For details on progress toward developing the research agenda outlined, click here (NEW!)

VISUALIZATION - Cartography for the 21st century

Alan M. MacEachren
Department of Geography
Pennsylvania State University, USA
Chair, ICA Commission on Visualization


the ICA Commission on Visualization (1)


This paper provides a brief overview of the evolution of Geographic Visualization (GVis) and the activities of the ICA Commission on Visualization. Initial research objectives for the Commission are detailed and a draft research agenda being developed is outlined.

Proceedings of the Polish Spatial Information Association conference, May, 1998, Warsaw Poland

Cartography has undergone profound change over the past decade. The change is stimulated by a flood of new georeferenced data combined with new scientific and societal demands and uses for these data along with rapidly evolving geoinformation technologies. These technologies are part of wider developments in information representation and processing with which they are becoming increasingly linked (key among these are: geographic information systems (GIS), visualization in scientific computing (ViSC), and virtual reality (VR). One result of the links among cartography, GIS, and related visual information technologies has been emergence of Geographic Visualization (GVis) as a research focus and as a set of tools that promise to fundamentally change (and enhance) the way scientists and others conceptualize and explore georeferenced data, make decisions critical to society, and learn about the world (MacEachren and Ganter 1990; Taylor 1991). While GVis (like GIS) is an interdisciplinary endeavor, cartography is well positioned to take a lead in advancing the frontier.

Approximately a decade ago, a report to the U.S. National Science Foundation on Visualization in Scientific Computing (ViSC) prompted a dramatic change in the approach of science to visual analysis and visual evidence (McCormick, et al. 1987). That report characterized visualization as a method (and a product) that integrates of the power of digital computers and human vision and directs the result toward facilitating scientific insight.

Within cartography and geography, the ViSC report prompted attention to the role of maps and other visual displays in the process of science (DiBiase 1990) and an integration of the (then largely separate) research streams directed to mapping technology and cognitive cartography (Buttenfield and Ganter 1990; MacEachren and Ganter 1990; Taylor 1991). In 1993, the International Cartographic Association Commission on Map Use formed a working group on visualization to address implications of these developments. A product of the first meeting of that group was a conceptual model of map use (figure 1 - The (cartography)3 perspective on map use) that emphasized an expanded view of use prompted by integration of ViSC principles and cartographic representation methods (MacEachren, 1994). The expanded perspective on map use focused on three facets of change in use: (1) in goals of map use (a shift from information retrieval toward information exploration), (2) in target audience for use (a shift from the general public toward individuals), and (3) in flexibility of use (from inflexible static maps toward highly manipulable dynamic maps).

The implications of this model of an expanded map use space are, perhaps, more apparent if we move to an even more abstract depiction of that space--a 2D matrix that highlights the 8 corners of the cube (figure 2 - Examples of map use at the extremes of map use space). Virtually all cartographic research, from the 50s through the 80s, that was directed to use of maps is confined to one cell of the matrix--use of static maps by individual "average" map readers to retrieve specific information. Thus, not only did we know little about how highly interactive computer interfaces influence the use (and usefulness) of maps (and of other spatial representations), we had not explored the ways in which static paper maps were used by experts as a prompt for visual thinking. As a result, although we could (perhaps with some justification) claim that the roots of scientific visualization are to be found in geographic cartography (see Collins 1993), we were (in 1994) just beginning to branch out from these roots.

Commission on Visualization - initial research foci

In 1995, the International Cartographic Association approved formation of a new Commission on Visualization?a Commission with a term running through fall 1999, the eve of the next century. The Commission's focus is on use of dynamic maps as prompts to thinking (dynamic maps are maps that change in response to user action or to changes in data to which they are linked). A variety of Commission activities over the past three years have prompted an evolution of the (cartography)3 conceptual model of map use. The current model (figure 3 - A revised view of map use space that incorporates DiBiaseís (1990) perspective on the multiple roles of visualization in inquiry) deemphasizes the tension between a communication and visualization approach to cartography by treating GVis as a superset that includes attention to effective presentation, but expands the bounds of geoinformation use well beyond those traditionally considered within the purview of cartographic communication (MacEachren and Kraak 1997). As outlined above, GVis is a new rapidly evolving area of inquiry. While the focus of GVis is certain to evolve, at this stage in development GVis can be defined as a form of information visualization that emphasizes development and assessment of visual methods designed to facilitate the exploration, analysis, synthesis, and presentation of georeferenced information. GVis has a combined emphasis on development of theory, tools, and methods and on understanding how the tools and methods are used to prompt thinking and facilitate decision making.

Initial goals set for the ICA Commission on Visualization over its four year term include pursuing a series of research priority areas that relate to the expanded perspective on map use associated with GVis, a perspective that includes an emphasis on highly interactive tools, used by experts, in pursuit of unknowns. The specific research goals initially identified are listed below, along with examples of research completed thus far (in most cases by Commission members or those affiliated with the Commission through its meetings and publications).

1. Explore implications of the change from a cartography focused on optimal maps toward one that emphasizes multiple perspectives (particularly those multiple perspectives facilitated by direct manipulation interfaces and by hypermedia methods). See Fernandes, et al (1997) and MacEachren, et al (1997) for two perspectives on direct manipulation interfaces and Cartwright (1997) for advances in the integration of multimedia and visualization concepts in the development of multi-perspective tools for exploring geoinformaiton.

2. Develop a conceptual model and associated tools for depicting spatial-temporal process information, with particular attention to links between dynamic mapping and temporal GIS. See Kraak and MacEachren (1994) for a framework through which to explore visualization of spatial data's temporal component, Acevedo and Masuoka (1997) for issues associated with time-series animation of urban growth, Buziek (1997) for consideration of human perception as a factor in animation design, and Edsall and Peuquet (1997) for consideration of temporal legends as controls for manipulating both database queries and visual displays in temporal GIS.

3. Develop a conceptual model and associated tools for the visualization of data quality/reliability information. See Ehlschlaeger, et al (1997) and Davis and Keller (1997) for work on modeling uncertainty and using animation to depict multiple potential realities and Evans (1997) for assessment of user comprehension of visualization and dynamic methods for representing uncertainty.

4. study methods for and implications of linking cartographic visualization tools to GIS. See Rhyne (1997) for a conceptual approach to design of integrated systems, Mitas, et al (1997) for work on the integration of process models, GIS, and dynamic visualization for the study of landscape processes, Qian, et. al (1997) for conceptualization of operations common to GIS and GVis.

5. Investigate the impact of exploratory spatial data analysis (ESDA) methods on scientific thinking and hypothesis formulation along with that of map-based spatial decision support tools on decision making strategies and outcomes?as well as the impact of alternative computer interface design strategies in both contexts. See MacEachren (1995) for a theoretical perspective on the role of ESDA for science, Leitner and Buttenfield (1996) for work on the role of reliability visualization in a decision support context, and Howard and MacEachren (1996) for a conceptual approach to interface/system design for GVis.

6. Study the potential of three-dimensional representation tools and the corresponding implications of both three-dimensional display and the associated general trend toward realism (versus abstraction) in scientific representation. Emphasis here has been on the first component, particularly the application of VRML and related VR technology to georeferenced data. See Fairbairn and Parsley (1997) for an introduction to VRML for mapping and Neves, et al (1997) for consideration of the use of GIS as a metaphor for immersive virtual worlds.

7. Address implications, for our approaches to map design, of the ability to link many representation forms together in hypermedia documents. See Kraak and van Dreil (1997) for a conceptual approach to design of hypermaps, Cartwright (1997) for development of metaphors to support information access in a complex web environment of linked documents, Aschedowne, et al (1997) for an approach to design of virtual atlases that link information through the WWW, and Fabrikant and Buttenfield (1997) for an approach to envisioning large data archives.

Commission on Visualization - A GVis research agenda for the 21st century

Most of the research cited above appeared in a special issue of Computers &;Geoscience (May, 1997) or in the Proceedings of the June, 1997 ICA Conference in Stockholm. Just prior to that conference, the Commission met for three days in Gävle, Sweden to concentrate on developing a comprehensive research agenda designed to set research priorities that will lead GVis into the 21st century. This research agenda is still a work in progress and is presented as a draft outline below. Details will be filled in over the next year.

Research priorities are categorized under four themes that reflect several aspects of visualization as an interaction between humans and computers directed toward exploring and understanding geographic phenomena. These four themes are outlined below. (2)

1. Representation

Perhaps the most obvious research focus for visualization of georeferenced information involves issues of the visual display itself.

1.1 Extending the object of representation

It is clear that the integration of ViSC and VR tools with cartographic tools extends the object(s) of geographic representation in various ways. Some of these ways were identified in initial goals of the commission: the representation of space time phenomena and processes and the representation of data reliability. To this list, a need has been identified to develop methods for the visualization of algorithms used to process spatial data, the visualization of data structures and query processes, and the application of GVis methods to spatialization of information in general (i.e., of non-spatial information).

1.2 Extending forms of representation

As a complement to extending the range of objects to which GVis methods are applied, substantial research effort must be directed to taking advantages of advances in computer graphics that make the new forms of representation possible. Much of the initial GVis research has concentrated on adding dynamic components to traditional 2D and 2.5D representational forms (e.g., animated planimetric maps and flythroughs of perspective views). Priority areas identified for research emphasize two developments, the rapid increase in computer power that makes dynamic manipulation of map parameters possible and the increasing availability of tools for building virtual environments. Research is needed to determine how to integrate methods for dynamic manipulation into GVis tools (e.g., to merge exploratory data analysis methods with map animation) and to integrate VR technology with geographic data and principles of geographic representation (e.g., adding geofunctions to VRML or using immersive VR to explore abstract georeferenced data, such as output from a global climate model).

In addition to research on the technical problems of using new technology to best advantage, research is also needed to consider the implications of new representation forms. Questions here relate to the semiotics of extended representational environments, the relative merits of abstract versus realistic representations, and what the concept of representation means in a virtual world. One component of an approach to addressing these questions involves considering cognitive aspects of visualization tool use (detailed below). In addition, these questions can (and should) be approached from the perspective of the philosophy and sociology of science.

2. Interface design

The Commission has, from the start, directed attention to extending cartographic principles, developed for static maps, into the realm of dynamic maps and related displays. In addition to research focused on what and how we represent georeferenced information, therefore, attention must also be directed to mechanisms we provide that allow users to interact with those representations. Research priorities identified here relate to the following:

2.1 Typology of visualization operations

Cartographers have directed considerable attention to developing typologies of variables used in visual (and tactual and sonic) geographic representation and guidelines about appropriate use of those variables. To develop a coordinated approach to interaction with manipulable maps, however, it is essential that a complementary typology of visualization operations be developed. A step in this direction is a typology proposed by Keller and Keller (1992), but that typology does not focus specifically on visualization of georeferenced information, nor does it consider the need to integrate visualization operations with query operations in the context of GIS-GVis integration (see below).

2.2 Controls for operations

Interactive displays of georeferenced information require development of controls through which users can interact. Although there are many basic tools generally available that can be adapted for use in GVis environments (e.g, buttons, sliders, etc.), there has been no systematic attempt to develop a typology of control forms for manipulation of spatiotemporal information. Developing such a typology could lead to more consistent interfaces for geoinformation processing environments and more logical matches between operations and controls.

2.3 Facilitating information access in complex hyperlinked information archives

As the World Wide Web and related technologies provide a mechanism to link geoinformation in complex ways, there is a critical need for research directed to methods that facilitate navigation through that web of information. Potentially productive approaches involve the development of appropriate metaphors for navigation in these complex information spaces and the spatialization of non-spatial information as a method to identify relationships among information objects.

2.4 Intelligent GeoAgents

An intelligent GeoAgent can be defined as a virtual entity that understands our needs and prefrences related to geographic data access and/or its analysis and representation. Intelligent GeoAgents should act as filters and interpreters of information and contexts. One potential role for intelligent GeoAgents involves embedding cartographic expertise within database objects to be displayed. Each object could be bound with a GeoAgent that evaluates the context of use and the display context within which the object finds itself, and select a display form accordingly. Another role for GeoAgents is as information seekers that search the web for information that meets criteria of a fuzzy query. The GeoAgent would adjust to the information context it finds itself in, adapting the query parameters in response to such things as links found among information objects, density of information in particular locations within attribute space, or simply success at matching the initial query.

2.5 Collaborative visualization

An emerging area of research in ViSC generally is collaborative visualization--the development of environments that facilitate the use of manipulable visual displays for exploration of ideas and/or decision making by two or more individuals (perhaps located at a distance). Tools for collaborative visualization of georeferenced information have considerable potential for use in contexts such as urban planning, environmental management, and scientific interpretation of models of climate or other environmental processes.

3. Database-visualization links

One of the most common arguments for developing visualization tools is that such tools can help us cope with the rapidly increasing volume of information being generated by our information society. This potential can only be realized in the case of GVis, however, if GVis is integrated with other technologies for storage, access and analysis of that georeferenced information. Three key research priorities are identified in this context.

3.1 GIS?GVis integration

Traditional GIS environments have treated visual display as an output, thus an endpoint of a query or analysis process. GVis, on the other hand emphasizes information exploration and hypothesis generation, thus it is as likely to produce queries as be the result of one. For GVis to reach its potential, a new view of both GIS and GVis system design is required. Research must address the range of issues associated with merging the goals and functions of GIS and GVis.

3.2 Spatial data mining?GVis integration

As data volumes continue to increase, methods of "mining" data are being developed to sift through the vast amount of information in a search for interesting patterns and relationships. Visualization has a potential role at all stages of the data mining process (from preprocessing and error identification through selection of information to be mined, to interpretation of results from data mining). For georeferenced information, this integration of GVis in the data mining process will require, not only adaptation of approaches to data mining that recognize the unique properties of spatial information, but the adaptation of GIS tools to facilitate the application of both GVis and data mining methods.

3.3 Generalization and Visualization

Research priorities discussed under this theme were prompted by a joint session of the Commission on Visualization and the Working Group on Generalization that took place in Gävle, Sweden, June, 1997. Many applications of GVis require a facility to change spatial or temporal scale. Dynamic visualization offers several new challenges for research in generalization. One is to develop methods for making effective use of the level of detail node in VRML, in order to achieve an appearance of seamless change in scale--a problem that requires adapting many generalization concepts developed for 2D maps to 3D worlds. Similarly, the emphasis on spatiotemporal information characteristic of many GVis applications creates a need to begin to address issues in spatiotemporal generalization. A third generalization issue needing attention involves the generalization of hyperlinked networks of information. Just as a base map with too much detail can impede visual analysis of thematic information superimposed on that base, a too detailed network of links among information objects can impede visual exploration, searching, or decision making based on that network of information. Methods to generalize these networks are needed. Cartographic generalization is, of course, an active research area in its own right replete with many unanswered questions. One avenue of GVis research with potential to support advances in generalization research involves the use of GVis methods to facilitate understanding of the generalization process.

4. Cognitive aspects of visualization tool use

The promise of visualization is based on an assumption that human vision and cognition has powerful information synthesis and pattern seeking capabilities that can effectively complement the raw information processing power of digital computers. Harnessing this power of vision, however, requires developing a more complete understanding of spatial cognition and perception of visual displays. While we have a solid base of knowledge about perception and cognition as it relates to static paper maps, we know much less about the cognitive and perceptual issues associated with 3D and dynamic displays. Seven priority topics are identified here.

4.1 Cognitive aspects of dynamic representation

Our knowledge of cognition related to dynamic scenes (even real world scenes) is limited. Key questions relate to issues such as the role of animation in understanding process, the implications of various parameters of an animation (e.g., display changes per second, smoothness of transition between frames, etc.) on that understanding, differences in cognitive processing required to interpret temporal versus nontemporal animation, and methods of retaining orientation in flythroughs.

4.2 3D representation and virtuality

As technology makes it possible to create increasingly realistic representations of the geographic environment, it is important to consider the cognitive implications of this realism. Questions to be addressed include the cognitive processes involved in identifying the correspondence between 2D (planimetric) representations and 3D representation, exploring the changing relationship between sign-vehicles and referents as representations become increasingly realistic, and developing and understanding of the integration of visual and sonic information in the context of immersive and non-immersive virtual worlds.

4.3 Schemata, metaphors, and human-computer interaction

Facilitating human-computer interaction requires development of logical approaches to interface design and the creation of appropriate interface controls that allow users to manipulate parameters of the display. Also required, however, is a comprehensive understanding of the cognitive aspects of interaction with the display. Do, for example, different control forms (e.g., a time wheel versus a time line) prompt different knowledge schemata that result in different interpretations of what is seen? A similar question involves the implications of possible interface metaphors for the strategies that a user takes to data exploration and the interpretation of results from that exploration.

4.4 Hypermedia navigation

The web is a complex maze of information in which users frequently are frustrated in their efforts to locate the information they require. Attention to cognitive aspects of wayfinding in this information environment is needed as a complement to research (discussed above) directed to design of interface tools that facilitate information browsing. One potentially fruitful avenue of research is to explore the transferability of conceptual models of wayfinding in real spaces to wayfinding in information spaces. An additional topic to investigate involves strategies used to maintain orientation and context in multidimensional information spaces.

4.5 Expert-novice distinctions

At issue here is the impact of expertise on use of GVis and differences in design strategies that should be developed for GVis users with differing kinds or levels of expertise. At least two forms of expertise must be considered, that in the technology being used and that in the domain of knowledge to which the technology is applied. A better understanding of expert strategies for the application of GVis tools to data exploration or problem solving could be used to design knowledge-based GVis tools that prompt novices to use expert strategies.

4.6 Influence of GVis methods on the scientific process / scientific understanding

There is an implicit assumption behind ViSC that visualization will facilitate science. While anecdotal evidence may seem to support this contention, anecdotal evidence by its nature is generally positive. There is little systematically collected empirical evidence to either support or refute the claim and we know relatively little about how scientists actually use sophisticated visualization tools and methods. Research is required to test the underlying assumption that GVis facilitates science and to develop an understanding of the implications for science of visual geographic representation methods.

4.7 Role of visualization in decision-making

The primary question here is whether GVis tools change how decisions are arrived at and/or the outcome of decisions. Assuming that some changes are produced, it is important to explore the nature of those changes and whether decisions are more consistent or otherwise "better." Additional questions to address include the implications of various components of a GVis environment on decision making (e.g., the kinds of information displays provided, the kinds of interaction allowed, etc.) and the role of data reliability visualization on strategies taken to decision making.

1 This paper discusses the evolution of geographic visualization as an area of research and application affiliated with cartography. Emphasis is placed on work of the International Cartographic Association Commission on Visualization (chair: Alan M. MacEachren; co-chair: Menno-Jan Kraak) and on a draft research agenda being developed by the Commission. The Commission activities, particularly the developing research agenda are joint activities of the full and corresponding members of the Commission. Thus, the Commission as a whole should be considered (and is listed as) a co-author of this paper--but this does not imply that all Commission members agree with ideas presented here.

2 References are omitted from this draft, but that omission should not be interpreted as an indication that no progress has yet been made. In most cases, there is at least a preliminary base upon which to build (in some cases, research already cited above provides this base). A discussion of relevant research associated with each theme will be included in the expanded research agenda document.


return to top of page