Why do I wish to participate in the Workshop on Cognitive Models of Dynamic Phenomena and Their Representation?

I would like to participate in the workshop because of my research interest in geographic representation. I believe that the workshop will provide an excellent opportunity to discuss issues and exchange ideas on cognitive models useful in representing geographic processes. In particular, the workshop distinguishes itself from other types of meetings and conferences by its multidisciplinary setting that brings together scientists and experts from geography, cognitive psychology, information science, computer science, and other related fields to allow in-depth discussions on the workshop topic from multidisciplinary views. My research background will allow me to provide perspectives from conceptualizatoin and representation of geographic processes to workshop discussions.

My research interest relates to the workshop topic in two ways. First, I am interested in how to represent geographic processes to meet the needs of communication and understanding. I believe cognitive models are critical in developing effective geographic representations. My research has been applying the idea of using cognitive models in buiding representations for wildfires and hydrological processes. I study how scientists and practitioners perceive, organize and communicate their knowledge about processes to construct conceptual models, and then I use these conceptual models as the basis to build geographic representation. In doing so, geographic representation schemes are derived based on information granularity and structures of user needs. Hence, useful information can be more easily extracted from GIS data to facilitate a better understanding of processes that generate the data.

The second aspect of my research related to the workshop topic is how to design query algorithms that work with geographic representations to support information discovery from GIS databases. Developing geographic representations based on cognitive models is the key to provide an adequate support for information discovery because it is necessary to understand information granluarity and structures in human cognition to be able to identify what information needs to be computed from GIS databases. Obviously, a GIS cannot support a user query that asks for information beyond what can be represented in the system. Most GIS user queries request information that is pertinent to a better understanding of static or dynamic properties of geographic phenomena. Cogntive models of geographic phenomena can suggest information elements and their structures to facilitate communication and understanding of geographic phenomena. For example, if continuous movement is an important factor required to understand a process, a representation scheme with dynamic (or mathematic) mappings of the process to time and locations is more appropriate than animating snapshots of images. A geographic representation that supports information elements and structures suggested in cognitive models can greatly ease the design of query algorithms. Furthermore, GIS support for data mining and knowledge discovery can be greatly enhanced.

Based on my research background, there are a couple of areas that I can contribute to the workshop. My research experience on knowledge acquisition and representation of wildfires allows me to share congition of wildfire phenomena from various user perspectives. I can also share my recent experience on hydrometeorologic processes and my views of the commonality and differences between the two types of geographic processes. My research suggests that many kinds of geographic information cannot be support by map-based representation since essential behaviors or characteristics of geographic processes cannot be represented in a static form. Map animation cannot fully overcome the difficulty when users are looking for multi-dimensional or comparative information, such as "which fire is spreading faster?" or "where did the storm produce the longest rainfall versus the most intensive rainfall?" In the workshop, I will be able to provide example processes or their behaviors for various cognitive models and how representation schemes (display or conceptual) may influence communication or understanding of these processes. Because of my research focus on geographic representation and information query support, my major contribution should reside in identifying cognitive models, using them as the foundation for building geographic representation to support information query and analysis.

The theme of this position paper accentuates representing geographic processes compatible with human cognition to enable GIS support for information query about process dynamics. Information query support is arguably the most essential function for all information systems, including GIS. Data stored in a GIS will be of little use if important information cannot be extracted from the data. In a GIS, extracting useful information is commonly initiated by user queries. Even when a user is familiar with a GIS, there are three key determinants in the success of a user query: (1) if the user is able to describe the needed information through the interface protocol provided by the GIS; (2) if the requested information is embedded in the GIS databases; and (3) if the GIS is capable of computing and representing the requested information. The position paper stresses that the compatibility of cognitive models and GIS representations is the fundamental issue to all three determinants. Geographic representation needs to account for how users conceptualize geographic worlds and what forms of information they use to develop and address their understanding of the worlds.

Cognitive models and GIS representation

Representation is the conceptual core of an information system. A representation scheme determines data elements and their associations which an information system can use to hold data and depict reality. Obviously, an information system cannot support the kinds of information that its representation schemes cannot incorporate. Because of a strong cartographic tradition, GIS representation has followed the map metaphor to protray reality as a 2D static world. The map metaphor fits well with cognitive models that we use to acquire slowly changing large-scale environments, type C and D spaces in Zubin=92s space typology (Zubin 1989). Maps arguably "are the most efficient and effective way of communicating metric properties of larger scale places, especially configurations" (Montello 1998, p. 151). A GIS based on these 2D static representations provides sufficient support for location information query and computing, such as geometry manipulation, spatial search, overlay, and buffering.

However, geographic processes are dynamic; cognitive models of processes are much more complicated than the above map-based models. Processes consistently evolve and interact with other environment factors in space and time. Processes are usually organized in a hierarchical structure in which processes at a higher scale tend to control the behaviors of processes at a lower scale. Using wildfires as an example, previous studies of the author (Yuan and Albrecht 1996, Yuan 1997) revealed that there are four models of wildfires needed to support analysis and modeling of fire danger, fire behaviors, fire effects, and fire history. The four models are derived through written materials analysis and user interviews, which also suggest these are the models used by researchers and practitioners for information and knowledge acquisition about wildfires. Combination of the four models form an information life cycle through spatial or temporal aggregation to support transforming information from one model to another (Figure 1), compatible with how users apply knowledge about wildfires from study to study. The information cycle also accommodates the needs for multiple views of a fire process as progression energy exchange among plants and the atmosphere (location snapshots with updates of fuel and weather conditions), fire entities moving across a landscape (entity model with fire runs across space through time), environmental impacts of burns (fire effects and environment recovery in burnt areas), and spatiotemporal mosaics of historical burns (fire mosaics).

Although the information cycle is based on wildfire studies, it can be generalized to depict three information elements essential to understanding geographic processes: process, space, and time. Furthermore, the information elements can be used to identify cognitive models of geographic processes. A key cognitive question on geographic processes is how do we determine each of the three information elements. Do we have pre-defined spatial units and time steps to describe process conditions at individual locations (location snapshots)? Do we have an identified process and use the process to determine space and time according to its rate of spread (fire entity)? Do we use the footprints of a process to determine space and have time arbitrarily defined according to observations (entity snapshot)? Or do we define space by a set of process footprints and have time determined by the occurrences of these processes (fire mosaics)? Information we seek to further understand a geographic process depends on how we perceive the process and our prior knowledge. Each of the above four cognitive approaches results in distinct information needs and challenges a GIS to represent the needed information and provide operations for query and analysis to produce the information from its databases.

Representing geographic processes in accordance with cognitive models

Representing geographic processes in concert with information structures and needs in cognitive models can greatly enhance information support and usability of a GIS. The above four cognitive approaches can be generalized into two alternatives to perceiving geographic processes: (1) from location, time, to processes (location snapshot and fire mosaics), and (2) from processes, time, to location (fire entity and entity snapshots). The former view focuses on information about fire conditions or fire history at individual locations, and the latter view focuses on spatial distribution of fires and burns at different times. The two alternatives suggest a representation scheme to enable bi-directional mappings among processes, time, and space in GIS databases (Figure 2). The representation is distinguished from the map-based model in two ways. First, it applies semantic networks to structure concepts about processes and their elements. Hence, intrinsic structures of processes that correspond to human observations and knowledge of process dynamics are represented explicitly. Hierarchical structures and scale dependency among processes can also be incorporated in the semantic network. Second, it supports bi-directional information linkages so that information can be inferred from processes or locations. Because of the bi-directional mappings of process, time, and space, the representation can facilitate information queries based on either processes or locations. Based on the representation, information can be readily available to support a query about "how a supercell progresses > from t3 to t7?" or "how long location S2 has hail?"

Towards a more powerful geographic information query support in GIS The foundation for a GIS to provide an adequate query support lies in the design of data models and algorithms (Worboys 1990). Users submit queries to a GIS based on their perception and understanding of geographic worlds. Although map-based models handle static geographic phenomena well, adequate GIS query support cannot be achieved without incorporating the approaches that users take to perceive and understand geographic processes. While current GIS map-based representations cannot capture the dynamic aspects of geographic processes, improvements are needed to equip GIS representation to incorporate process behaviors to facilitate spatiotemporal information query and analysis.

The process-based representation framework in Figure 2 greatly enhances information query support for spatiotemporal behaviors of processes, including frequency, duration, movement, and rate of movement. These kinds of information are central to understand geographic processes, but are not easily handled by representation and computation capabilities in current GIS. With the process-based representation, support for frequency queries is straightforward because the number of events occurring in an area can be counted by the number of instances (links) at a location. Likewise, it also eases support for queries about duration by having processes and locations directly linked to the periods during which processes occurred. Furthermore, information about movement or the rate of movement of a process can be computed by tracing time and location objects linked to the process. More information about computing process behaviors based on the process-based representation is discussed in Yuan (1998).

In summary, cognitive models of geographic processes present a basic framework of user perception and understanding. It is also the basic framework for GIS representation to support user queries on the dynamic behaviors of these processes from geospatial databases. Four basic conceptual models of processes are elicited from surveys of users in wildfire studies. By integrating the four basic conceptual models, a process-based representation is proposed. Because the process-based representation is compatible with cognitive models of geographic processes suggested by the four conceptual models, it supports information queries on process behaviors, such as frequency, duration, movement, and rate of movement of processes. Arguably, frequency, duration, movement, and rate of movement are the basic characteristics that we use to understand the behaviors of a process. The compatibility of the process-based representation and cognitive models of geographic processes enhances GIS support for user query and analysis on information about process behaviors.

References

• Montello, D. R. (1998) A new framework for understanding the acquisition of spatial knowledge in large-scale environments. In M. J. Egenhofer and R. G. Colledge eds. Spatial and Temporal Reasoning in Geographic Information Systems, Oxford University Press, New York, New York. Pp. 143-154.
• Zubin, D. (1989) Natural language understanding and reference frames. In D. M. Mark, A. U. Frank, M. J. Egenhofer, S. M. Freundschuh, M. McGranaghan, and R. M. White eds. Languages of spatial relations: researchable questions and NCGIA research agenda, Technical Report 89-2A, National Center for Geographic Information and Analysis, University of California, Santa Barbara, CA. 1989. Pp13-16.
• Worboys M F 1990 Reasoning about GIS Using Temporal and Dynamic Logics. NCGIA Technical Report 90-4.
• Yuan, M. and Albrecht, J. 1995. Structural analysis of geographic information and GIS operations: from a user=92s perspective. In Spatial Information Theory, edited by Andrew Frank and Werner Kuhn (Berlin: Springer-Verlag), pp. 107-122.
• Yuan, M. (1997) Knowledge acquisition for building wildfire representation in Geographic Information Systems. The International Journal of Geographic Information Systems, 11(8):723-745.
• Yuan, M. (1998) Representing Spatiotemporal Processes to Support Knowledge Discovery in GIS databases. Proceedings: the 8th International Symposium on Spatial Data Handling. Forthcoming.

• 1994 Ph.D. in Geography from The State University of New York at Buffalo.
  Dissertation: Representation of Wildfire in Geographic Information Systems.
  Major Advisor: Dr. David M. Mark.
• 1992 M. A. in Geography from The State University of New York at Buffalo. Thesis Project: Use of Arc/INFO Macro Language (AML) to Construct an Interactive User Interface for City Landuse Management. Major Advisor: Dr. Hugh W. Calkins.
• 1987 B. S. in Geography from the National Taiwan University. Senior Project: A Study of Waterfall Evolution in the Upstream of the Keelung River, Taiwan.
  Major Advisor: Dr. Shing Wang.

• 1994-present Assistant Professor of Geography, University of Oklahoma, Norman, Oklahoma.
• 1990-1994 Graduate assistant and graduate teaching assistant, Department of Geography, the State University of New York at Buffalo, Buffalo, NY and the National Center for Geographic Information and Analysis.

• Yuan, M. forthcoming. Representing Geographic Information to enhance GIS support for complex spatiotemporal queries. Transactions in GIS.
• Yuan, M. forthcoming. Representing Spatiotemporal Processes to Support Knowledge Discovery in GIS databases. Program Committee of International Conference on Spatial Data Handling.
• Yuan, M. forthcoming. Modeling geographic information to support spatiotemporal queries. In Life and Motion of Socio-Economic Units edited by A. Frank, the European Science Foundation (ESF).
• Yuan, M. 1997. Knowledge acquisition for building wildfire representation in Geographic Information Systems. The International Journal of Geographic Information Systems, 11(8):723-745.
• Yuan, M. 1996. Modeling semantical, temporal, and spatial information in geographic information systems. M. Craglia and H. Couclelis eds. Geographic Information Research: Bridging the Atlantic, Taylor & Francis, pp. 334-347.