Robert Edsall and Donna Peuquet

Department of Geography

The Pennsylvania State University

University Park, PA 16802

email: edsall@geog.psu.edu / peuquet@geog.psu.edu


The incorporation of time into GIS has introduced challenges in all realms of GIS research, from the development of appropriate data structures and algorithms to analysis and querying methodology and display. For users of any software, however, the user interface is key for the software's productive use. It is through the interface that (potentially complex) requests are expressed and results of those requests are presented. The current paper describes the GUI design for a Temporal GIS. The specific example described here has been implemented as the user interface of TEMPEST, a prototypical Temporal GIS. Emphasis in the discussion is on the unique issues that arise in representing time at the user-interface level, and effective strategies that can be developed .


Data associated with space-time processes have long been difficult to visualize and analyze, at least in a GIS context, without time-based data models and interfaces. This necessary emphasis on process -- analysis of data according to both spatial and temporal components -- has created an imperative in the realm of GIS research for efficient and user-friendly software that deals with questions of time as well as it does with questions of space. A prototype graphical user interface (GUI) has been developed which facilitates analysis of events, patterns, and processes within geo-referenced data. The representation of time in a GIS user interface presents intriguing and unique challenges to the designer, and it is these challenges and their proposed solutions that will be the primary focus of this paper. The analogous representations of locations and attributes are by no means meant to be diminished in importance by this focus; in fact, representation of these other dimensions is an ongoing and parallel research project.


The GUI described herein was designed within the conceptual framework of the Triad model, first suggested by Peuquet (1994), in which data are stored and presented based on three "dimensions": where (location-based), what (object-based), and when (time-based). The interface is intended to be the front end of the TEMPEST GIS prototype, which, as its name suggests, is being developed specifically for the representation and analysis of spatiotemporal data (Peuquet and Qian, 1996). TEMPEST evaluates queries according to what has been called the "dimensional dominance" of the query (Langran, 1993). The "dominant" dimension of a multidimensional query is that which is most constrained by the query; for example, the query "what features have ever existed at this location?" is most constrained by the specification of location, its dominant dimension.

The design of the interface is driven by this typology of historical or process-driven queries; with each type of question comes different ideas for the integration of the vital temporal component of the data represented and analyzed. The Triad model provides an organizing structure for the development of individual tools for the interface; the creation of various elements of the GUI involved first distinguishing the operation as oriented toward either spatial, temporal, or attribute information, then considering how the operation might be related to the other dimensions.


Because the user of any information system interacts with the system in every way through the user interface, its design is of vital importance to the productive use of the system. In essence, as Frank (1993) and Gould (1993) suggest that, to the user, "the user interface is the system." Howard and MacEachren (1996), in their typology of design for interfaces to geo-referenced data, emphasize that a successful GIS "facilitates creative thinking by allowing a user to change displays quickly and in predictable ways," (p. 62) with a user-friendly, efficient, and useful interface being the key to this success.

Following Gould's (1993) lead, the TEMPEST interface design is focused first on "well-reasoned conceptual design." Since issues of temporal query and display have not been fully addressed in a GIS context, the proposition of a universal interface for GIS (Raper and Rhind, 1990) requires conceptual extensions for spatiotemporal data. These extensions are based on the types of queries which might be made of spatiotemporal data, some of which are outlined by Peuquet (1994). Three classes of temporal queries are suggested: the change in object or feature (e.g. "where was this object two years ago?"), changes in the spatial distribution of an object or set of objects (e.g. "did any land use changes occur in this drainage basin over the last 15 years?"), and temporal relationships between multiple geographic phenomena (e.g. which areas experienced a landslide within one week of a major storm?"). Effort was made in the TEMPEST interface to recall existing design elements in familiar GIS GUIs, drawing parallels between spatial and temporal constructs (Langran, 1992; Parkes and Thrift, 1990). For example, the concept of "size" in a spatial sense is outlined by units of length, whereas in a temporal sense it is outlined in units of duration. The familiar click-drag operation of a mouse to outline a spatial length (say a region on a displayed map) has been translated in the TEMPEST GUI to outline a duration on a graphical timeline.


The general look of the interface is a reflection of TEMPEST's reliance on the Triad model. The screen is divided into three overlappable windows, one for each "vertex" of the triad. Figure 1 shows a typical interface of TEMPEST in action: (a) a "location" window displaying a spatial representation (typically a map), (b) an "object" window which lists not only spatial objects like lakes, thunderstorm cells, and power lines, but also temporal objects like events and conditions, and (c) a "time" window giving graphical information about the temporal component of the data. The most innovative graphical devices, by necessity, are found in the time window. To facilitate understanding of the temporal analysis capabilities of TEMPEST, the tools developed for time-based queries are purposely analogous to those built for the more familiar location- or attribute-based queries. Such features as a magnifying tool for multi-(temporal-)scale queries and an area box tool for temporal searches are borrowed and adapted from traditional GIS tools and operators. The integrative approach of the TEMPEST data model requires that the time-based queries be interdependent with the location- and attribute-based queries, a need recognized by Langran (1993) and Worboys (1994). A typology of these interrelationships was presented in Peuquet and Qian (1996). The location window, thus, is an interactive and dynamic spatial representation with varying detail for varying scale, flexible symbolization schemes, and a dynamic legend. The object window displays a dynamic hierarchical "tree" of objects, customizable based upon such parameters as location, duration, and size. All of the components of the interface, like the components of the Triad model, are cooperative and interdependent.

A fundamental design consideration for any application, GUI or otherwise is, of course, the intended audience : the user. The TEMPEST GIS is designed for not only "public" applications such as the presentataion and communication of multidimensional information, but also "private" applications like the cognitive, abstract formulation of hypotheses through interactive and individual exploration and visualization (MacEachren, 1994). Its ability to facilitate exploration of spatial and temporal data and the formation of hypotheses during and after that exploration is a priority in the interface design (Monmonier, 1990; MacDougall, 1992). Thus, prime consideration was given to the flexibility and customization of the system through the interface such that such exploratory (especially temporal) queries might be straightforward.

Figure 1. The general TEMPEST interface, showing the three basic display windows (clockwise from top), location, object, and display, overlaid on top of the main tool and menu window

To this end, a conceptual distinction has been made between linear and cyclical views of time. The linear view, more familiar to humans in our everyday experience, is that time marches inexorably forward and that changes occur steadily (like aging) or suddenly (like an earthquake) , as observed. The cyclical view, on the other hand, relates directly to conceptually perceived or anticipated temporal patterns (like diurnal wildlife activity). This distinction is manifested in the GUI in the temporal querying tool, a graphical representation of time which can alternate between a "time line" and a "time wheel" at the user's request. The line representation (Fig. 2a) is effective when querying observed raw data, with no presumption or imposition of temporal rhythms or patterns. For example, a query about the gradual change of an undisturbed meadow to old-growth forest land over many decades would likely be linear and continuous (e.g., "from 1950 to 1990, what land cover changes did this acre undergo?"), and best represented graphically by a time line. The wheel representation (Fig. 2b), on the other hand, is useful when querying spatiotemporal data which may have a known or anticipated cyclical nature. The user may specify the period of the cycle represented and choose to query only those dates which correspond to a specified duration within the cycle. For example, suppose a researcher is interested in the variation of rainfall each monsoon season over several years. She would customize her query time wheel to a yearly period and then select those months (or weeks or days) of interest to limit her investigation. This flexibility of investigative methods encourages creative exploration of the databases for the formulation of new hypotheses as well as the proof of existing ones.



Figure 2. The TEMPEST temporal query tool: (a) the time line representation: a user selects a continuous temporal region for query, (b) the time wheel representation: a user selects units of time relative to a preconceived temporal cycle.

The temporal querying tool described above is designed separately from the display date tool, though they are both created to be graphical representations of the temporal variable. The display date is defined to be the real-world time that is represented in the display. The display date is shown in the GUI both graphically and textually in the display date tool (Fig. 3), an interactive scrollable time line. A separate timeline is generated for each spatiotemporal data layer (attribute). These multiple timelines can be manipulated by the user in several ways: (1) They can be controlled independently of one another, allowing a user to inspect the spatial patterns and processes of one attribute while holding the other(s) still in time. This might be of use to a researcher examining the effects over time of a single event, like the diffusion of certain cancers through space after a nuclear power plant accident. (2) The timelines can be "bound" to one another, such that the displayed layers move simultaneously, to inspect for spatial correlation. For example, a meteorologist might wish to bind a cloud-top brightness data layer to one of ultraviolet radiation at the surface in order to examine how these phenomena relate to one another. (3) A user can bind and "offset" timelines such that the data layers move simultaneously, but with a specified time lag. A researcher hypothesizing that landslides occur approximately four days after a specific region experiences a major precipitation event would find this feature of the interface useful.

Figure 3. The display date tools, showing the "rainfall" layer bound to the "fires" layer with no temporal offset.

Among other tools specifically designed for the analysis of temporal data is a time series graph (Fig. 4) showing changes in an attribute over time at a specified location or region. This is a natural extension of capabilities in other systems which shows the change in an attribute over a certain spatial distance at a specified time. The graph, not unlike the display date tool, is scrollable and capable of displaying multiple time series simultaneously. Here, too, the concept of "zooming" with a magnifying tool has been borrowed from standard spatially-based GIS interfaces for multiple temporal-scale inspections of the time series. Time series analysis functions such as Fourier transformations and filtering may be added to future versions of the interface.

Figure 4. The TEMPEST time series graph.


The development of the interface for TEMPEST has served to confirm that "the representation of phenomena in time as well as space is significantly more complex and more difficult than their representation in space alone" (Peuquet, 1994, p.442), but that potentially powerful methods for this representaion can be developed. Our goal was the incorporation and extension of existing analogous query and display interface designs to a GIS interface created to examine and analyze temporal as well as spatial data. Priority was also given to the construction of tools which encourage creative and exploratory thinking by users. Our goal in this research was not only the practical implementation of this interface as the front-end of the TEMPEST Temporal GIS but also the suggestion of general strategies for the representation of time in GIS displays. The specific interface described here was created using a Sun UNIX workstation and the object-oriented interface developers' language Tcl/Tk. Development of this interface is ongoing and will continue to improve with creative input from not only other designers but also, and especially, users of the system.

Acknowledgments. The authors wish to thank Liujian Qian, Elizabeth Wentz, and Alan MacEachren for their assistance and advice in the development and design of the prototype interface. This research is sponsored by the National Science Foundation under Grant FAW NSF 90-27, and by the Environmental Protection Agency, Grant R825195-01-0.


Frank, A.U., 1993. The Use of Geographical Information Systems: The User Interface Is the System. In Medyckyj-Scott, D. and H. Hearnshaw eds. 1993. Human Factors in Geographical Information Systems 3-14. London, UK: Belhaven Press.

Gould, M.D., 1993. Two Views of the User Interface. In Medyckyj-Scott, D. and H. Hearnshaw eds. 1993. Human Factors in Geographical Information Systems 101-110. London, UK: Belhaven Press.

Howard, D. and A.M. MacEachren, 1996. Interface Design for Geographic Visualization: Tools for Representing Reliability. Cartography and Geographic Information Systems, 23(2): 59-77.

Langran, G. 1989, A review of temporal database research and its use in GIS applications. Int. Journal of GIS, 3(3): 215-232.

Langran, G. 1992. Time in Geographic Information Systems. New York: Taylor and Francis.

Langran, G. 1993. Manipulation and Analysis of Temporal Geographic Information. Canadian Conference on GIS, Ottawa. ??

MacDougall, E. B. 1992. Exploratory Analysis, Dynamic Static Visualization, and Geographic Information Systems. Cartography and GIS, 19(4): 237-246.

MacEachren, A.M. 1994. Some Truth With Maps: A Primer on Design and Symbolization. AAG Resource Paper Series, Washington, D.C.

Monmonier, M. 1990. Strategies for the visualization of geographic time series data. Cartographica, 27(1): 30-35.

Parkes, D. and N. Thrift, 1980. Times, Spaces, and Places. New York: John Wiley and Sons.

Peuquet, D. 1994. It's About Time: A Conceptual Framework for the Representation of Temporal Dynamics in GIS. Annals of the American Association of Geographers, 84(3): 441-461.

Peuquet, D. and Duan, N., 1995. An event-based spatiotemporal model (EDSTM) for temporal analysis of geographical data. Int. Journal of GIS, 9(1): 7-24.

Peuquet, D. and Qian, L. 1996. An Integrated Database Design for Temporal GIS, in Proceedings, Seventh International Symposium on Spatial Data Handling, Delft, The Netherlands, International Geographical Union, pp. 2.1-2.11

Raper, J.F, and D.W. Rhind, 1990. UGIX (A): The design of a spatial language interface for a topological vector GIS, in Proceedings, Fourth International Symposium on Spatial Data Handling, 23-27 July, Zurich, Switzerland, International Geographical Union, pp. 405-412.

Worboys, M.F. 1994. Unifying the Spatial and Temporal Components of Geographical Information, in Proceedings, Sixth International Symposium on Spatial Data Handling, Edinburgh, Scotland, International Geographical Union, pp. 505-517.