Terrestrial ecosystems support a high plant diversity where different plant types coexist. However, the mechanisms that support plant coexistence are not entirely clear. Savanna ecosystems that are nutrient and water limited are characterized by a unique ecological feature: the coexistence of trees and grasses. Tree-grass interactions in savannas are typically viewed as being competitive and are based on the Gaussian principle of niche or habitat differentiation. Trees and grasses are reported to suppress the growth of each other and the interactions are viewed as competition. However, tree-grass mixtures persist in a range of rainfall conditions in savannas. This study examined tree-grass interactions to understand the ecological processes that may sustain tree-grass coexistence in dry savannas (< 800 mm of rainfall) of southern Africa. Water and nitrogen resource-use patterns of trees and grasses were investigated and the effects of competition between trees and grasses on resource storage in perennial grasses were examined. An ecological perspective of the role of trees in two human land-use types in African drylands is provided and the functionality of trees in these land-use types was reviewed. Seasonality of plant available water imposes intense water limitation to plants in savannas. Yet, trees and grasses coexist. The water relations between trees and grasses are poorly understood. In Chapter 2, the principal water-sources for trees and grasses in different seasons were identified using the natural variation in H and O stable isotope composition of source waters. Seasonal differences in the stable isotope composition of water in trees and grasses indicated that there was water-source use partitioning as well as overlap. Trees and grasses used water from the topsoil after rainfall indicating overlap of water-sources. Trees shifted to groundwater or subsoil water when there was no water in the topsoil, indicating partitioning of water-use. Grasses always used water from the topsoil. By labelling deep-soil (2.5 m depth) with a deuterium tracer, hydraulic-redistribution in all the studied tree species and water transfer to grasses via the topsoil was confirmed. However, this occurred only in the dry-season. Results indicated possible shifts in tree-grass interactions during different periods of the year. Furthermore, dry-season hydraulic-redistribution indicated potential facilitation affects by trees to their understory grasses. A key question in savanna ecology is how trees and grasses coexist under nitrogen limitation. In Chapter 3, the sources of nitrogen for trees and grasses in a semi-arid savanna were investigated using natural abundance of foliar δ15N and nitrogen content. 15N tracer additions were used additionally to investigate the redistribution of subsoil nitrogen by trees to grasses. Foliar δ15N values were consistent with trees and grasses using mycorrhiza-supplied nitrogen in all seasons and a switch to microbially-fixed nitrogen during the wet season. Based on seasonal variation in mineralization rates in the Kruger Park region, the use of mineralized soil nitrogen by trees and grasses seemed highly unlikely. The foliar δ15N values were similar for all the studied tree species differing in the potential for nitrogen-fixation consistent with the absence of nodules indicating the lack of rhizobially fixed nitrogen. The tracer experiment showed that nitrogen was redistributed by trees to understory grasses in all seasons. Redistribution of nitrogen by trees and subsequent uptake of this tree redistributed nitrogen by grasses from the topsoil was independent of water redistribution. Although there was overlap of nitrogen sources between trees and grasses, dependence on biological sources of nitrogen coupled with redistribution of sub-soil nitrogen by trees could be contributing to the co-existence of trees and grasses in semi-arid savannas. An important plant response to competition and resource limitation is an increase in root reserves. In Chapter 4, the root characteristics of perennial grasses in the presence and absence of trees as a proxy of competition in South African savannas in three sites that differed in rainfall were investigated. The hypothesis on which this investigation was based was that competition from trees and water limitation will result in increased storage in roots of grasses under trees. However, no significant effect of variation in rainfall of the different study locations on root characteristics of grasses were found. Furthermore, most root characteristics were not significantly influenced by tree presence with the exception of nitrogen-content. The root nitrogen content showed an increase with rainfall and tree presence through potentially higher mineralization rates and nitrogen availability in the under-tree canopy environment. The study sites occurred in the drier rainfall range in South Africa. Therefore, it is likely that trees and grasses in these dry savannas might have a positive relationship conforming to the stress-gradient hypothesis. Alternatively, grasses and trees might be using complementary water and nutritional resources. The mix of trees and grasses is critical for the functioning of the savanna biome, which supports a large fraction of the human population and sustains the highest densities and diversities of herbivores in the world. Both, increases and decreases in tree densities have been reported from savannas globally, which are attributed to human activities and climate change. Changes in tree densities could drastically impact ecosystem functioning and lead to land degradation and large economic losses. Consequently, the sustainable and heterogeneous nature of various savanna land-use types is compromised. In Chapter 5, the significant role of trees in dry savannas (< 800mm rainfall) based on nutrient and water-redistribution capabilities of savanna trees is illustrated. An ecological perspective of the role of trees in two human land-use types in African drylands: agroforests and rangelands which include silvo-pastoral systems and mixed-game-livestock farming systems, is provided. The causes for the loss of trees in these land-use types is evaluated and the role of trees for better land and sustainable natural resource management is highlighted. Chapter 6 synthesises the conclusions of all the preceding chapters highlighting the importance of facilitative interactions in tree-grass coexistence in savannas that are mostly overlooked. A simple Gaussian model of niche or habitat differentiation may not be a holistic and functional explanation of plant coexistence but rather the role of biotic interactions that include symbionts, parasites, or predators that will influence not only the competitive ability of plants but also facilitation, may be more pragmatic. Plant-plant interactions are complex and a multitrophic approach may be necessary to understand the functioning of these interactions and their roles in ecosystems.