Tropical rainforest areas are difficult to classify in the digital analysis of remote sensing data because of spatial heterogeneity. Often many technical solutions are adopted to reduce the ‘problem’ of spatial heterogeneity. This thesis describes theory and methods that now use this heterogeneity during the digital image classification. With spatial heterogeneity, spatial aggregation levels such as patches,patch-mosaics and landscapes can be distinguished. Consequently, vegetation patterns can be related to functional management units at different decision-levels. The developed theory and methods thus save two birds with one stone: (a) the classification is completely digitally, and (b) the classification provides information on deforestation that meets the needs of decision-makers. This thesis further recommends approaching all land cover classifications from a heterogeneous perspective for understanding and controlling environmental processes on a global level. This can enhance a sustainable development of tropical rainforest areas for the benefit of future generations.