The interpretation of remotely sensed images in a spatiotemporal context is becoming a valuable research topic. It helps predicting future trends and behaviors, allowing remotely sensed users to make proactive and knowledge-driven decisions. These decisions are useful for urban sprawl prevention, estimation of changes regarding productivity, and planting status of agricultural products, etc. However, the process of change prediction is usually characterized by several types of imperfection, such as uncertainty, imprecision, and ignorance. Fusion of several decisions about changes helps improve the change prediction process and decrease the associated imperfections. In this paper, we propose to use an adaptive possibility fusion approach to take into account the reliability of each change decision. This reduces the influence of unreliable information and thus enhances the relative weight of reliable information. Decisions about changes are obtained by applying previous works and represented as spatiotemporal trees. These trees are combined to obtain more accurate and complete ones. In addition, an uncertainty propagation module is developed to estimate the uncertainty in the output of the knowledge fusion module from the uncertainty in the inputs. This helps us to identify robust conclusions. The proposed approach is validated using SPOT images representing the Saint-Denis region, capital of Reunion Island. Results show good performances of the proposed approach in predicting change for the urban zone in the Saint-Denis region.