Savannahs, which are defined as a heterogeneous mixture of herbaceous and woody plant components, occupy one fifth of the global land surface and is the largest biome in South Africa. The woody vegetation structure of savannahs is particularly important as it influences the fire regime, nutrient cycling and the water cycle of these environments and provides fuelwood to sustain the local human populace. Remote Sensing has been proven in numerous studies to be the preferred tool for quantifying and mapping this woody vegetation structure (in this study, defined as woody biomass, woody canopy volume and woody canopy cover metrics) over large areas, mainly due to its superior information gathering capabilities, wide spatial coverage and temporal repeatability. Active remote sensing sensors such as Light Detection and Ranging (LiDAR) and Synthetic Aperture Radar (SAR) are particularly useful in studying woody biomass and other canopy structural metrics, because of their capacity to image within-canopy properties. Passive optical imagery acquired over multiple seasons can also provide tree phenological information which can be used to ascertain the best period for monitoring tree structure, i.e. when tree canopies has sufficient leaves while the grasses are dry. The combined strength of these active (SAR and LiDAR) and passive (optical) sensor technologies, are yet to be applied to their full potential in the dynamic and heterogeneous savannah environment, with a special relevance in Southern African landscapes. This PhD study aimed to evaluate various methods for estimating and upscaling woody structural metrics of South African savannahs using integrated SAR and optical remote sensing datasets and LiDAR datasets as training and validation. Before this aim could be tackled, two current global-scale remote sensing woody structural products (25m JAXA ALOS PALSAR Forest/Non-Forest or FNF and 30m Landsat-based Vegetation Continuous Field or VCF) were evaluated, within the South African context, with the help of high resolution airborne LiDAR datasets. These datasets were resampled to match the products’ criteria and definition used to depict forests. It was found that the FNF product grossly under-represented the distribution of forests in savannah environments (20-80% CC ranges), due to the inadequate HV backscatter threshold chosen in its creation. The FNF product also showed a limited ability in detecting closed forest cover class (90-100%) and Natural Forest and Scrub Forest tree structural classes. The Landsat VCF product displayed strong CC underestimation with increasing variability and mean error from CC values of greater than 30%. The moderate accuracies at the 10-20% CC range (and in the Open Woodland tree structural class) suggests that the VCF product could be potentially applicable in low CC environments such as grasslands and sparse savannahs but can also marginally detect closed canopy environments (90-100% CC range). These results provide the justification for developing new, locally calibrated woody structural products for South Africa. Next, the aim of this study was addressed, firstly, by developing methodologies for the estimation of key woody structural metrics (above ground biomass, woody canopy cover and woody canopy volume) for the Greater Southern Kruger National Park Region using multi-frequency SAR parameters (X-, C- and L-band backscatter and polarisations). Secondly, the most suitable SAR frequency was then tested against and in combination with various Landsat-5 TM optical features (textures, vegetation indices and multi-seasonal band reflectance) for improved regional modelling of woody canopy cover. In both cases, In-situ field measurements of woody vegetation structure were “scaled-up” to landscape and regional scales by using LiDAR, SAR and/or optical sensor products to produce reliable maps of woody structural metrics. A Random Forest modelling approach was predominantly used to meet the modelling challenges in this study and the LiDAR datasets were used for model calibration and validation.