This dissertation incorporates time into base geographic data, including boundary, elevation and land use/land cover. It first examines the characteristics of time in base geographic data. Although changes can be discrete, stepwise or continuous, all are simplified as discrete. Currently, there are two major spatial representations, objects and fields, but there is not an existing temporal approach that is suitable for both. Since space is represented differently in existing geographic information systems (GIS) and there are similarities between space and time, this dissertation argues time should be represented in different approaches too. Parallel to the object-field dichotomy, this dissertation introduces a feature-based temporal model for objects and suggests the layer-based approach for fields. The layer-based approach can be further classified into current-state-with-changes (CSC) and sequential snapshots. In the feature-based temporal model, a feature can have multiple temporal spaces and themes. All are functions of time, and can be simplified as constants during time intervals when changes are treated as discrete. This makes it possible to model a feature's history with several time intervals and corresponding states. In the CSC approach, discrete change in fields is modeled using the current state and a series of spatio-temporal changes. The sequential snapshots approach is recommended for fields where space is frequently fragmented through time. However, there is no strict distinction between CSC and sequential snapshots. The feature-based temporal model can be implemented in an object-relational (OR) database, which supports abstract data types. In an OR feature table, a record can model a temporal feature. Based on common spatial and temporal reference systems, a prototype temporal GIS is developed to integrate different temporal approaches. The temporal system is developed from scratch because existing GIS limits the advantages of the OR schema. The prototype temporal GIS not only integrates the three temporal approaches, but also can work with other models. In conclusion, this dissertation makes a step in moving traditional static GIS towards temporal GIS, and contributes to geographic representations in three of the nine research objectives identified by University Consortium for Geographic Information Science.