Towards Exploratory Visualization of Spatio-Temporal Data - CiteSeerX

6 downloads 49489 Views 204KB Size Report
May 27, 2000 - 3rd AGILE Conference on Geographic Information Science ..... In the comparison mode a bi-directional color scheme is used in the map: ...
3rd AGILE Conference on Geographic Information Science – Helsinki/Espoo, Finland, May 25th – 27th, 2000

Towards Exploratory Visualization of Spatio-Temporal Data

Natalia Andrienko, Gennady Andrienko, and Peter Gatalsky GMD – German National Research Center for Information Technology, Schloss Birlinghoven, Sankt Augustin, D-53754 Germany Tel: +49-2241-142329 Fax: +49-2241-142072 e-mail: [email protected], [email protected] URL http://allanon.gmd.de/and/

Abstract. In the paper we focus on the problem of supporting visual exploration of data having spatial and temporal reference. We suggest some methods and tools based on cartographic visualization of the data. The tools involve a dynamic, highly interactive map display that can change its properties in real time, in particular, perform animation. We seek to advance our tools beyond mere animation towards facilitating exploratory analysis of spatio-temporal data. We diversify our approaches depending on properties of data and the character of their variation in time: changing existence, position, values of thematic attributes etc.

1. Introduction Presenting characteristics of spatial objects or phenomena that vary in time has always been a challenge for thematic cartography. A general approach is map iteration: states of a phenomenon at different moments are represented in a collection of maps arranged in a chronological order. For certain particular kinds of data special presentation techniques were developed, e.g. arrows or bands are traditionally used to represent movement. Modern computers offer new opportunities for visualization of time-dependent spatially referenced data. They enable, in particular, cartographic animation: a screen map represents dynamics of phenomena by rapidly changing its appearance. Animation produces a strong visual effect on a viewer when it demonstrates some rather apparent dependencies and trends like urban growth (see, for instance, Tobler, 1970) or shrinking forest areas. On the other hand, the usefulness of animation for data exploration is doubtful. Map iteration is still a more valuable tool of visual analysis. Thus, it is much more appropriate for comparison of states of a phenomenon at different time moments and detecting changes that occurred between two moments. There is a potential for further enhancement of effectiveness of the map iteration technique by proceeding from static map images to highly interactive displays on computer screens, see (Andrienko and Andrienko 1999) for a description of a number of map manipulation techniques facilitating data exploration. Still, the iteration technique has an evident limitation: it is hardly suitable for long time series requiring large number of maps. Currently a number of researchers work on development of map-based software tools for visual exploration of spatio-temporal data. Monmonier (1990) proposed temporal focusing and temporal brushing techniques. Kraak et al. (1997) suggest so called “active legends”. Such a legend provides an information about time reference of the data presented in the map and allows the user to control the animation. Harrower et al (1999) advocate including of two models of time in the active legend. The first model is a linear representation of time. It is useful for selecting subintervals on the time axis. The second model treats time as cyclic. On its basis, for example, a time of the day may be chosen, and only the data referring to this time will be included in animation. In such a way the user can ignore daily variations and consider a longer-term development. It can be noted that all these works are mostly concerned with selection of time moment(s) to be represented in the map. In our research we consider various aspects that can play role in visual exploration of spatio-temporal data. Among them, besides different animation modes and techniques of focusing on time moments or

137

3rd AGILE Conference on Geographic Information Science – Helsinki/Espoo, Finland, May 25th – 27th, 2000

intervals, are the methods used for presentation of the data in the map, user-map interaction, alteration of presentation properties in a way productive for analysis, various transformations of data etc. We realize that diverse tools are needed depending on which aspects of spatial phenomena vary in time: existence, spatial location, geometry (shape and size), or thematic (attribute) data. In the following section we summarize possible analytical tasks and questions that arise when an analyst explores time-referred data of various kinds. In section 3 we deal with interactive features of map display, visualization methods, and display time manipulation controls that we developed. Then we give several examples of application of our techniques to different kinds of spatio-temporal data. In section 4 we describe techniques for visualization of data about events that can be treated as instant, e.g. records of occurrence of earthquakes or of observation of plant and animal species. The next section touches upon visual exploration of movement of objects in geographic space on the example of seasonal migration of birds. Section 6 describes techniques and tools suggested for exploration of temporal variation of thematic data referring to area objects.

2. Summary of analysis tasks Exploratory analysis tasks can be classified according to two dimensions. One of them reflects which temporal characteristic is in focus of exploration, e.g. existence, spatial location, thematic data, and so on. The second dimension concerns whether the analyst is interested to see the state of the data at some moment of time (snapshot), or how data changed at a moment t2 as compared to some another moment t1, or what happened during an interval [t1, t2]. Questions related to a time interval can be divided to those concerning data dynamics (i.e. the process of development) and those concerning summary characteristics of the data on the interval. Table 1 presents our classification of questions that an analyst may pose in the course of exploration of spatio-temporal data. Table 1. Shape and size

Thematic data

What shapes/sizes had the objects at t?

What were values of an attribute at t? How were they spatially distributed?

What objects Where/ how far remained, did each object move? appeared, died? How did the spatial pattern change?

What is the difference between shapes/sizes at t1 and t2?

What is the difference between values and their distribution at t1 and at t2?

What objects How did the existed, objects move? (trajectory) appeared, died during [t1, t2]?

How often did the objects change? How much?

What are average (minimum, maximum, dominant) values on [t1, t2]?

How did the shapes/sizes change?

How did the values and their spatial distribution develop in time?

Existence

Single moment t

Two moments t1 and t2

Interval [t1,t2]

Instant events Durable objects What events What objects Where was each object at t? happened and existed and where? where?

What is the difference (in number, kind, spatial pattern) of events between t1 and t2?

What events happened during (summary) [t1,t2]?

Interval [t1,t2]

Spatial location

How did the number, kind, spatial pattern of events/objects change in (progress) time?

How fast did the objects move? Did they meet? How did the speed change?

When did maximum changes occur? Were there still periods? Is there any temporal trend? Was (where was) the development monotonous /periodic?

138

3rd AGILE Conference on Geographic Information Science – Helsinki/Espoo, Finland, May 25th – 27th, 2000

By the moment we haven’t yet embrace all these cases in our work. We have developed some techniques and tools for exploring data about instant events, moving objects, and changing thematic data. In the next section we describe the interactive controls utilized for the exploration.

3. Time controls and dynamic map display Interactive exploration of data is supported by a specially designed collection of widgets further referred to as time manager (see Figure 1). The time manager is connected to a dynamic map display. It allows the user to select time moments or intervals the data presented in the map to refer to. For selecting an interval, the user specifies its starting moment and length. To select a moment, the length of the interval should be set to one time unit.

Figure 1. Time manager

The current display moment or interval can be shifted forth and back along the time axis either by pressing the buttons “step forth” (“>”) and “step back” (“>” or “>>…”). In this mode the tool iteratively shifts the current moment/interval by a specified number of time units (step). The user may control the speed of animation by varying the parameter “delay”. With the time manager the user may choose what is represented in the map at each display moment: o

the instant view: the map represents the state of the world at the currently selected moment;

o

the interval view: the summary of events, movements, etc. that occurred during the current interval;

o

the overall view: the summary of what occurred from the beginning of the whole period the available data refer to up to the current moment (the end of the current interval).

The selected type of view may be used both statically and dynamically. The static usage means that the user examines the map reflecting some fixed moment or interval. In the dynamic mode the user observes how the map changes as the current moment/interval moves along the time axis. This may be done in the course of either user-controlled or automatic animation. In the following sections we describe examples of use of our tools for exploration of different types of time referenced data.

4. Visualization of instant events. Within the project “Naturdetektive” (see URL http://lo.san-ev.de/natdet) German schoolchildren registered when and where they have seen certain plant or bird. For some plants different states are distinguished: appearing of first leaves, start of blossoming, appearing of fruits. The observations registered were to be presented in map form. One of the requirements to the presentation was to show how participation of children in the project evolved with the time. For this purpose we suggested the dynamic map display manipulated through the time controls described above. The map display is based 139

3rd AGILE Conference on Geographic Information Science – Helsinki/Espoo, Finland, May 25th – 27th, 2000

upon the treatment of the data as instant events. The events are represented by signs (icons) put on the map. Variation of icons is used to encode qualitative information about the events: what plant or animal was observed and, when appropriate, in which state: leaves, flowers, or fruits. An example map with observations can be seen in Figure 2. The interested readers are welcome to view the application in the World Wide Web (WWW) at the URL http://lo.san-ev.de/nd/show00.asp

Figure 2. Visualization of observations of nature. Another application was visualization of data about earthquakes. In this application the events (occurrences of earthquakes) are represented by geometric signs. Characteristics of earthquakes, such as magnitude, depth, or radius, are encoded by the degree of darkness of the signs: darker shades correspond to higher values. The user may interactively choose which of the available characteristics is to be presented. The readers can run an applet showing the data about earthquakes in Europe in 1980-1983 at the URL http://borneo.gmd.de/descartes/java/show1field/eq.html

5. Visualization of spatial movement We applied our tool to an example data set containing telemetric observations of migration of four storks to Africa in autumn 1998 and back to Europe in spring 1999. In our tests we found especially interesting and useful the dynamic interval view, i.e. consideration of the interval view in the course of animation. The interval view in this case shows route fragments made by the objects during the current interval. During the animation (illustrated in Figure 3 by a sequence of screenshots) the routes look like worms crawling on the map. It is not merely fascinating; these "worms" help to study important dynamic characteristics of movement. The length of a "worm" shows the speed of movement. The length being reduced signalizes that the movement of the object slows down, and extension of the worm means that the movement becomes faster. When an object stops its movement and stays for some time in the same place, the corresponding "worm" reduces to one point. It would be practically impossible to do such observations using an ordinary animated presentation playing a sequence of images showing states (in this case positions) at successive moments of time.

140

3rd AGILE Conference on Geographic Information Science – Helsinki/Espoo, Finland, May 25th – 27th, 2000

Figure 3. Snapshots of a dynamic presentation of the behavior of white storks in Africa We invite the readers to explore the movement of the storks as well as test our tools to support this exploration by running the Java applet that is available in the WWW (http://borneo.gmd.de/descartes/java/birds/index.html).

6. Visualization of changing thematic data At the present moment we work on development of techniques to support exploration of dynamics of attribute data, for example, demography or economic indices (Figure 4). For the beginning we consider numeric attributes. We have found that the tools successfully used earlier for data about events and movement are insufficient for exploring attribute data referring to area objects such as countries or districts. Therefore we have designed additional interactive controls enabling various kinds of comparison: with the previous time moment, with a fixed moment, with a selected object, and with a selected number. In the comparison mode a bi-directional color scheme is used in the map: shades of two different colors are used to represent values higher and lower than the reference value. For example, in comparison with the previous time moment the areas where the values of the studied attribute have increased are painted in brown and those with decrease of values are blue. The controls for comparison can be seen to the left of the map in Figure 4. In order to support better investigation of temporal trends, we have added a new data display, timeseries plot, linked to the map. The analyst can select some object(s) using mouse move or click in the map or plot, and the corresponding object(s) will be highlighted in both displays. The plot may be seen in the bottom part of Figure 4. The work described will be further developed within the SPIN project (Spatial Mining for Data of Public Interest) accepted for funding in the IST Programme.

141

3rd AGILE Conference on Geographic Information Science – Helsinki/Espoo, Finland, May 25th – 27th, 2000

Figure 4. Tools for exploration of time-dependent numeric data referring to areas

7. References Andrienko, G. and Andrienko N., 1999. Interactive Maps for Visual Data Exploration, International Journal of Geographical Information Science, 13 (4), pp. 355-374. Andrienko, G., and Gatalsky, P., 1999. Spatio-Temporal Visualization in Naturdetective and beyond, Abstracts of 5th EC-GIS Workshop, 1999, pp. 97-98. Bertin, 1967, 1983, Semiology of Graphics. Diagrams, Networks, Maps, The University of Wisconsin Press, Madison, 1983. Harrower, M., Griffin,A.L., MacEachren, A.M., 1999. Temporal Focusing and Temporal Brushing: Assessing their Impact in Geographic Visualization, Proceedings of 19th International Cartographic Conference, Ottawa, Canada, August 14-21, 1999, 1, pp. 729-738. Kraak, M.J., R. Edsall & A.M. MacEachren, 1997. Cartographic animation and legends for temporal maps: exploration and/or interaction, Proceedings 18th International Cartographic Conference, Stockholm, Sweden, 23-27 June, 1997,1, pp.253-261. Monmonier, M. (1990). Strategies for the visualization of geographic time-series data. Cartographica, 27(1), 30-45. Tobler, W., 1970. Computer movie simulating urban growth in the Detroit region, Economic Geography, Vol. 46, No. 2, 1970, pp.234-240.

142

Suggest Documents