Fires are a regular feature of savanna ecosystems worldwide. Although Namibia is the most arid country across Sub-Saharan Africa, the seasonal occurrence of fires is widespread. Humans and biophysical controls are known to govern the spatio-temporal patterns of fire. Yet, the interplay among the controlling factors and their individual contribution to the generation of fires lack generality. An overall impact of fire on vegetation and its structure is controversial – especially in drier regions. Remote sensing provides a unique means for the assessment and modelling of fire regimes and vegetation. Earth observation missions such as the Moderate-Resolution Imaging Spectroradiometer (MODIS) offer consistent records of fire and quantitative vegetation parameters. The scale of observation in space and time impose an inherent source of uncertainty with any remotely-sensed dataset. As such, background contamination and phenology usually complicate the discrimination of sparse green vegetation. Unmanned Aerial Vehicles (UAV) introduce new 3D opportunities for optical remote sensing, however their full potential remains to be explored. In the present study, remote sensing and spatial modelling are the primary tools for a quantitative investigation of fire and vegetation parameters across Namibia. Several spatial datasets are applied to achieve this task. These range from readily-available thematic products from Earth observation over higher-resolution RapidEye and UAV imagery to vector datasets. Fire regimes are analysed and modelled using a set of common statistical and machine learning techniques. Field measurements and upscaling techniques are combined in order to comparatively explore the estimates of Leaf Area Index (LAI). Imagery generated from an UAV mission facilitates the reconstruction of vegetation structure in 3D by means of a photogrammetric approach known as Structure-from-Motion - Multi-View Stereopsis (SfM-MVS). Woody individuals are then delineated in order to yield approximate stand structures. The results show that productivity is the major control of fire activity in Namibia. A distinct increase in both Burned Area (BA) and Fire Occurrence (FO) with a mean annual precipitation above 400 mm is observed and located in the northern parts of the country. Although humans are known to account for the majority of ignitions, their activities also consume the fuels that are required for burning. Hence, increasing densities of population and livestock reduce fire activity across the country. A case study from Owamboland in northern Namibia confirms the uncertainties that are associated with the spectral remote sensing of low-productivity ecosystems. As such, a mean underestimation of 0.34 (±0.2) is found with the estimates of LAI from MODIS (MOD15A2), which are compared to an empirically-calibrated model of LAI. In contrast to the general underestimation by MOD15A2, overestimations of LAI are apparent in the case of a recent fire in the region.