Geographical information systems (GIS), expert systems (ES) and spatial decision support systems (SDSS) are becoming important tools for supporting managers and planners in making decisions for resource and environmental management. In recent years, attention has turned to the integration of existing GIS, ES systems and other problem-solving techniques to develop more powerful SDSS systems. Such systems should lead to significant competitive advantages, such as cost savings, the ability to couple analytical modelling with heuristic reasoning, and automated explanation facilities for interpreting and justifying the results of modelling studies. However, early attempts have also demonstrated a number of drawbacks, such as user unfriendliness, lack of flexible model management capabilities and poor adaptation to users' needs. To try and overcome some of these problems, this research establishes a new approach to the development of spatial decision support systems within an integrated framework of GIS, spatial modelling and expert systems techniques and technologies. In this approach, knowledge-based techniques are introduced into the design of knowledge-based spatial decision support systems (KBSDSS), with emphasis on the design of a representation scheme based on spatial influence diagrams and mechanisms for structuring, representing and formulating spatial problems, together with automation of the solution process. Spatial influence diagrams are graphic knowledge representations for resource and environmental problems, consisting of information about all problem variables or parameters and their relationships. They can be seen as spatial analogues of influence diagrams developed for decision analysis. However, spatial influence diagrams are deterministic cases of influence diagrams without decision components. Algorithms are developed to formulate and evaluate spatial influence diagrams using domain-specific knowledge in the system to represent and evaluate specific spatial problems according to the decision marker's perspective.