Conceptualization of complex spatio-temporal data that have both object- and field-like properties, such as hazardous chemical leak data and rainfall et al., has not yet been incorporated into GIS data models. A new approach is proposed in this paper for organizing data about such complex spatio-temporal data in a GIS as a hierarchy of geographic regions, geographic features, geographic processes, and geographic events. In this model, discrete objects are used to show how geographic events and processes progress in space and time, and fields are used to model how regions and features of geographic themes vary in a spatio-temporal frame. We using the hazardous chemical leak data (HCLD) from T=120 s to t=240 s as an example to demonstrate the construction and application of the proposed approach, and designed the algorithms to stores the HCLD, so that attributes about geographic regions, features, processes and events are readily available for information query. Moreover, the spatio-temporal behavior and interaction of geographic events and processes are computed by using the model, which enhances support for spatio-temporal query and analysis of complex spatio-temporal data in geographic information systems.