Thematic Navigation in Space and Time

Interdependencies of Spatial, Temporal and Thematic Navigation for Cartographic Visualization - By Example of a European Map of Cultural Heritage; For the SVG Open 2005 conference

Keywords: Interactive SVG Maps, Navigation, Spatio-Temporal Visualization, Information Visualization

Mag. Andreas Neumann
PhD Student
www.carto.net
Institute of Cartography, ETH Hoenggerberg
Zurich
-
Switzerland
neumann@karto.baug.ethz.ch

Biography

Andreas Neumann got a masters degree in Geography/Cartography from Vienna University, in 2001. In 1999 he joined the Cartographic Institute of the Swiss Federal Institute of Technology, first as a system administrator and later as a teaching and research assistant. At the same institute Andreas advised several thesis and student projects in the webmapping domain. Currently he works on a PhD as part of a ETH project called "Distributed Publishing of Cartographic Information on Demand", with the topic "Navigation in Space, Time and Topic". Besides university he is occasionaly doing consulting work in the domain of SVG for Webmapping and Online GIS. From 2001 to 2003 he worked as a GIS specialist at a Zurich based geology consulting company. Andreas was one of the initiators of the SVG.Open 2002 conference in Zurich, the first international SVG developers conference, co-organized by the W3C consortium.


Abstract


As GISystems become more mature and spatial research questions more complex, time is regarded as a crucial component to GIS data models. Cartographic visualization can be one useful method of presenting dynamic spatial data and visualizing changes over time, hence assisting the study of spatio-temporal patterns and phenomena. The rise of interactive mapping applications, together with the growing demand to incorporate time in GIS, however, fuel the need for efficient navigation tools in space, time and topic. This paper summarizes the current results of an ongoing PhD work that contributes in the following areas: Study of the effects that changes in one of the dimensions should trigger on the navigation and visualization of the dependent dimensions, compilation of a taxonomy of spatiotemporal and thematic questions, in order to understand how navigation tools can assist in answering these questions and development of a sophisticated user interface for thematic navigation in space and time by creating a prototype of a european map of cultural heritage. The prototype is implemented on top of SVG for clientside rendering of the interactive GUI and map visualization. The serverside components build on PostgreSQL and the spatial extension PostGIS as well as the scripting language PHP. The visualization includes a map component with the corresponding spatial navigation tools, an interactive timeline for temporal navigation and hierarchic navigation in the various topics, and finally a network diagram that displays relations between persons, events and artwork. All three visualization components are linked to each other. Changing one dimension in any of the spatial, temporal or thematic dimension, has effects on the navigation in the dependent variables and in the visualization results. Highlighting one element in either of the three visualization forms, highlights the corresponding elements in the two other visualization types.

This paper is in larger sections a re-print and refinement of [Neumann2005]. The more detailed section on modeling of time was shortened, navigation options were refined and completed. This papers contains more examples. The technical section (last chapter) was enhanced.


Table of Contents


1. Introduction
     1.1 Motivation
     1.2 Time in GIS - Past Research
     1.3 Terms and Definitions
2. Spatial, Temporal and Thematic Navigation Tools
     2.1 Why Navigation?
     2.2 Navigation Metaphers
     2.3 Spatial Navigation
     2.4 Temporal Navigation
     2.5 Thematic Navigation
3. Interdependencies of Spatial, Temporal and Thematic Navigation
4. Prototype implementation of an interactive 2D map of european cultural heritage
5. Summary and Outlook
Acknowledgements
Bibliography

1. Introduction

1.1 Motivation

Already the early GIS pioneers discovered that “theme”, “location” and “time” are major cornerstones around which one should design a data model: “It can be argued that any observation that does not record explicitly or implicitly all three of the above attributes is not useful information. Equally important, a report of information or data which does not specify all of these attributes will be deficient in its information content and therefore should not be part of a geographic information system.”[Sinton1978]

Today, more than 25 years later, time is still not an integral part of GISystems and incorporating time in standard GISoftware, no matter if it concerns data modeling, visualization or navigation, is still a very complicated task. While considerable research dealt with incorporating time in GIS data models and visualization of dynamic data, graphical user interfaces for temporal navigation are still in their infancy. Furthermore, little is known about the relationships between space, time and topic, especially in the context of navigation. New interactive media require sophisticated navigation tools in order to facilitate the study of spatio-temporal patterns in the context of various interrelated topics. SVG, with its rich support for visualization, interactivity and network connectivity can be used as one technology to implement these navigation tools and interactive mapping applications.

1.2 Time in GIS - Past Research

While time was always an important topic in geography, geography was initially defined as a “science of space” after the quantitative revolution in the 1960s geography. This fact considerably influenced the introduction of the first operational GISystems. Geographers are therefore at least partially responsible for the fact that GIS technology neglected the temporal component for a long time. [Harrower1999] partially explains the primacy of space over time with the close relationship between cartography and geography. “Perhaps if cartography had allied with history, time would be treated with equal or greater import than space.” Another reason is, that many GIS projects are simply at an earlier stage, where multi-temporal datasets and spatio-temporal analysis is not yet a priority. ([Langran1992], p. 5) distinguishes several stages of GIS capabilities: The first stage includes initial data capturing, none to few analysis and primarily static output. The second stage introduces centralized data management and exchange, simple analysis and static modeling, and first interactive graphics. The third stage, at which most projects have not arrived yet, features incremental updates, multi-state analysis and dynamic modelling with animated and fully interactive maps as output. Finally, and probably the most likely cause for the unubiquitousness of spatio-temporal GIS functions, the limited toolset available for spatio-temporal analysis and modelling severely hinders a broad introduction of spatio-temporal methods within GIS. It is quite obvious that time deserves more attention and needs better solutions than just adding it as an additional conventional attribute to spatial datasets.

Geographers in the 60s and 70s mainly studied the change of the society over time. Through his studies on innovation waves [Hägerstrand1952] and human migration [Hägerstrand1970], Hägerstrand can be regarded as one of the fathers of time geography. He was one of the first geographers who declared time as an essential component in his spatio-temporal model. As part of this model, he introduced the space-time pathes of individuals where the map was two-dimensional and the third dimension was time. Stationary people are represented as vertical lines, a location is represented as a vertical tube. If people move, their movement is drawn with a sloped line. The slower the movement, the steeper the line will be. Additionally, a space-time prism can indicate the locations that can be reached in a particular time interval, which obviously depends on the maximum travel speed the individual can accomplish. In the pre-computer graphics time, it was time consuming and expensive to produce such space-time cube illustrations. However, with the rise of new visualization technology and interactivity, researchers revisited this concept for the visualization of lifelines and movements of individuals ([Kraak2003] and [MooreEtAl2003]).

space_time_cube_moore_etal.png

Figure 1: Illustration of the space-time cube with explanations of the individual components/terms, Source:[MooreEtAl2003]

space_time_cube.png

Figure 2: An Individual's path in a space-time cube, Source:[Carlstein1978]

[HaggettEtAl1977] uses a three-dimensional matrix to represent relationships of times, locations and activities (attributes). [Sinton1978], as already mentioned in the motivation, is even clearer: he not only states that all three dimensions need to be present in GIS data models, but also distinguishes between fixed, controlled and measured ways to aquire data. He mentions that in most cases, GIS specialists fix the time component and control and measure the space or attribute dimension.

With the establishment of GIScience as an independent scientific discipline [Goodchild1990] at the end of the 80s/beginning of the 90s, spatio-temporal modelling, visualization and analysis became central research questions. Prominent researchers in the space and time domain are [Langran1992], who primarily approached this topic from a conceptual and technical perspective (modelling and implementation), and [Peuquet2002] who additionally added a more theoretical and interdisciplinary aspect to the topic. Peuquet also dealt with perceptual and cognitive issues and included a broader view on the topic integrating also aspects from philosophy, psychology, linguistics, anthropology, neuroscience and artificial intelligence. [OttSwiaczny2001] provide a good overview over the current research on time in GISystems and GIScience. They discuss topics such as conceptualizing entities in spatio-temporal GIS, integrating and implementing time in GIS, analysis and visualization. They also provide a comprehensive research bibliography.

One of the more recent and promising projects dealing with spatio-temporal maps and temporal navigation is the TimeMap project by the University of Sydney, an open source Java based web application [Johnson2004]. Additional research can be found at the homepage of the ICA commission on visualization and virtual environments [MacEachrenKraak2005] and the Geovista homepage of the Department of Geography, Penn State University [GeoVista2005]. Today, modelling and visualizing time and spatio-temporal navigation in GIS is a truly interdisciplinary research topic, including domains such as geography, social and life sciences, psychology, philosophy, GIScience, GIS, cartography, computer science, information visualization, multimedia design, mathematics, statistics, etc. Substantial input is currently contributed from information visualization, a discipline that deals a lot with interactive graphics, visualizing large datasets and data mining issues ([CardEtAl1999]).

1.3 Terms and Definitions

Following are definitions of the most relevant terms and concepts as used in this article:

Space is the basic organization concept in geography. In spatial data, elements or entities have a location or series of locations associated with them. These locations are either zero, one, two or three dimensional. In the further discussion and the prototype implemented during this PhD project, it is almost exclusively dealt with two dimensional spatial data and representations. Locations are usually expressed implicitly or explicitly in a coordinate system. As to [Yattaw1999] space can either be measured metrically or temporally with clocks (e.g. “drive time”). This shows already how tightly interlaced the spatial and temporal dimensions usually are. Spatial data is usually modeled as discrete elements (e.g. point features, linear features, polygon features) where the spatial and temporal extents are attached as attributes directly to these entities, or as a continuum (surface data, e.g. grids, TINs or isolines) where the basis of the representation is space and/or time. Individual objects are specified as attributes attached to a given location in the space and/or time continuum. ([Peuquet2002], p. 270). It is beyond the aim of this article to discuss spatial data structures in detail.

Time quantifies or measures the interval between events, or the duration of events. Time has long been perceived as a dimension in which each event has a definite (but not necessarily unique) position in a linear sequence, but as differing from spatial dimensions in that "motion" through time appears restricted to having only a forward direction” ([WikipediaTime2005]). Time is often considered as the fourth cartographic or geographic dimension. Different domains and philosophical schools view, perceive and describe time in different ways ([WikipediaTime2005]and [Peuquet2002]). “In our culture, time is commonly viewed as a line without endpoints that stretches infinitely into the past and the future” ([Langran1992], p. 27), but other, alternative topologies exist as well, such as multiple parallel lines, tree structures, circularity, discreteness and non-existence. According to ([Yattaw1999], p. 88), time cannot be measured tangibly like space but is perceived by its effects (change). From a conceptional point of view one can distinguish between “continuous” or “mechanical” time and “discrete” or “body” time. Mechanical time is defined through the natural rhythms of the earth revolving around the sun (measured or defined through clocks, calendars with days, seasons or years as units). Body time is more related to human activity (e.g. hours, decades and generations). Both, discrete and continuous time concepts, are useful for modeling and visualizing spatio-temporal phenomena and may lead to different discovery patterns and understandings.

Depending on the accuracy of time oberservations and measurement one can differentiate between ordinal scale (e.g. more or less accurately defined geologic eras) and interval scale where the level of accuracy and precision is defined by its granularity, the smallest chronon used (e.g. second, hour, day, year) in measuring time. Scale, like in space, is also an issue here, as granularity may well depend on the topic to be modeled. In geology a granularity of 1000 years may sufficient, where as data needs to be captured every minute or hour in meteorology. ([Frank1998], p. 45) furthermore distinguishes between linear and cyclic time in his taxonomic model of time. Cyclic time is used in case of reoccurring events and needs to be brought into relation with an absolute, linear time scale for accurate calculation against an origin. A further distinction needs to be made between world time (also called event or valid time) and database time (also called transaction time). Langran adds a third facet, the “display” time. The database time always lacks a bit behind the world time. Both are relevant and should be stored in the data model. “Realtime” GISystems try to update the database immediately after an event occurred, they have a better synchronicity.

[Langran1992] (p. 28) introduced the notion of cartographic time, which “distills the characteristics of time that are essential for representing spatiotemporality in the most pragmatic and generic fashion” in contrast to treating time strictly as a continuum with all minor, often irrelevant details. Conceptually, she distinguishes four different models of spatiotemporality ([Langran1992], p. 37ff): The space-time cube, as propagated by [Hägerstrand1970] and others, is the first concept. The second model is a snapshot sequence of time slices. The third concept is called “base state with amendments” where a base state first stores an initial state and only the differences are stored to describe the next snapshot. The fourth model mentioned by Langran is the space-time composite. This composite is constructed by flattening the space-time cube into a two-dimensional data structure. Differences in the time dimension show up as new objects in two-dimensional space. In reality there won't be an ideal data model that suits all requirements. The data model should be chosen in the context of the envisaged data type, the update frequency, database size, visualization type and interactive environment and technical framework.

If attributes are added (in this article also refered to as thema or topic) we end up in a triadic model, as propagated by ([Peuquet2002], p 203; see figure 2). Peuquet suggests in her generalized diagram on how knowledge is cognitively arranged, that our cognitive storage comprises of three subsystems for “what”, “where” and “when”. Each subsystem works cognitively distinct, but highly interrelated and in parallel, and has its own distinct category hierarchies, based on objects, places and processes. The “what system“ is based on recognition, comparing observed evidence with a gradually accumulating store of known objects. In contrast, the “where system” builds on direct perception of scenes in the environment, picking out relevant, invariant bits from the rich flow of sensory information. The “when system” operates through the detection of change over time in both stored object and place knowledge, as well as sensory information.

triadic_model_peuquet.png

Figure 3: Diagram of how knowledge is cognitively arranged in memory. Source: [Peuquet2002], p. 203

In an entirely different classification scheme, based on dimensions, ([OttSwiaczny2001], p. 4) extend Harrowers multidimensional cube of space, time and attribute. Within this cube they visualize the potential number of dimensions. The attribute dimension is categorized into nominal, ordinal, interval and ratio scale. In order to characterize change and movements, all three dimensions are relevant.

multidimensional_cube_ott_swiaczny.png

Figure 4: Multidimensional cube of space, time and attribute with potential number of dimensions. Source: [OttSwiaczny2001], p. 4

The relationship between space and time is usually a very close one. The quotation “Nothing in the world is purely spatial or purely temporal; everything is process” ([Blaut1961]) clearly emphasizes this fact. In physics, after Einsteins discoveries, the term spacetime indicates, that time and three-dimensional space together form a single four-dimensional object ([WikipediaSpacetime2005]). A “point” in spacetime is denoted as an event (e.g. the explosion of a star) and must be defined in all four dimensions. In geography or cartography, however, the simpler Newtonian concept of noninteracting space and time is often used. “While it may well be that a geographer's spatial and temporal dimensions do interact, spatiotemporal data are more generally useful when space and time are recorded separately.” ([Langran1992], p. 29).

Navigation, in this context, can be described as the task of determining position within the information space and finding the course to the envisaged information and other relevant related information. Navigation helps explore information spaces that are too large to be conveniently displayed in a single window. The information space consists of spatial, temporal and thematic dimensions. Therefore, one distinguishes between spatial, temporal and thematic navigation. These three modes build on different methodologies and tools, but are closely interrelated. Good navigation tools provide answers to questions, such as “Where am I?”, “Where can I go?”, “How will I get there?”, or “How can I get back to where I once was?”. Navigation tools, together with spatio-temporal visualization, help to solve spatio-temporal tasks more efficiently and provide both overview and detail. The several navigation modi, available tools, navigation metaphers and the advantages or disadvantages of specific navigation methods are discussed later in this article.

2. Spatial, Temporal and Thematic Navigation Tools

2.1 Why Navigation?

The “Where am I?” question is one of the most basic questions. Disorientation makes people feel insecure. “Lost in Hyperspace” is a common problem of internet surfers and multimedia users, especially for inexperienced ones. In hyperspace we have fewer sensories at hand to find our way. We can't rely on touch, balance and smell. The use of audio is limited and our viewing angle is often narrow. Moving through hyperspace or multimedia products can be compared to moving around in the real world with diving goggles.

Good navigation tools guide the user through the information space of a product or webpage, helping explore the content and functionalities of a project. Navigation enables the exploration of information spaces that are too large to be conveniently displayed in a single window. Efficient navigation tools provide both overview and detail. Ideally, multiple user profiles are supported: beginners, intermediates and experts. Experts need more precise and efficient navigation tools and a more direct access to the information offered. Good navigation tools are intuitive and don't force the user to first study extensive manuals or help documents. They are consistent throughout the whole product. This includes the appearance (e.g. colors and fonts), the placement and the functionality of the tools. Navigation tools should not be dominant in the screen layout of a product. A history function allows the user to go back to previous views.

There is no single best, one-size-fits-all navigation method available. The suitability of navigation tools depends on the audience, the user's experience and motivation, the task to solve and the type and size of the information space. A rough rule of thumb is that the larger and more complex the information space is, the more sophisticated and efficient the visualization and navigation tools should be. Good visualization systems provide multiple methods for solving the envisaged tasks, supporting different preferences and capabilities of their audience.

When introducing new methods for user interfaces and navigation tools,[HarrowerSheesley2005] mention an interesting fact: people aquire interface skills primarily through exposure and repetition. Navigation methods that were present in early days of emerging technologies and commonly used, not necessarily represent the best possible implementation. It is not easy to introduce new and potentially better navigation tools and make them appealing to the user. Harrower suggests that design decisions made in early days of emerging technologies should be critically questioned before introducing them. The past has shown that often technical limits or the lazyness of developers have driven these decisions, rather than trying to really understand the users' needs.

2.2 Navigation Metaphers

Depending on the target audience, different navigation metaphers might be suitable. For projects with a broader audience several navigation metaphers can be used in parallel. The following list shows a few potential metaphers that can be used:

2.3 Spatial Navigation

Spatial navigation in two dimensional maps consists primarily of zooming and panning. Zooming enlarges or reduces a map to see it more clearly or to get a better overview. Alternatively, we can say that the user is changing the map scale if he is zooming. Panning describes the process of repositioning or re-centering the map. The term “panning” is derived from “panorama”, from “machines that unrolled or unfolded a long horizontal painting to give the impression the scene was passing by”[WikipediaPanning2005]. For zooming, we distinguish between physical and semantic or logical zooming. Whereas physical zooming is a mere geometric magnification, semantic zooming also changes the shape of geographic objects and adds more detail through more fine grained selections. Semantic zooming usually requires a LOD concept and is closely related to automatic generalization. [HarrowerSheesley2005] differentiate between “precise” and “fuzzy” spatial navigation tools. Precise zoom and pan tools are usually used in scenarios where users know exactly the position either through exact coordinates or geographic names. They are often non-visual tools. Among the graphical spatial navigation tools there are also various levels of precision. Zooming and panning by dragging and drawing a zoom box directly in the map is substantially more precise than having to zoom and pan with predefined zoom steps and discrete panning. Finally, one can distinguish continuous versus discrete or non-sequential versus sequential panning and zooming. Panning and zooming with fixed steps are discrete and sequential, whereas zoom sliders or scrollbars would be continuous.

Zoom OptionsPan/Scroll OptionsSpatial Reference/Context
plus/minus zoom buttons (discrete steps, sequential)pan hand (continuous, non-sequential)linked reference map indicates map extent in context/overview map
zoom slider (continuous)recentering with mouse clickcoordinate display on mousemove
fullview or predefined zoom step buttons with scale cues (e.g. street, city, province, country level)pan buttons at map edges/corners or somewhere in the navigation interface (discrete steps, sequential)“rocket functionality” lets one quickly zoom out and back in to see larger map context (see http://www.map24.com/ (Java version)
zooming by resizing a rectangle in linked reference mappanning by repositioning a rectangle in linked reference mapgrid lines
zooming and by entering a scale value in text input boxpanning with smart scrollbarsscale bar
zoom to selected objectspanning with arrow keys (keyboard)scale cues (e.g. street level, city level)
zooming and panning by dragging and drawing a zoom box
zooming and panning by searching for geographic names (gazeteer) – non visual
zooming and panning by explicitly entering coordinates in text input boxes – non visual
zoom and pan history lets one step forward and backwards between previously extents

Table 1: Zooming and Panning Options, Spatial Reference Options, Source: [Neumann2005]

Measuring the efficiency and functionality of spatial navigation tools is not an easy task. [HarrowerSheesley2005] suggest four functionality and two efficiency criteria. The functionality critera are sequential vs. non-sequential map browsing (fixed steps vs. continuous zooming an panning), precision, local-global orientation cues, live-linked manipulation. The efficiency criteria are the interface workload and the information-to-interface ratio. The workload can mainly be determined in a qualitative way. The NASA TLX worksheet and the GOMS model are two possible qualitative methods to determine the workload. The TLX worksheet includes six workload sources: mental demand, physical demand, temporal demand, performance, effort and frustration level. Each criteria should be rated on a 5 point Likert scale. The information-to-interface ratio is derived from Edward Tufte's concept of the data-ink ratio. Navigation tools should not dominate the user interface. Everything not relevant to the information and important for the functionality should be omited. [HarrowerSheesley2005] detected two superior spatial navigation tools: the linked reference map and smart scrollbars where the scroller length indicates the position and percentage of the current sector in the context of the whole map.

screenshot_nisyros.png

Figure 5: Example for a well designed user interface with spatial navigation tools and a linked reference map. Source: [Flueler2005]

2.4 Temporal Navigation

As discussed above in the section on “conceptualizing and modeling time”, time cannot be seen and visualized directly, but only through the changes in the spatial and thematic dimensions. It is only natural, therefore, that temporal controls are usually closely tied to visualization, serving both navigation and visualization purposes. Depending on the topology, range and granularity of the time, linear, circular or tree representations and controls are more suitable. Following is a table of possible time controls and representations:

Linear controls and temporal visualizationCircular controls and temporal visualizationOther controls and temporal visualization
interactive timelines / lifelinesknobsday / night buttons
temporal sliderinteractive watchescalendar widgets
smart time scrollbarswheels of months or seasonstext input
small multiples along timelinecogwheelsselection lists (e.g. list of days, months, years)
VCR like controls: play, pause, stop, fast forward/backward buttons

Table 2: Temporal Navigation Options - Linear, Circular and other controls and temporal visualization, Source: [Neumann2005]

filmfinder.png

Figure 6: Film Finder as an Example for the use of Sliders for Temporal and thematic navigation. Source: [AhlbergShneiderman1994]

Interactive timelines are particularly useful since they can serve both visualization and navigation. Timelines can be used to show events, processes and the distribution of attribute values along time. Processes or uncertainty can be visualized using gradients. Timelines can be combined with sliders and scrollbars. Small multiples are a series of small pictures representing the same topic over time. These can be ordered along a timeline. They are propagated among others by [Tufte1990] and [Monmonier1992]. An interesting option is to use cogwheels for controlling time. A series of cogwheels with different sizes and ordered by granularity can intertwine with each other and allow the user to navigate through time at various speeds. Finally, the space-time cube could be an interesting navigation tool to enable combined spatio-temporal navigation and visualization. Slicing planes could be moved within the cube to cut out space or time clippings of the whole space-time cube.

time_cogwheels.png

Figure 7: Cogwheels for Navigation in and Representation of Time, Source: [Drewe2005]

2.5 Thematic Navigation

The optimal technique for navigation in a topic depends very much on the nature and data type, as well as the size of the information space to be visualized. Disciplines dealing with the visualization and navigation within topics are “information visualization”, “document visualization” and the “semantic web” community. Basically, one can distinguish between hierarchic data structures and graphs or network data structures. Hierarchic data structures are often visualized using interactive trees where one interactively opens and collapses sub-hierarchies. A typical example for such a tree structure is the file-browser, present in any modern GUI to operating systems. A special form of a hierarchic tree is the hyperbolic browser developed and propagated by Inxight, where items can be interactively moved to the center of the visualization. Linked branches can be opened or collapsed or automatically appear as they move closer to the center of the visualization. Objects farther away from the visualization center are perspectively scaled down and later hidden. A nice demonstration utilizing this technology for interactive grocery shopping can be explored at [Inxight2005]. Treemaps are alternative tree representations, originally introduced by [JohnsonShneiderman1991]. Within the treemap, one starts with a large root rectangle, filling the entire canvas. Within that rectangle are smaller rectangles, one for each subordinate node of the node just considered. This procedure is repeated until all nodes are represented. There is no limit on the depth of the tree. Color and text labels can be used to represent different attributes.

hyperbolic_tree.png

Figure 8: Hyperbolic Tree, Source: [Inxight2005]

Increasingly, interactive maps are used to display a topic in two or three dimensions. The information space is defined by two or three attributes. Objects that are closer to others, according to the attribute space, appear closer in the map. Links to other objects can be visualized using linear or network structures and arrows indicating directions. When using a LOD concept, important objects appear first, less important ones only after zooming into the information space. Objects that are close to eachother are often grouped in galaxies, in which the user can dig into deeper, gradually revealing more details. Exponents of this technology are e.g. graphical front-ends to search engines, such as [Grokker2005] or [Kartoo2005]. Another form of information landscapes is the themescape. A themescape is a thematic terrain where the elevation indicates theme strength. Peaks indicate where concentrations of closely related objects appear. An example application for such themescapes is available as part of the “Aureka” product, a patent browser of a company called Micropatent (http://www.micropatent.com/). Using the map metaphor, one can use spatial navigation and analysis tools to explore the themescape. This includes zooming, panning, LOD or perspective views. Hierarchical Menus and Sitemaps, as common in more complex webpages are additional useful tools that can be used to help users navigate in information space.

themescape.png

Figure 9: Themescape, Source: http://www.micropatent.com/

kartoo.png

Figure 10: Screenshot KartOO, Source: [Kartoo2005]

grokker.png

Figure 11: Screenshot Grokker, Source: [Grokker2005]

newsmap.png

Figure 12: Screenshot Newsmap "Marumushi", Source: [Marumushi2005]

Depending on the data type and scale of the topic, statistical analysis, charts and diagrams might be useful for visualization and navigation. Examples include scatterplots, histograms or parallel coordinate plots. It has proven useful to provide several visualization and navigation options and link them together using brushing and linking techniques. Query builder, such as available in many database and GIS products help to filter data.

3. Interdependencies of Spatial, Temporal and Thematic Navigation

A study of the interdependencies in spatial, temporal and thematic navigation includes questions, such as: What effects should changes in one of the three dimensions (space, time or topic) trigger in the two remaining dimensions? How should the user interface adapt to the new constraints? What do changes in either dimension mean for the cartographic visualization within the map? A systematic reflection of these dependencies leads to the following table. The first column is the fixed, given dimension that is set interactively by the user. The other two remaining dimensions are changed by the visualization system as a response to the change in the first dimension set by the user.

Fixed (Given) dimension (changed by user)First, Variable dimensionSecond dependent dimensionEffects on Space (Map Design and Map Content)Effects on Time (Visualization of Time and Time Controls)Effects on Topic (Thematic Choice, Depiction and Tools)
SpaceTimeTopic
  • Adapt map content according to map extent (generalization)
  • Display objects according to relevance in topic
  • Resymbolize map in relation to map extent
  • Choose different scale for base maps
  • System should display in timeline only objects relevant for the given space
  • Other elements could be hidden or dimmed
  • System should only give thematic choice available in the area selected by the user
TopicTime
Time (Span) SpaceTopic
  • Display objects according to relevance in time span
  • Adjust global relevant bounding box according to new time span
  • Adapt to new global time scale
  • adjust granularity in global timeline and clockspeed in animation
  • display objects in timeline according to relevance
  • System should only give thematic choice available in the time-span selected by the user
TopicSpace
TopicSpaceTime
  • Adjust global relevant bounding box according to new topic selected by the user
  • Display different topic in map according to rules in data model
  • Show new relevant time spans in timeline view according to new topic
  • Highlight new active topic in timeline view
  • Re-order neighbor topics in timeline view
TimeSpace

Table 3: Interdependencies of Spatial, Temporal and Thematic Navigation, Source: [Neumann2005]

4. Prototype implementation of an interactive 2D map of european cultural heritage

A prototype is currently being implemented to demonstrate the feasibility of the proposed navigation tools, as well as their close interrelationships. The topic of the prototype is a map of european artists that displays where the artists worked and lived. The prototype allows the user to navigate in time and topic. Thematic navigation includes the choice of various art styles (e.g. painters, sculptors, architects, poets, etc.). Furthermore, the user can display network diagrams that show links from selected artists to related artists, e.g. friends, collaborators, apprentices, teachers. Links to Google and Wikipedia are provided for artists and events. Temporal navigation and visualization is implemented using timelines. They show the life span of artists as well as important events in their life. If data is available, travelling of artists is visualized in the map. To put the artist's life into context, important events, such as wars, political events, inventions, etc. are visualized in a separate timeline. Timelines are linked to the map and vice versa. Highlighting emphasizes the corresponding elements in the temporal and spatial visualization. If there is a corresponding network diagram available, it is linked as well. The screenshot below shows the current version of the prototype. As this is work in progress, the screen layout and available navigation tools are subject to change. The current version of the tool can be found at [NeumannPrototype2005].

screenshot_cultural_heritage_europe.png

Figure 13: Screenshot of the current, preliminary version of the prototype "European Map of Cultural Heritage", Source: [NeumannPrototype2005]

The project was inspired from the Twistory software [Boer2003], a Macintosh freeware history browser that displays lifetimes and travels of historic figures. This PhD project implements similar features than the twistory project, but with higher emphasis on cartographic quality, navigation and interactivity. Furthermore, the prototype utilizes advantages from the fact that it is web-based. Resources are linked to search engines and online encyclopedias, such as Wikipedia. The prototype is technologically based on SVG (Scalable Vector Graphics) and ECMAScript on the client, with a PostgreSQL/Postgis spatial database backend that collaborates with Apache and PHP on the server. Postgis is an OGC compliant spatial extension to PostgreSQL that features spatial queries, reprojection of data, geometric processing (such as overlay and buffering) and output in SVG format (among other formats) [Postgis2005]. The application follows the client-server approach where only portions of the whole dataset are loaded on demand. Spatial data is filtered and generalized to deliver the user scale-adequate map data. The generalization process is utilizing the Postgis function "Simplify()" which is an implementation of the Douglas-Poiker line filtering algorithm. It can be regarded as a data reduction step, rather than a sophisticated generalization. Basic topographic geometry data was obtained from the WDBII freeware datasets and corrected and enhanced in ESRI ArcGIS. A shaded relief with natural colors was provided by Tom Patterson ([Patterson2005]). Thematic Data (database of artists) was obtained from the Paul Getty Research Institute ([Getty2005]).

SVG was chosen as base technology for the client part of the prototype for several reasons: The superior rendering quality and advanced features, such as text on path, the ability to do network requests, such as "getURL()" or "XMLHttpRequest()", the various interactivity, scripting and animation features and finally the growing SVG support in various open source and commercial server and desktop projects. The Postgis spatial database e.g. directly outputs SVG geometry. A tutorial available at [NeumannWilliams2005] introduces the mechanisms necessary to load additional geodata into SVG mapping applications. The downside from developing SVG applications is the current state of SVG renderers available. None of the current viewers implements the whole SVG 1.0 or SVG 1.1 specification and many of the viewers are still in early development. Browser and OS support varies. Hopefully, the upcoming webbrowsers will feature mature native SVG support directly in the browser, without the need for plugins.

5. Summary and Outlook

This paper summarizes the various options for navigating in space, time and thema. This relatively new research topic is closely related to information visualization and interactive mapping research topics. Implementing tools and controls to navigate through large information spaces using small bandwidth connections across the internet is still a challenge. SVG already offers interesting features and the upcoming version 1.2 with it's support for sXBL will offer even better tools for implementing interactive applications within a web context. The work on the prototype is still in early stages and thus no user tests have been carried out. Further results of the research will be published at the website carto.net or maybe at next year's SVG.Open conference.

Acknowledgements

I'd like to thank my PhD supervisors Prof. Sara Fabrikant, Dr. Hansruedi Bär and Prof. Lorenz Hurni. I'd like to thank the SVG community and W3C SVG working group for providing numerous examples and their continuous support with SVG and programming problems. The PostgreSQL/Postgis and PHP developers enabled my serverside framework, which proved to be reliable and easy to maintain. My girlfriend Juliana Williams helped correcting my typos and disentangling complicated sentences. She also helped a lot with data preparation and her math skills.

Bibliography

[AhlbergShneiderman1994]
Ahlberg, Christopher and Ben Shneiderman (1961): Visual Information Seeking: Tight coupling of dynamic query filters with starfield displays, in: Proc. of ACM CHI94 Conference (April 1994), 313-317 + color plates. Reprinted in Baecker, R. M., Grudin, J., Buxton, W. A. S., and Greenberg, S. (Editors), Readings in Human-Computer Interaction: Toward the Year 2000, Second Edition, Morgan Kaufmann Publishers, Inc., San Francisco, CA (1995), 450-456.
[Blaut1961]
Blaut, J.M. (1961): Space and Process, in: The Professional Geographer, Volume 13, p. 1-7.
[Boer2003]
Boer, J.D. (2003): Twistory Project, http://www.king.igs.net/~deboer/twist/
[CardEtAl1999]
Card, Stuart, Mackinlay, Jock and Ben Shneidermann (1999): Readings in Information Visualization, Using Vision to Think, Morgan Kaufmann, San Diego.
[Carlstein1978]
Carlstein, T. et al (1978): Human Activity and Time Geography, editor: Edward Arnold, London.
[Drewe2005]
Drewe, Paul (2005): What about time in urban planning & design in the ICT age?, in: Proceedings of the CORP conference 2005, Vienna, 25 pages.
[Flueler2005]
Flueler, Isabella (2005): Slope Stability on Nisyros Island, http://www.carto.net/flueler/nisyros/
[Frank1998]
Frank, Andrew U. (1998): Different types of times in GIS, in: Spatial and Temporal Reasoning in GIS, Oxford University Press, Oxford, 40-62.
[GeoVista2005]
GeoVISTA (2005): GeoVISTA Center, http://www.geovista.psu.edu/
[Getty2005]
Getty Research Institute (2005): Getty Vocabulary Program, http://www.getty.edu/research/conducting_research/vocabularies/
[Goodchild1990]
Goodchild, Michael (1990): Spatial Information Science, in: Proceedings 4th Symposium on Spatial Data Handling, Zurich, Switzerland, p. 3-12.
[Grokker2005]
Grokker (2005): Grokker, http://www.grokker.com/
[Hägerstrand1952]
Hägerstrand, Torsten (1952): The propagation of innovation waves, in: Lund Studies in Geography B, Human Geography, Volume , p. 3-19.
[Hägerstrand1970]
Hägerstrand, Torsten (1970): What about people in regional science?, in: Regional Science Association Papers, Volume 24, p. 7-21.
[HaggettEtAl1977]
Haggett, P., Cliff, A.D. and Frey, A. (1977): Locational Models, Halstead Press, New York.
[Harrower1999]
Harrower, Mark (1999): A Conceptual Framework for Mapping Space-Time, in: Proceedings at the 95th Annual Meeting of AAG, Honolulu, Hawai, 14 pages.
[HarrowerSheesley2005]
Harrower, Mark and Benjamin Sheesley (1999): Designing better map interfaces: A framework for panning and zooming, in: Transactions in GIS, Volume 9 (2)
[Inxight2005]
Inxight (2005): Hyperbolic Browser - Shopping Demo, http://www.inxight.com/demos/grocery/index.htm
[Johnson2004]
Johnson, Ian (2004): TimeMap Open Source Consortium, http://www.timemap.net/
[JohnsonShneiderman1991]
Johnson, B. and B. Shneiderman (1991): Treemaps: a space-filling approach to the visualisation of hierarchical information structures, IEEE Proceedings Information Visualization, , p. 284-291.
[Kartoo2005]
Kartoo (2005): KartOO - Visual Meta Search Engine, http://www.kartoo.com/
[Kraak2003]
Kraak, Menno-Jan (2003): The space-time cube revisited from a geovisualization perspective, in: Proc. of the 21st ICA Conference, Durban, South-Africa, 8 pages.
[Langran1992]
Langran, Gail (1992): Time in Geographic Information Systems, Taylor and Francis, London, Washington DC.
[MacEachrenKraak2005]
MacEachren, Alan and Menno-Jan Kraak (2005): ICA Commission on Visualization, http://kartoweb.itc.nl/icavis/index.html.
[Marumushi2005]
Marumushi (2005): newsmap, http://www.marumushi.com/apps/newsmap/newsmap.cfm
[Monmonier1992]
Monmonier, Marc (1992): Time and motion as strategic variables in the analysis and communication of correlation, in: Proc. of the 5th Int. Symposium on Spatial Data Handling, Charleston SC, p. 72-81.
[MooreEtAl2003]
Moore, A.B., Whigwham, P., Holt, A., Alridge, C. and K. Hodge (2003): A Time Geography Approach to the Visualization of Sport, in: Proc. of the 7th Intern. Conference on Geocomputation, 13 pages.
[Neumann2005]
Neumann, Andreas (2005): Navigation in Space, Time and Topic, Interdependencies of Spatial, Temporal and Thematic Navigation in 2D Interactive Maps, in: Proc. of the 22nd ICA conference in A Coruna, 10 pages.
[NeumannPrototype2005]
Neumann, Andreas (2005): European Map of Cultural Heritage, http://www.carto.net/neumann/phd/prototype/
[NeumannWilliams2005]
Neumann, Andreas and Juliana Williams (2005): Dynamic Loading of Vector Geodata for SVG Mapping Applications Using Postgis, PHP and getURL(), http://www.carto.net/papers/svg/postgis_geturl/
[OttSwiaczny2001]
Ott, Thomas and Frank Swiaczny (2001): Time Integrative Geographic Information Systems, Springer, Heidelberg.
[Patterson2005]
Patterson, Tom (2005): Introducing Natural Earth, http://www.shadedrelief.com/natural/index.html
[Peuquet2002]
Peuquet, Donna J. (2002): Representations of Space and Time, The Guilford Press, New York/London.
[Postgis2005]
Refractions (2005): Postgis, http://postgis.refractions.net/
[Sinton1978]
Sinton, David (1978): The inherent structure of information as a constraint to analysis, in: Harvard Papers on GIS, Volume 6, p. 1-17, Reading, Massachusets, 1978.
[Tufte1990]
Tufte, Edward (1990): Envisioning Information, Graphics Press, Cheshire CT.
[WikipediaPanning2005]
Wikipedia (2005): Panning, http://en.wikipedia.org/wiki/Panning
[WikipediaSpacetime2005]
Wikipedia (2005): Spacetime, http://en.wikipedia.org/wiki/Spacetime
[WikipediaTime2005]
Wikipedia (2005): Time, http://en.wikipedia.org/wiki/Time
[Yattaw1999]
Yattaw, Nancy J. (1999): Conceptualizing Space and Time: A Classification of Geographic Movement, in: Cartography and Geographic Information Science, Volume 26, No 2, 1999, p. 85-98.

XHTML rendition made possible by SchemaSoft's Document Interpreter™ technology.