Influence of drought or elephant on the dynamics of key woodland species in a semiarid African savanna.
Extensive drought - and elephant-related dieback of Colophospermum mopane and Acacia tortitis, respectively, offered an opportunity for increasing understanding of the causes of drought-related patch dieback, the factors influencing elephant utilization of woody plants, and the response of woody plants to both aforementioned determinants of savanna structure and function. The dendrochronological analysis of both species was undertaken to estimate potential rates of replacement, following extensive mortality. Areas of discrete dieback were compared with adjacent paired areas of 'healthy' vegetation, which revealed, on average, 87% and 13% loss of basal area by mortality, respectively. 'Live' and 'dead' plots did not differ in soil type, topography or mean slope, but differed in vegetation structure, soil surface condition, and soil chemistry. Although there was evidence of self-thinning, neither inter - nor intra-specific competition explained dieback. 'Dead', by comparison with 'live' plots, had changed from functioning as sinks of sediment and water to sources, were less likely to retain water because of a poor soil surface condition, and were predisposed to drought effects because of a greater proportion of fines, and Na concentration. Dieback resulted from insufficient soil water for survival during a drought owing to the development of a dysfunctional landscape during 50 years of livestock ranching. Spatial heterogeneity within a landscape was suggested to enhance woodland resilience to severe droughts by ensuring the survival of plants in run-on sinks or 'drought refuge' sites. Stem sections were removed from 40 multi-stemmed C. mopane trees and prepared for examination under a dissecting microscope. It was impossible to age C. mopane, owing to a hollow and/or dark heartwood. Nevertheless, the distribution of stem diameters suggested a single recruitment event. Fire scars attributed to the last recorded fire in 1948 could explain the trees' multi-stemmed growth form and indicate that most trees of VLNR were > 50 years of age. Growth rings were identified in 29 A. tortilis trees of unknown age, but were not correlated with annual rainfall records. Growth rates varied between trees; mean ring width ranged from 1.4 to 3.5 mm (overall mean 2.4 ± 0.1 mm). A technique was proposed for predicting growth rate from annual rainfall, using selected data, and several factors potentially influencing ring width in semiarid environs were identified. Permanent ground-based transects were located within riparian (n = 16) and Acacia (n = 5) woodlands to monitor elephant utilization. Elephant had not changed the population structure of the woodlands by 2000, but had reduced stem density from 215.6 stems ha -1 (1996) to 84.4 stems ha -I (2000). Acacia tortitis trees in the woodlands had branches removed, were debarked, uprooted and broken. Acacia tortitis trees in the riverine had lower levels of utilization, whilst Acacia nilotica trees were mostly debarked. The method of elephant feeding varies within and between woody species, provided it is within the mechanical constraints of a certain size or species. Elephant behaviour is concluded to depend on spatiotemporal variation of forage abundance/quality, abundance of a preferred species, and species response (coppice or mortality). Elephant can cause a change of vegetation state, and increase spatial homogeneity of a plant population. The remnant population of woodland trees should provide the potential for recolonization, in which case the system would reflect the stable limit cycle. However, if browsing inhibits seedling recruitment, the system could reflect either a multiple stable state system or an artificial equilibrium imposed on a stable limit cycle.