In quantitative landscape ecology and environmental modelling, the spatial delineation of a landscape and its differentiation into landscape units is often based on subjective criteria. At the same time, the precise spatial definition of a landscape as a study area is often crucial for deriving accurate model outputs. In this study we present a promising solution to this problem by quantitatively assessing the complex term 'landscape' using an empirical statistical approach. This analysis is based principally on a digital elevation model (DEM) derived from stereoscopic ASTER sensor data. DEMs were produced for varying spatial subsets of two ACACIA landscapes in Northern Namibia: (1) Lower Kunene Hills (LKH) and (2) Upper Kunene Hills (UKH).