005) have recently shown a pronounced linear relationship between average annual precipitation and maximum tree cover for a large number of savanna sites in Africa. We propose that tree-tree competition may be a likely mechanism generating the precipitation-rainfall pattern and that a self-thinning analysis can be helpful to further explain the relationship. Here, we examine self-thinning in savannas along a strong rainfall gradient in Kruger National Park, South Africa. The rainfall gradient varies from 750 mm annual average precipitation in the southeastern portion of the park to approximately 350 mm in the far north. The park also has a pronounced soil divide, with the western half of the park largely existing on granitic substrate and the eastern portion of the park on basalt. The study makes use of long-term fire-suppressed plots, where fire has been excluded for over 50 years but canopy cover levels are as low as 30 percent. Results presented show that the intercept increases uniformly along the rainfall gradient, but that the scaling exponent remains invariant and corresponds favorably with results for African forests suggested by (Enquist and Niklas 2001). These findings are particularly important in addressing uncertainties in tree-grass interactions, where it appears that beyond a threshold tree size class, tree-grass interactions are largely the same. Potential impacts of soil characteristics and nutrient heterogeneity on self-thinning relationships are further hypothesized. Criticisms using self-thinning in savannas are addressed as well.