The study of landscape dynamics in Amazonia requires a good understanding of the processes that rule land-use conversion, agricultural land abandon and forest recovering. To achieve such goal, one needs to develop new methods to map as well as to analyze spatial patterns resulting from changes induced by these processes. These methods will in turn enable the formulation of reliable models to operate at regional scale.

On these basis, this work aims at the development of methodology to obtain spatial transition models using mainly remotely sensed data and topographic maps. This study was done using an area located in the northern part of the state of Mato Grosso (Brazil), that stands as a typical Amazonian Frontier Colonization region. Map algebra was applied in order to generate the selected variables needed for the change analysis, e.g. slope, altitude, vegetation, soils, urban attraction, road buffers and distance to selected landscape elements. This process was done within a Geographical Information system (GIS) environment combined with multivariate statistical techniques. It could be shown that the maps resulting from the logistic regression gave a good indication of the areas most favorable for each type of transition, consequently they could be interpreted as maps of the spatial transition probabilities.

In addition to the quantified effects of the above variables, the relief, represented by the slope and altitude variables, showed a negative influence on the selection of sites to be deforested, and a positive effect on the abandon and regeneration processes that take place in the cultivated and pasture fields. Besides these results, the methodology developed represents an important step towards the discretization of landscape dynamic simulation models into sub-areas.