The long-term economic viability and ecological integrity of savanna rangelands is being undermined by increasingly dense woody thickets at the expense of palatable herbaceous cover. This process is known as shrub- or bush-encroachment. Bush encroachment is a subset of a broader ecological riddle underlying the coexistence of woody and herbaceous vegetation that has been the subject of many ecological models. The ecophysiological mechanisms and interactions between trees and grasses on which most assumptions of ecological models are premised have seldom been tested empirically. This document synthesises the results of greenhouse and field-based investigations of the underlying ecological mechanisms and ecophysiological interactions between encroaching Acacia mellifera trees and grasses in a semi-arid environment. In a greenhouse study, I determined the contribution of N2 fixation to the N-budget of Acacia mellifera under conditions of both varying N availability and competition from grass. Tree seedlings had longer shoots and greater total dry mass in the absence of grass. The leaf δ15N values were lower with grass than without grasses. Thus, trees were more reliant on N2 fixation in the presence of grasses. N2 fixation may enable the tree seedlings to survive competition with grass at critical and vulnerable developmental stages of germination and establishment. In a field removal experiment, I monitored the growth rates, water relations and mortalities of shrubs around which neighbouring woody plants were removed (target) and control shrubs over three years. Results showed target trees to have benefitted from removal of neighbours, which was manifested in significantly faster growth rates, less negative predawn water potential and a relatively small degree of canopy die-back. Nonetheless, neighbouring trees appeared to prevent the whole plant mortalities resulting from severe environmental stress. Growing in close proximity with neighbours could therefore yield positive and negative ecophysiological effects. In another greenhouse experiment, I tested the effects of the separation of moisture uptake with depth between tree seedlings and grasses on two common substrate types. I also examined the influence of repeated grass clipping on the persistence of soil moisture. Results indicated a three-tier rooting pattern with a top layer exclusively exploited by grasses, an intermediate layer occupied by both grass and tree roots and deeper layers exclusively tapped by trees. Tree seedling biomass was negatively affected by grass competition although the biomass of grass was enhanced in the presence of tree seedlings on sandy substrates only. The repeated clipping of grass benefitted tree seedlings on rocky substrate more than it did on sandy substrate. The effects of heavy grazing on soil moisture availability to woody shrubs and thus bush encroachment may be contingent on substrate type, being more acute on rocky terrains. Grass competition suppresses tree seedlings but the removal of grass by grazing weakens this suppressive effect, particularly on rocky substrates. The insufficiency of space and soil resources on rocky substrates may necessitate increased investment in root biomass by plants. It is not known why grasses have lower densities on rocky substrates than on sandy substrates, but the obstruction by rocks disadvantages grasses against tree seedlings, leaving grasses vulnerable to grazing pressure. This may allow the woody plants on rocky substrates to benefit more from grass removal than on sandy substrates. Root restriction by rock barriers and, perhaps, sparse soil volume further lead to small tree sizes on the rocky substrate. Small shrubs are less likely to compete intensively for resources and cause density-dependent mortality. Intraspecific competition may maintain shrub sizes within the threshold that can be supported by available resource pools. I conclude from my results that the two-layer hypothesis of niche separation between savanna vegetation is valid although there is an overlap in the grass and tree rooting depth/moisture uptake. An additional factor that affects the success of A. mellifera is the substrate.Trees are more dense on rocky substrates but grow larger on sand. I have further shown that A. mellifera trees fix nitrogen when competing with grasses but do not do so when grasses are absent. A mechanistic model of savanna dynamics will need to integrate water use patterns, substrate and nutrients to make effective predictions about encroachment patterns. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2010.