Ecological systems are dynamic not static, that is, the condition of their components changes through time. The spatial evolutions of anthropoid ecosystems and the progressive transformation of spaces through time constitute a large preoccupation in space accommodation and have far reaching implications with respect to the environment. Therefore, this subject emerges more and more as a special interest issue in researches about the environment.

An underlying question arises, concerning the landscape evolution and the prospective of the state of a forest area in future:

We can answer this question by a spatio-temporal analysis, particularly the study of change through time of the environment. Thus we have an increasing urgent need to better understand the effects of dynamic process on the natural environment. Space-time analysis is required to model the evolution of temporal phenomena. A basic motivation for this study is to provide a tool for the analysis of spatio-temporal dynamics that corresponds to human conceptualizations of these dynamics.

In fact, the time constitutes a hierarchy of events and can produce transformations over a terrain landscape as appearance, disappearance, union or division of spatial entities. These transformations are called ``temporal phenomena '' [1]. The demand for forecasts about new situations, arising for instance as a result of changing environmental conditions, is on the increase. Models represent a possible tool in this respect as they make forecasting on the basis of current knowledge, combined with various types of information.

In a more general context of a reasoning system conception with geographic data, we have developed a geographic spatio-temporal reasoning module. We describe a method for analysis and forecast of a region area from a sequence of n thematic maps which represent the terrain state in different years. Such forecast is performed for a particular date provided by the user. The evolution modeling of regions is obtained with help of a sequence of aerial photographies compared through different dates. It allows the geographer interested in environmental prospective problems to get cartographical documents showing future conditions of a landscape.

Models are only useful if they address specific questions and refrain from attempting to reproduce the vast complexity of nature. In order to be clear and comprehensible, they must simplify the representation of what can be very complex systems. Understanding both the dynamics and the functional interrelations and processes of the environment concerned are essential for balanced planning and good management.

The method proposed provides the potentiality of evolution or disappearance in region areas, by analyzing forestry progression and regression zones. Thus, sensitive factors on regions evolution are considered for forecasting purpose. We used the method of analysis on an experimental area which reveals these spatial transformations. The results of this analysis are combined with out-image data that provide land area characteristics such as elevations, fire zones, distance to roads, distance to urban areas and expert knowledge.

The approach proposed is supported by the analysis of landscape dynamic of a mountainous region (Ariège's basin- France), in order to understand the evolutions of the whole forest.

In a test land area which reveals spatial changes through time, we propose a method for forecasting forest evolution in future years. For this, we shall use techniques of fuzzy subsets. This module developed is a specific spatio-temporal forecasting module which is a first attempt to solve this problem and leads to interesting results in this particular application domain.

The implemented approach models the evolution of the forest in a region in the south of France by using maps for the years 1942, 1962 and 1993. We used this method to study a region located in the Ariège mountains called "Soulane'' to describe the evolution of its landscape for the years 2000, 2025 and 2050. The experimental tests have shown promising results.

Another study is proposed. The test-site is located in a tropical rain country: the oriental piedmont of Andes mountain in Venezuela. This large area is typical of tropical deforestation in a pioneer front.