GIS （or Geographical Information System） works well on spatial data manage but has shortage in spatial analysis. Many spatial decision-making problems are dependent on spatial analysis so it is essential to extend the function of GIS analysis. Spatial decision support system has become an important research area in GIS and the traditional methods rely on analysis models while focus little on the importance of knowledge. Model-based decision support system presumes that the real world is structured and can be well simulated. But the fact is that most of the decision problems are ill-structured and cannot do without domain knowledge. How to acquire sufficient decision-related knowledge and well organize it imposes a great challenge both to spatial information science and AI （artificial intelligence）. In this paper

an object-oriented spatial knowledge representation is proposed. The foundation of this method is based on knowledge analysis and knowledge classification. The fact that domain knowledge has layer structure should be considered when knowledge base is designed. The layer structure and organization of domain knowledge class and classification model know are given in this article. The whole knowledge used can be classified into decision area class

decision area sub-class

knowledge class of decision area sub-class

sub-knowledge class and knowledge unit according to the layer structure of knowledge from the top to bottom. Each knowledge class encapsulates attributes

methods and some rules. The detail of spatial reasoning controlled by knowledge unit is illustrated. Based on our method proposed

an application in intelligent spatial decision support system in agriculture is introduced. The function of GIS offers basic spatial data and facts that are the input of spatial decision-making module. In this way GIS and knowledge-based reasoning can be integrated.