The present thesis investigates the explicit representation of specific knowledge for the automatic interpretation of multitemporal low resolution satellite images. In this context, the term specific knowledge refers to all and any type of knowledge that makes an individual capable or more competent to carry out one determined task. In the scope of this thesis, specific knowledge comprehends the necessary information for the interpretation of low resolution satellite images, for instance: the characteristics of the classes in the legend, the agronomic management, and the ecology of the region under interest. Thus, the present thesis proposes a framework for the knowledge based interpretation of low-resolution satellite images which concerns at reproducing the reasoning used by the photo-interpreter while performing the visual interpretation. This model employs three different kinds of specific knowledge: 1) Spectral knowledge, that associates the diverse observed spectral signatures in the input image to the correspondent classes in the legend, grouping under a single spectral class the classes of the legend whose spectral signatures are difficult to be discriminated. 2) Contextual knowledge, which indicates the diverse contexts for the discrimination of the classes in the legend with similar spectral signatures. 3) Multitemporal knowledge, which relates, considering the previous classification, the possible classifications at the present moment and their respective possibility of occurrence. The potentiality of this methodology was evaluated through a series of experiments. The dataset consisted of images of two regions inserted in the Upper Watershed of the Taquari River, situated at the east of the Brazilian Pantanal, a lowlands ecological sanctuary located in the States of Mato Grosso and Mato Grosso do Sul. The main objective of the experiments was to evaluate the contribution of each sort of knowledge. The results indicate that the use of knowledge based methods can automate great part of the interpretation process, increasing the productivity of the photointerpreters. In the future, the results of the present research can guide the development of systems capable to automatically perform any interpretation strategy, defined by the proper photointerpreter, speeding up the monitoring of land use based on low resolution satellite images.