With an emergence of geographic information systems (GIS), spatial data have been represented by a traditional layer-based data model using geometry and thematic attributes. In the real world, geographic phenomena also possess relationships among them and temporal attributes. Since late 1980, a feature concept has been developed that uses three dimensions for representation: space, theme, and time. A feature approach in developing GIS has the potential for more accurate representation of geographic reality because the features correspond to the basic level of human cognitive observation. The aim of this study is to derive a feature data model and to develop a feature-oriented GIS (FOGIS). The objectives are 1) to apply the object-oriented model and feature concepts to GIS and construct a feature data model, 2) to develop a spatial data input procedure using an interactive rule-based expert system, 3) to apply the input procedure for spatial data generation, 4) to design and implement a FOGIS architecture, and 5) to demonstrate the utility of FOGIS by entering, storing, analyzing, and displaying spatial data for multiple types of geographic features. A feature data model was derived from the feature concept. Classes in the feature data model were derived using abstraction methods in Spatial Data Transfer Standard (SDTS) and key concepts of object-orientation. A prototype system was then developed using the feature data model. This system consists of an input module, an import/export module, a feature database, and a display module. Specifically, the input module includes the expert system to provide a guideline to users for producing land use/cover data through an interactive question-and-answer sequence. The display module can be used to effectively visualize the hierarchical structure of objects in the feature data model. The developed system was applied for several geographic applications. The input module with an expert system was applied for thematic mapping of land use/cover in Athens, Georgia. The FOGIS was applied for the visualization of multi-temporal representations of buildings and coastlines in Camp Lejeune Marine Corps Base, North Carolina. Also, thematic relationships were used for spatial query on the roads in Camp Lejeune area.