At first appearance, a simple mark on paper may look like just that. However, such marks actually contain a variety of visual properties. The concept of visual variables was introduced by Bertin (1967), and is divided into two groups: planar and retinal. The planar variables represent the location of the mark. Because the image is on a flat surface, the mark will always have an x and y value to show its location. The retinal variables include size, value, texture, color, orientation, and shape.
Some of the basic visual variables may be further broken down into sub variables. Texture, for instance, may be described by the characteristics associated with patterns which include: the basic element of the pattern, spacing, value of lightness or darkness (MacEachren 1982), and arrangement of repeated elements within the pattern (Robinson and Petchenik 1976). Color may be described using the three characteristics of hue, value and saturation which refer respectively to the dominant wavelength on the light spectrum, the color’s lightness or darkness, and the color’s purity (Munsell 1905).
There exist almost unlimited combinations of these visual variables which means that visual marks may be used to represent a variety of related and unrelated properties. Bertin describes using different levels of organization to relate variable differences. On the nominal level is an associative organization in which all variables except size and brightness may be grouped together in an image. In other words, differences among variables may be overlooked except for size and brightness. Also on the nominal level is a selective organization in which a viewer may select one variable and perceive locations of marks based on that variable while ignoring others. Although Bertin suggests that this applies to all variables except for shape, Green (1999) proposes that by expanding Bertin’s domain of shapes (basic triangles, squares, etc.), there exist shape differences that support selectivity. Bertin also describes an ordered organization in which all variables except color, orientation and shape may be used to show different values. In other words, higher or lower values of a variable represent similar values of the data. According to Green, the color characteristic of hue may also be ordered, within a specified range on the color continuum. Finally, Bertin describes a quantitative organization in which there is a direct relationship between the ratio of the variable values and the ratio of the data values being represented. This can only be applied to size and the planar variables, although according to Green, this may also apply to brightness, if scaled.
The concept of visual variables as described by Bertin has since been modified and expanded. Such variables as focus which employs visual sharpness as a depth cue and photo-realism have been discussed (McGranaghan 1993). Other variables such as flicker which involves frequency and phase of blinking marks, and disparity, or left and right views of the same image, have also been proposed (Green 1999). The ability to effectively use these variables depends greatly on the capabilities of computer graphics systems which are used.
With
the advancement of computing technology, static images are less restricted to
paper. The role of viewing images
on computer screens has grown in prominence since Bertin brought forth the
concept of visual variables. Although
Bertin then acknowledged that there would be changes introduced with technology,
he limited his theory to static images. Nonetheless,
technological developments have encouraged and supported the discussion of
additional visual variables, which may even be assigned dynamic properties.
2.1.1
Dynamic Visual Variables
Dynamic visual variables are not limited to the static properties available to a mark on paper. Use of dynamic variables usually implies viewing on screen through film, video or computers. The dynamic visual variables which are discussed here involve animation and user interaction.
There are a variety of variables associated with animation which imply movement of some sort. First is the variable motion in which a visual mark has the ability to move. This implies the characteristics of velocity and direction to describe the movement. Next are sets of variables involving the use of time. Variables associated with time allow for the creation of dynamic marks which show visual changes over time. Such changes emphasize either location or attribute or else show movement. Emphasizing location or attribute is commonly done with the display of static images which have some kind of constant or repetitive action such as symbols which flash.
Showing movement involves three types of actions: re-expressions, time series, and fly-bys (DiBiase et al. 1992). Re-expressions involve making a change in data being represented, such as switching from one static image in time to another representing a different period in time, not necessarily chronological but with different data. Time series incorporate a sequence of static images which change over a period of time. To do so implies the variables of duration (of each image), rate of change (between images), and order (of images). This results in the background being held steady while some kind of change through time occurs to the image on top. A fly-by on the other hand does not show change in time but rather, through a series of static images, shows a change in the viewpoint. MacEachren (1995) adds the dynamic variables of display date, frequency and synchronization. Display date refers to when a change in data occurs at a particular location. Frequency takes into account the number of individual states of data per one unit of time and between changes. Synchronization refers to a correspondence in time between two or more series of data. The use of these dynamic variables facilitate the ability to involve viewer interaction.
Interaction implies the ability of
the viewer to become a user
who may actually modify a visual display. With
a regular static image or a pre-programmed animated scene, the viewer is passive
and may not make changes to the image. Interactive
environments permit a hands on experience between the user and the image
allowing the user to achieve a ‘feel’ for the data (McGranaghan 1993).
Such interaction between user and data is a key aspect of information
visualization.
The study of information visualization is based on the ability to use visual tools to assist in the exploration of data. This allows in-depth insight into data which enables the development of new questions and hypotheses based on that data (DeFanti and Brown 1989; MacEachren and Monmonier 1992). Computing ability has had a significant impact on the advancement of information visualization with computers serving as fundamental tools. The computer assists in creating multiple images of data which then allow for human interaction. This is particularly useful with cartographic images. Cartographic images on a computer screen may be manipulated, moved and changed to allow for real time visual interaction with the map which is not typically possible with paper maps. Geographic visualization therefore offers significant possibilities through the combination of information, data and maps.
Geographic visualization has been
defined within the framework of information visualization as emphasizing the
development of visual methods which further the analysis and exploration of
georeferenced information (MacEachren 1998).
It is looked upon as a means which will not only look for solutions to
questions but prompt the researcher to develop new questions.
2.2.1
Information Visualization and Cartography
In promoting visualization as a concentration in geography MacEachren and Kraak (1997) state, in their review of the goals set by the International Cartographic Association Commission on Visualization, that “Cartography has always been about visualization.” These goals include focusing on mapping systems which facilitate multiple perspectives, modeling the visualization of spatial-temporal process information, studying the potential of three-dimensional displays, and investigating alternative computer interface methodologies.
With the treatment of a map as a tool of visualization rather than a simple display of information, cartography is perceived as exploratory, leading to insight. This has prompted a shift in cartographic research from attempting to present the optimal map to determining methods of spatial data abstraction which prompt pattern identification through visualization which should produce multiple perspectives of the same data to be explored (MacEachren and Ganter 1990). MacEachren and Ganter define visualization as an act of cognition, or the “human ability to develop mental images, often of relationships that have no visible form.” They state that the most important role for cartographic visualization is to take schematic pieces of information and use them to prompt mental visualization of spatial patterns and relationships. They declare that for cartographic visualization to be successful, user interaction must be predominant. Visualization systems should therefore allow users to interact with the data in a variety of ways including creating new viewpoints and changing data conditions. The objective of such visualization is to encourage the “widest latitude in visualizing ‘what-if’ scenarios with the data” (Campbell and Egbert 1990).
To date, several interactive software tools have been developed which allow for visual exploration of spatially referenced data. These programs make possible the creation of maps depicting data chosen from a variety of options in which the user typically may easily change the selected data to create new maps.
One interactive mapping tool, DDViewer 3.0, was created at the Consortium for International Earth Science Information Network to allow demographic data to be downloaded from the internet and, using Java technology, process and display the data on a desktop computer (Sovik 1997). Originally developed to combat slow server time, DDViewer’s strength is its ability to easily remap data according to the user’s selection. It works by allowing a user to select a region of interest on a map and then control the appearance of the map by manipulating colors, text, and legends. The interactivity of DDViewer focuses on map development and presentation based on demographic data. Although the system reached version 3.0, it has not been updated since 1997.
Another tool for cartographic visualization, discussed by Dykes (1999), focuses on interactive graphics to create dynamic maps. This system is based on a scripting language which is used to build interfaces for and between programs. In this case it has been used to develop a system for mapping spatial data for use in exploratory analysis with the objective of revealing information in a spatial data set that is not readily visible in a static map view. One feature within this system is the ability to have graphics or tables pop up on screen when moving the mouse over a region. Another feature allows for the creation of cartograms alongside a regular map. A further dynamic feature allows a map to change colors as a cursor is placed over a specific region. For instance, in a region divided by localities, placing a cursor over one locality will alter the colors of the surrounding areas based on their relationship to the central one. Finally, this program allows for a variety of statistical data such as spatial autocorrelation and covariance to be presented in relation to a map on screen. A major limitation of this software, as discussed by Dykes, is a decrease in computation time as the size of the data increases. Increasing the speed would take away from the interactive options available to the user.
Descartes is another software tool, developed by Andrienko and Andrienko (1999) to support visual exploration of spatially referenced data for use by anyone (not just cartographers). Their goal in developing this system was to promote exploratory data analysis by allowing highly interactive, dynamic data displays as an extension of traditional cartographic displays. Descartes’ main features are that, given data, it creates an automated presentation of the data on maps and then, allows the user to interactively manipulate those maps. Once a map is displayed, the user may select pieces of data or certain parameters as presented on the screen to change the map in real time. Descartes appears to be a powerful system with many tools accessible to the user. Upon choosing what map to display, the map will open in a separate window showing also a legend, zooming tools, layer options (themes include physical and attribute options), and controls which affect the visual properties of the data. The maps are also sensitive to mouse movement in that moving the mouse over a section of the map will highlight that area as well as the same area in other maps if open.
In supporting the theory of their interactive system, Andrienko and Andrienko discuss three essential techniques of manipulating graphical displays of data: focusing individual views, linking multiple views, and arranging many views. Focusing techniques deal with selecting the subsets of data, variables to be viewed, and the layout on screen of the data. Each individual view only shows partial information of the data. By linking together and integrating individual views, a multiple view scenario is created which, when arranged in a logical order, then promotes comparisons.
A final tool discussed here for creating interactive animated maps is MapTime, a software program developed as part of a Master’s thesis at the University of Kansas by Stephen Yoder in 1996. The current download (http://www.geog.ukans.edu/reso-soft.htm [Accessed March 22, 2001]) is a version which was revised in 2000 in preparation for distribution by the association of American Geographer’s Microcomputer Specialty Group (Slocum et al. unpublished). Maptime uses proportional circles to map data at fixed point locations. It allows the user to interact with time series data through animation, small multiples (displaying various temporal elements alongside each other) and change maps (static maps showing change over time) (Slocum et al. unpublished). It also focuses on the ability to create animations of data over time. All animation relates to change in the circles. MapTime is geared towards the actual data; it shows the shape of the map as a backdrop and does not allow layers to be placed on the map. It gives the user a high level of control over the data through user interaction in spite of the flat image maps it produces. In discussing the limitations of MapTime, Slocum et al. note that the area of circles tend to distract from their locations. They particularly note a need for a variety of symbols to be used besides circles and a need for stronger graphical displays. Finally they consider a major weakness of the software to be the lack of ability to map greater than one variable or phenomenon at a time.
These
systems, while shown to be effective interactive tools for their individual
purposes, are nonetheless all limited to a traditional two-dimensional mapping
environment.
According to MacEachren et al. (1999a), virtual reality (VR) creates a fuller experience in visual perception as compared to traditional cartography. In relating geography to VR, MacEachren et al. present what they call the four I’s of geographic VR”: interactivity, immersion, information intensity, and intelligence. Interactivity concerns the user ability to change the viewpoint and move within the environment whereas immersion refers to the actual sensation of being inside the virtual environment. Information intensity concerns the level of detail of the objects inside the virtual environment and intelligence considers the level at which those objects display context-sensitive behavior.
In discussing types of geospatial virtual environments, MacEachren et al. distinguish between actual immersive three-dimensional environments from non-immersive ones. Immersive 3D environments are those in which the viewer appears to be inside the virtual world, such as when using goggles.
MacEachren et al. believe that a semi-immersive world in which the environment contains an abstract world will be more effective than a non-immersive world. This type of environment falls into their further division of 3D environments as those that present the tangible world versus those that display abstract and non-visible aspects of the world. Such abstract worlds are used particularly to explore statistical relationships in geospatial information. Choices concerning the abstractness of representation in location, time and attributes are particularly relevant to a geospatial virtual environment. However, MacEachren et al. admit to knowing very little about how abstract representations in multiple dimensions or even how each factor of multi-dimensional viewing environments are understood. They feel that there is a major challenge in successfully integrating abstract and iconic forms of representation.
Studies using immersion in virtual environments have determined that direct and immersive interaction with spatial data in a human-computer interface may be achieved by using VR techniques. Neves et al. (1997) created a “Virtual GIS Room” to integrate what they refer to as the three cognitive spaces (haptic, pictorial and transperceptual) for interacting with and analyzing spatial data. They conclude in their study that such a virtual immersive environment does improve user interaction with data. Fairbairn and Parsley (1997) stress an emphasis on the participant who does not receive data passively but rather actively initiates behavior to dynamically explore a data set. Such techniques may be used to assist in decision making. In fact, according to Pitt and Nassauer (1992), successful systems for data visualization invite even unsophisticated users to make changes to the viewing environment.
Besides user interaction, animation may be incorporated into virtual environments. Animation may be applied in different ways which include showing change over time, showing objects moving within the world, or having the whole image move across the screen (as if the viewer is moving) to create a fly-by.
In
his comparison of the animation of time
vs. the animation of space, Dorling
(1992) suggests that a map held in time which shows continuous movement is more
effective to the viewer than a map held in space which animates time by having
images appear and disappear or change colors with the progression of time.
In other words, Dorling is suggesting that animation should have a
natural feel to it which may be accomplished by the incorporation of continuous
types of movement. Dorling believes
that not only should time be frozen to animate space effectively, but that
change in time may be shown more effectively without using time to show it.
Monmonier (1992) agrees, stating that animated timed sequences may pass
by too quickly and are therefore too complex for effective cognitive processing
of what is being displayed. Dorling
concludes that movement has been determined to be the most important depth cue
in bringing 3D objects to life in visual imagination.
Using animated objects in a time-static map may therefore create
effective cartographic virtual worlds.
Creating a virtual environment which includes both user interaction and animation may be particularly applicable to demographic data and mapping. Historically, many techniques have been developed to display demographic data. Visualizing population on a map has proven to be a challenge; many maps have certain aspects effectively represented, yet others have features which are misleading (Dorling 1993). Dorling (1993) discusses computerized cartographic techniques for studying census data to demonstrate the value of the spatial nature of such data. Dorling observes that as the physical terrain is shown with more reality, there is less space to show the social reality. Dasymetric mapping, for example, entails shading only inhabited areas of a map, which has a tendency to bring the focus of the viewer to those areas that are less inhabited, such as parks or isolated regions. Another map design, using equal population grids, results in non uniform squares which are distorted to represent the same population per square. Dorling proposes that raw data may be transformed into meaningful scales. He presents an example of an equal population cartogram which uses circles, rather than the distorted squares of the equal population grid, to show population and therefore minimizes visual bias to show a more socially just map. Dorling feels that the enormous amount of demographic data available means that such data must be approached with efficiency to expand our knowledge of society.
Another method which has been analyzed for displaying population density is continuous surface representations. Such representations transform the representation of population density into a terrain landscape. In exploring such models, Wood et al. (1999) suggest that the new terrain provides “interesting analytical and visual insights into population distributions.” Wood et al. discuss a variety of data transformations to show assorted population landscapes. While their maps seem to contribute to analysis of urban population, this type of map may not be suitable for populations on alternative physical landscapes.
In
attempting to overcome such limitations, MacEachren and DiBiase (1991) developed
a system of animated mapping using ‘chorodots’.
In their system, they combine the concepts of a choropleth map with a dot
map to show the temporal change in population data.
Their result is a system where change in the data is more visible than a
typical animated dot map which presents dots popping up and disappearing.
Each
of these mapping formats have their strengths as static maps.
Creating 3D demographic maps with animation and user interaction
generates alternative factors to mapping population. Although dynamic maps take their basic principles from static
maps (Monmonier 1996), a fresh outlook contributes to dealing with the
challenges introduced by dynamic maps.
Among challenges in creating virtual environments for geographic visualization is determining a balance between realism and abstraction (MacEachren et al. 1999b). The characteristics of a virtual reality environment have been shown to be well taken advantage of even when the world presented maintains a high level of abstractness (Dykes et al. 1999). The properties of the symbols used within a virtual environment are a major consideration in the development of such environments.
In keeping with cartographic convention, a thematic map designer’s role is to present a symbology which allows the average person to understand the most complex symbols on a map (Dent 1985). For information to be perceived with just a glance, Bertin (1967) sets a limit of three variables which may be used to differentiate multiple symbols. Symbols which are appropriate for representing geographic phenomena should be used in an environment which also takes into consideration the limitations of the human eye (Dent 1985). Dent suggests that strong map design means both that symbols are not overloaded with visual variables and that the map itself is not overloaded with symbols, therefore assisting the map viewer’s ability to make effective quantitative and spatial judgments of the map and its data.
Maps are to be designed with the application of similar concepts of symbology. Maps must maximize the content that is displayed using clarity, precision and accuracy to avoid too much complexity (Tufte 1983). Well designed maps are simultaneously powerful and simple so that individual graphical elements effectively reveal as much data as possible without being too complex to decipher (Tufte 1983). Applying multiple elements has become easier to design with current technology. This may lead to intricate and detailed maps which, while visually appealing, are overloaded with data and serve no purpose other than to show off the capabilities of a new technology (Tufte 1983).
In considering symbology in a 3D environment, Kraak (1994) asks whether the combination of graphical variables with 3D depth cues are still effective in a virtual environment. In focusing on the placement of objects in space, Kjelldahl and Prime (1995) determine that cues such as a bottom surface and a horizon do improve the ability to recognize the placement of objects. But they did not actually study the effect on the visual properties of the objects in space.
There
are few symbolization or design guidelines that exist for designing dynamic maps
which are used for visualization and exploration (MacEachren and Monmonier
1992). In particular, there is a
need for developing techniques for evaluating the effectiveness of symbols used
in virtual reality since it is not known how the average map user will interpret
traditional symbols when used in a dynamic mapping environment (Slocum 1999).
Such issues lead to Slocum’s (1999) call for cartographers to be in the
forefront of creating a symbology which is less abstract than traditional
symbols. This connects to Rhyne’s (1997) more general call for
cartographers to participate in the development of desktop visual display tools
to support the visualization of geographic data on the World Wide Web.
True virtual reality immerses a person into a 3D scene using goggles or other simulating devices and tools. Though a 3D environment on a computer screen allows for virtual reality type movements, the limitations of viewing a scene on a flat screen means the virtual reality environment is not truly immersive.
There are a variety of coding formats that have been or are being developed to display virtual reality type environments on the screen. These include Java3D, VRML, Extensible Three Dimensions (X3D), and Moving Picture Experts Group 4 (MPEG-4). These formats were conceived of in response to the growth of common access to the internet and were thus created to display images over the World Wide Web. To view these images, special browsers have been developed to display the images and allow for viewer navigation.
The most commonly used browsers are plug-in attachments to be used with Netscape Navigator® or Microsoft Internet Explorer®. The need for such a browser, however creates the problem that an internet user must have already installed, or be willing to install, such a browser. Furthermore, the ability to create web pages with 3D content is not readily accessible to the average person. Either the software is difficult to use, or the coding language must be learned. Nonetheless, a growing number of 3D formats, 3D web pages, and 3D exposure may be a sign of more widespread use of 3D on the web. In support of web-based 3D environments, Adobe Systems Inc. recently announced the creation and future release of a major new tool, Adobe Atmosphere, for the creation of interactive 3D web sites (Adobe 2001).
One format commonly used for displaying web-based 3D graphics is VRML. VRML is a coding format used to represent geometric objects and behaviors associated with those objects. These behaviors include the ability to react to change in time to create animation, or the ability to respond to user interaction. VRML is actually neither virtual reality nor a modeling language but rather a 3D interchange format similar to Hyper Text Markup Language (HTML) which integrates three-dimensions, text, and multimedia to create 3D scenes for viewing on the World Wide Web (Carey and Bell 1999).
VRML was first created in 1994, and a second version, VRML 2.0, was released in 1997. It is currently supported by the Web3D Consortium which was created to provide a forum for 3D open standards and applications on the web (Web3D Consortium 1999a). They have played a major role in the development of VRML although the Consortium is open to all internet related 3D technologies. Currently, they are working to create a new format of VRML, known as X3D, or Extensible 3D, which has a coding format based on XML (Extensible Markup Language). XML is being used as a common coding standard for web technologies (Web3D Consortium 1999b).
A 3D scene in VRML is known as a world. VRML worlds and the objects within them are coded in a “hierarchical format,” (Carey and Bell 1999) sometimes called a “scene graph.” Table 2.1 and Figure 2.1 show some basic components of a VRML world and are followed by a brief description of their contents. Shown are examples of some of the coding used in the experiment described in this thesis as well as available in the current International Standards Organization (ISO) specifications of VRML.
Each VRML (.wrl) file contains some basic elements to describe the general scene. The beginning of every file, as shown, must contain a line stating what format is being used. All other items following a ‘#’ are comments. The type of navigation to be used in the VRML browser may be defined among the options walk, fly, examine, none, or any. Viewpoints throughout the world may be pre-set. A background may be used to fill in the space not used by objects within the world.
Table
2.1. Example of VRML code
Each entity in a scene is called a node which in turn contains the fields that describe the node’s properties, or data. Transform is a grouping node which sets a coordinate base for all objects within the group which are considered children. The location of the group is then set using translation, its orientation set with rotation. An example of a node, as shown, is a geometric figure such as a box which contains the field for its size to describe the height, width and length of the box. Its appearance is then described, giving it a material to describe the surface using properties of color.
Additional features of VRML include the ability to create a prototype to reuse a part of a scene. A node may also be defined for reuse or to be used in an event. Furthermore, Javascript® may be used in the coding of events.
As
mentioned, behaviors may be associated with the objects in that scene.
These behaviors are based on events which occur in the world such as time
passing or a user clicking on an object. When
an event occurs, the coding then routes
information to make a change in the scene.
In the sample code above, a timer is used to make the box rotate
continuously.
There
is an X3D working group within the Web3D Consortium which is developing an open
source X3D browser using Java3D as their basis since Java3D is an application
programming interface using the Java language to draw 3D graphics (Web3D
Consortium 1999b). Another format
is MPEG-4, an object oriented standard, which incorporates audio and visual
content designed for low bandwidth (Koenen 1999). The powerful nature of MPEG-4 allows it to incorporate VRML
objects, which led to the development of the MPEG-4/Web3D working group within
the Web3D Consortium (Web3D Consortium 2000).
As of March 2001, there were an additional fifteen active working groups
in the Consortium. Among them is
the GeoVRML working group.
The GeoVRML working group was created in 1998 with the goal of developing and recommending practices for representing geographical data using VRML (GeoVRML.org 2000a). Their focus is to allow maps and other 3D terrain models to use geo-referenced data for representation and viewing in a web browser (GeoVRML.org 2000a). To do so, they have created extensions to VRML which increase its usability to display geographic applications. These extensions include metadata, scalability, and the ability to use higher coordinate precision than standard VRML. Furthermore, GeoVRML allows for coordinate systems, animation to be interpolated around such geographic coordinates, and the ability to query the coordinate system. GeoVRML techniques are seen as having the “potential to provide functional and transparent communication between geographical information and 3D Web visualization tools” (Rhyne 1999).
There are ten nodes included in the GeoVRML specification. These nodes are GeoCoordinate, GeoElevationGrid, GeoInline, GeoLocation, GeoLOD, GeoMetadata, GeoOrigin, GeoPositionInterpolator, GeoTouchSensor, and GeoViewpoint. Examples of these nodes and GeoVRML in use may be found at the GeoVRML.org website (GeoVRML.org 2000b).
According
to the GeoVRML.org website, GeoVRML 1.0 was licensed and released in 2000.
This led to the significant inclusion of a GeoVRML export function in the
3D Analyst Extension of ArcView/ArcInfo 8.1®.
In early 2001, GeoVRML was given native support in one of the popular
VRML browsers, Cortona VRML Client (GeoVRML.org 2000c).
A variety of VRML browsers, typically plug-ins for Netscape Navigator® and Microsoft Internet Explorer®, have been developed to display virtual worlds written in VRML. No single plug-in has become a standard though a few have become popular. One particular plug-in stands out because it is widely used, continuously being developed, and is able to read GeoVRML nodes. That plug-in, Cortona VRML Client, developed by Parallel Graphics (Cortona VRML Client 2001), is considered fast, powerful and easy to use (ZDNet 2000). Additionally, it is currently supported with new versions which are being developed and released to increase its usability, particularly with Java® and Javascript®. Cortona VRML Client offers the ability to plan, pan, turn, or roll within the scene. Descriptions of these commands may be found in Appendix A.
Some of the useful features of Cortona VRML Client include the ability to turn the world at any point chosen, go to any point, or restore the last viewpoint used whenever lost. These features are controlled with the mouse, generally by clicking the mouse somewhere in the screen, holding it down, and moving the mouse until the intended movement is reached. VRML browsers are difficult to control however and often, desired movements within a VRML world are difficult to actually achieve (Fairbairn and Parsley 1997).
By combining the ideas of user interaction, animation, symbology, and cartographic convention with a demographic map to create a web-based 3D environment, one may test whether the interactive and dynamic elements of such a cartographic image enhance the ability of the map user to correctly understand data displayed on the map. The use of VRML for map presentation raises many questions and prompts the investigation of issues used within such a VRML world including map representation, navigation and user interaction, and symbol animation.
Map representation is looked at in regards to flat maps versus 3D terrain maps. Navigation and user interaction are looked at to see the extent to which map users take advantage of the navigation tools made available to them. This includes looking at their use of pre-set viewpoints within the VRML worlds along with self created viewpoints. It also includes a look at the length of time required to complete the research task depending on the availability of navigation tools. Symbol animation is looked at by setting some map symbols rotating and others stationary. These issues are all tested by looking at the ability of map users to accurately estimate symbol size based on a reference symbol within a VRML world. Whereas a map user may be able to click on specific locations or objects to call up actual symbol data, this may be impractical when there are large amounts of data displayed.
A series of scenes containing the same map are created for viewing in the VRML environment. Placed atop those maps are groups of cylinders assigned different values and characteristics. Some of the cylinders rotate while others are stationary. Half of the worlds contain the map as a flat image, and half show the map with 3D terrain. One world of each type of map is then presented in each of three levels of user interaction. The first level allows no interaction by the map viewer. The second level displays a pre-programmed interaction in the form of a circular fly-by. The third level gives the map user the opportunity to make use of the navigation tools available in the VRML viewing environment.
By analyzing estimates made by participants in the experiment, along with their actions, a determination will be made as to whether any of these individual factors or combinations of factors allow for significantly accurate estimates of the data represented by the cylinders.