5. The Research Agenda

5.1 The Process

Following the plenary sessions during which reviews of papers were presented, the questions put forth by the discussants and the audience were placed into the following nine categories:


Based on these categories, the participants of the workshop broke into working groups comprised of 6 to 7 members. Working groups, using as the basis of their deliberations the categorized questions, identified specific research questions and tasks at the granularity of a Ph.D. or Master's thesis. A representative of each working group then reported in a plenary session the group's findings. These findings were discussed and modified via suggestions by all participants of the meeting.

At a later stage, the researchable questions and tasks that were formulated by the working groups were placed into the following framework:

This list of researchable questions and tasks was subsequently distributed to the participants of the Specialist Meeting who were asked to (1) comment on the completeness and representatives of this list, and to (2) rank the questions in terms of relative importance (most important to least important). This was not to suggest that some of the questions raised at the workshop were not important, but rather to identify which questions should be addressed first in terms of a research agenda. The workshop participants were told that the purpose of this ranking process was to determine a research agenda for the NCGIA, for the participants of the Specialist Meeting, and for researchers interested in spatio-temporal issues in GIS. The remainder of this section describes researchable topics and key questions under each of these areas. Note that the questions under each topic are listed in rank order as identified by participants in the Specialist Meeting.

5.2 Language, Culture, and Human Cognitive Representation and Behavior

The overall objective in this area is to get a better understanding of how people deal with geographic space and time. It relates to the discussions in the sessions on Reasoning and Philosophy, the Social Science Perspective, and Spatio-Temporal Cognition, where the following questions and research tasks were identified:

Of particular interest were questions on linguistic and cultural issues, on which two working groups focused their attention, and the distinction between propositional and image representation. A fourth working group theme, closely related to cognitive representation and behavior, was topological vs. metric spatial knowledge. Since it had more overlap with the category on formal systems, details will be provided in Section 5.3.

5.2.1 Language and Linguistic Issues

Language is used to describe time, space, and sometimes both together. In natural language, however, there seems to be an asymmetry between time and space. Space seems more important than time--spatial concepts are acquired sooner than temporal concepts in children (and apes). There are more terms to describe spatial relationships than temporal relationships. Time is ephemeral and private; you cannot revisit events.

What kinds of models of the world does language yield? The temporal model that our language provides is very strange. Most languages do not allow you to assign a temporal name to a spatial object. Language gives you wonderful economy in describing space. The ambiguity is often considered a disadvantage, but is it? There is a default meaning that we apply based on context. Language allows ambiguity. Is the ambiguity helpful or is it a negative? Verbal instructions will probably become more important in GIS. "Take the next exit" works better than looking at a map in a Vehicle Navigation System. What element of the instructions are spatial vs. temporal? Language can more precisely capture uncertainty than graphics. What kind of spatial-temporal phenomena are better understood/utilized via language (e.g., driving directions) and what kinds are better understood/utilized via graphics (e.g., maps).

There are differences in the representational power of language. In giving directions you can choose how specific to be. There is flexibility to jump to any scale as required. In a temporal GIS, granularity should change as required. If there is no change, then the temporal scale has the flexibility to get coarser.

Three approaches can be employed to explore language issues: (1) one could study language by looking at forms of knowledge that are relatively linguistic--implication is that language is a window to cognition; (2) one could also conduct experiments to access cognitive processes that cannot be observed directly through language; and (3) one could also employ both of the above approaches (i.e., experiments on language).

Several questions were raised about what would be interesting findings about language for space and time:

5.2.2 Cultural Aspects with Respect to Spatio-Temporal Reasoning

Culture is an issue for GIS and temporal GIS, because GISs have for the most part, been developed from a mono-language community. What kind of influence does culture have on understanding and representing time? Different cultures manage time differently. This may be a good way to approach the problem of working with a cross-cultural space-time GIS. There are three different temporal scales which evolve: (1) the evolution of humans, (2) the individual, and (3) the culture that individuals are members of. The cultural aspects of time are built on all three scales. One is the biological makeup of our bodies which makes us share fundamental approaches to time. Life itself is a process. The heartbeat is the basic clock. In terms of the individual scale, our time of birth is exactly recorded. In between these two there is the effect of culture on our conception of time. First, is the comparative view of time. The multi-cultural view is relevant in this country. If we have an agency with varying people, how does that affect GIS. Then there is the historical concept of time. Also, there are the varying subcultures within a nation (social and economic influences).

People perceive their landscape by taking the physical landscape and warping it to the perceived (cultural) landscape. Land ownership is a good example of an application in GIS which has a necessary temporal component. Are there places where the decisions a society makes is actually different. In the U.S. there exists the Hopi Indian land dispute which is based on historical land ownership. One proposed solution is to buy some land and give it to the Hopi or Navajo. There are different conceptions of land-ownership, which often boil down to a political situation. In Latvia, they are trying to recreate old geographies to get back to the old land ownership rules. An interesting study would be to stand back and look at these cultural-problems and how they evolve over time. Would cross-cultural studies of databases or data-collections be fruitful in understanding how different cultures deal with time?

One needs a temporal GIS, because people are not working in the current time. One can simultaneously have separate cultural views of the same space. Some of the GIS cultural problems are interface problems. In Britain, some corporate cultures are changing GIS to fit their corporate culture. To what extent is a GIS obligated to conform to a given subculture? Can we identify a research question that will allow us to see how GIS affects a culture?

5.2.3 Propositional vs. Image Representation

A smallgroup discussion on propositional vs. image schemas pointed out that the difficulty to define them and to distinguish between them. A propositional schema seems to be something with a truth value, whereas image representations use space, particularly distance, direction, and orientation, in meaningful ways. This dichotomy does not do justice to the richness of possible representations and to making hybrid nature of many of them. What has been called imagery, for example, may include visual, spatial, and kinesthetic/motor imagery, all of which are separable.

The group concluded that are three types of representations: (1) those in the mind (supposedly), (2) those in computer programs, and (3) those in the world, i.e., on paper or on a computer screen or a physical model. It is difficult to talk about representations without talking about the processes performed on them.

5.2.4 Researchable Questions

Spatial cognitive representations

Influence of the task environment on spatio-temporal cognitive representations Communication about space and time, and the influence of culture

5.3 Formal Systems

The need for formal systems for spatio-temporal reasoning comes from the desire to build GISs that can be used to perform spatio-temporal reasoning tasks. Formalisms are also useful in comparing different models. Questions and tasks for this category were supplied by the discussions in sessions on Computational Issues, Spatial and Temporal Cognition, and the GIS perspectives. They included:

  1. local metric neighborhoods;
  2. strip maps (are they the same as route maps?);
  3. categorical information;
  4. hierarchical information;
  5. temporal sequences; and
  6. propositional representations.

Participants spent more time on identifying specific issues in the areas of categorizations of time (temporal taxonomies) and what primitives are; the difference between topological and metric knowledge; the linkage among different scales (micro-macro issues); and concerns about the implementation of temporal GISs.

5.3.1 Temporal Taxonomies

A number of models exist for representing time. The following is a list of models that could be explored (individually or in comparative studies) for implementation in a temporal GIS: (1) mechanical time, (2) event-driven time, (3) absolute time, (4) discrete vs. continuous time, (5) relative time, (6) direct determination/measurement, (7) indirect determination/measurement, (8) work-flow or process time, (9) partial orderings and (10) fully ordered. Processes operate in the world and are embedded in time. Elements of a taxonomy of time can be continuous or discrete, as are processes. Cycles are a property of processes, not time--we can derive time through measurements. Processes can be observed and measured. Observation can be direct or indirect. There are a number of ways to represent time. We can use interval methods, such as calendars and clocks which are absolute, or we can use ordinal methods which are relative. We can also represent time with cyclic representations--the storage of repetition of information or by propositional information.

Examples of temporal topologies included linear complete ordering, branch (forward and backward) versioning, dendritic as an example, braided (lattice) partial ordering, and cyclic or quasi-cyclic (a la Lorentz attractors). Problems arise when defining the granularity of measurement, determining connections between time and measurement, dealing with imprecise relative dates and indirect dates, and attempting to discretize space, time, and attributes--Sinton's model (1978) suggest that, "you cannot measure all three."

Several research tasks and researchable questions related to temporal taxonomy are:

5.3.2 Primitives (location, time, motion)

These discussions were motivated by the question of whether or not space and time are separate categories. Can we substitute space for time as Kant did? In some models, we can substitute space and time to answer different questions. An example is the use of graph theory to answer questions about connectivity (topology) and order (as in a pert chart). But, plausible interpretations of 2-d maps (such as a fertility time-slice) led to the rejection of these inferential techniques until the addition of the time dimension. Therefore, in these instances, time and space cannot be separated as has been done historically in geographic thought.

More detailed research tasks and questions about space, time, and motion were formulated:

5.3.3 Topological and Metric Spatial Knowledge

There is a need to define what we mean by topological, metric, configurational, qualitative and quantitative metrics when discussing spatial knowledge types. This need arises because of the diverse academic backgrounds of researchers involved in this area of inquiry (geography, psychology, mathematics, computer science, cognitive science). In this group, it was discovered that different disciplines define similar terms using various terminology. There was confusion as to what was meant by terms such as qualitative and quantitative metric. One of the tasks of this group was to formulate common definitions.

Alternative geometries formally defined by mathematicians as determined by what spatial properties remain invariant under transformations of one type or another. Some disciplines use the term geometry to refer only to metrics--this discussion group generally agreed that all of the spatial knowledge types listed above constitute geometries. There were several questions/issues raised concerning these geometries:

In terms of how humans reason spatially, there were a number of issues raised by this group:
  1. walking two or more legs of a path, estimating straight-line direction to start;
  2. humans can do this non-randomly, as has frequently been empirically demonstrated, but exactly how precisely?
Another area of discussion concerned the explanation of qualitative metrics and its existence:

5.3.4 Micro-Macro Issues

Do processes have natural/characteristic spatial and temporal scales? To answer this question, studies that examine data at different granularities (i.e., scales) in space, in time, and in space and time are needed. This would require a fairly detailed database, in addition to problem definition and guidance on the gathering of the data. Examples of some research topics include examining various physical models (such as hydrologic surficial flow or forest fire) using different resolutions of digital elevation models, examining social questions in a like manner using data sets (perhaps census at a fine resolution and at an aggregated level)--perhaps focusing on housing the market as an example, and examining epidemiological or unemployment data as a process.

Are spatial and temporal scales logically linked? Some links between spatial and temporal scales are clear. For example, daily commuting is associated with maximum length of work trips and that limits extent of urban size. Structuring time structures space. Do processes examined at one scale correlate or describe processes at another scale? Can we use aggregation techniques that will address this issue? What are their rules? Examples of some research topics are to examine biodiversity at a patch or landscape scale. Particular research tasks are to develop aggregation rules and determine if they hold over a variety of scales; examine biodiversity as a dynamic process rather than a collection of species at one time; examine a process at a very cartographically large scale (carbon sequestration) and determine if a smaller cartographic scale will result in a different temporal evolution or if the process can be represented at all; and translate time series methods to spatial temporal series.

Participants of this working group formulated further questions that should be investigated:

5.3.5 Implementation Issues for Temporal GIS

Numerous critical implementation issues must be considered in the development of a temporal GIS. A number of questions were raised that were considered either computer engineering issues or computer science issues.

Computer Engineering Questions: An important first step in developing a temporal GIS is to create a taxonomy of spatio-temporal information systems. Such a taxonomy should be based on current and proposed data models, temporal utility, indexing models, and query facilities. Indexing methods need to be developed for spatio-temporal behavior, features, events and processes. Transfer formats for temporal spatial data (an interlingua) also need to be developed. Efficiency and implementation issues concerning existing formal spatio-temporal models include data representation, spatial indices and query optimization. Need to exploit cognitive science research in representation of spatio-temporal data.

Computer Science Questions: These include the development and implementation of spatio-temporal query algebras, computational formalisms for temporal topology, and parallel computing algorithms for spatio-temporal data.

5.3.6 Researchable Questions

System design and development

Error, imprecision, and uncertainty in spatio-temporal data

Implementation of spatio-temporal GISs

Domain-specific aspects of formal systems for spatio-temporal reasoning

Modeling processes of natural systems in a temporal GIS

5.4 Bridging Human and Formal Systems

A critical role is taken by the linkage between human thinking and a formal system (such as a GIS). This involves communication, as well as interaction and presentation. User interfaces considerations may be based on temporal metaphors such as 3D and 4D (dimension shifting); multiple dimensional rendering; animation (scale shifting); simulation which involves statistical process modeling; other dimensions that include sound or sonic metaphors (music or time sequences) and color including raster false color (synthetic imagery).

At the level of research methodologies, participants suggested to (1) conduct specific case studies on the system level and the user level, and (2) compare cross-sectional studies to longitudinal studies.

5.4.1 Communication and Cartographic Issues

Communication is the transfer of information to the user. Communication is both aspatial and spatial. Aspatial communication includes charts, graphs, and reports for time series whereas spatial communication involves both static and dynamic mapping techniques. Static techniques include dimension shifting, which displays data on a temporal axis (e.g., accident data, time slice map with space constant), and symbology, as when displaying migration and other flow data. Dynamic techniques include animation and dimension shifting. Animation provides the illusion of movement to show time variability with time-dependent multivariante symbols and icons such as graduated symbols, histograms and time-lines. Can one consider either procedural or database animation, for example, the illumination of terrain by the sun over time to show the movement of shadows or the depiction of changes in the shoreline at different times.

There are a number of issues and questions surrounding sonification of spatial/temporal data. Are musical sound types effective for display of data (empirical)? If so, what kinds of data would sound convey effectively, and what sound types would be effective in conveying the intended message? What would the components of a sound interface be for a temporal GIS? One view would be a sonic tool box?

Given current methods for cartographic representations (point, line, areal and volumetric symbology), how can we show temporal and spatial changes? What sort of maps can be derived from animation's? How can missing information be supplied and flagged to the user? Given sparse data samples, can a model be produced that interpolates the gaps in the data.

Investigate different kinds of time series applications and develop a taxonomy of these applications. Evaluate a variety of display techniques for time series analysis. For example, animation vs. static displays (such as dimension shifts).

How does one navigate through a temporal, spatial database? Is a flight simulation metaphor useful for this? What might the components of a Temporal Query Builder for a GIS be?

5.4.2 Researchable Questions

Many of the questions raised here are specific to the design and implementation of user interfaces for spatio-temporal information systems. In addition, many of the questions were raised specific to human representation and understanding of spatio-temporal phenomena, and formalization of these models in a temporal GIS.