In the first part of this paper, the author presents a set of 20 universal analytical GIS operations, i.e. operations that are independent of data structures, yet cover the full range of analytical capabilities of current vendor-based GIS. This set represents a user's task-oriented view of GIS functionality rather than that of a technically oriented developer. Their function is readily apprehended by any spatially aware individual and does not require any knowledge about abstract spatial concepts. These operations constitute the foundation of a shell that has been developed at the Institute for Spatial Analysis and Planning in Areas of Intensive Agriculture (ISPA) and goes by the name Virtual GIS (VGIS). VGIS provides an ideal tool box for the environmental modeler, who wants to concentrate on modeling issues rather than the intricacies of GIS.
In the second part, the flow chart-based user interface of VGIS is introduced as a visual programming and prototyping tool similar to STELLA® (HPS, 1994), but working with real GIS data and offering the full functionality of GIS. The conceptual modeling capabilities are exemplified by applications from environmental modeling.
The domain of universal GIS methods is an area of Geographic Information Science (Goodchild, 1992a) that so far has been neglected by research and for which therefore no body of theory exists. Driven by the heterogeneous market forces every vendor and a multitude of academic developers produced myriads of commands to perform GIS operations. The few existing taxonomies of GIS operations (Aronoff, 1991; Burrough, 1989 and 1992; Goodchild, 1992b; de Man, 1988; Rhind and Green, 1988; Unwin, 1990; Schenkelaars, 1994) are limited either by the data structure that they are based on or by the scope of applications for which they had been developed. They all lack formalization and do not attempt to be truly universal. Section 2 introduces the author's work on a task-oriented systematization of data structure-independent GIS functionality.
On the ecological modeling side, Constanza and Sklar (1985) state that "most ecological modeling work to date has focused on temporal changes. They tend to simulate a point in space and extrapolate the findings for an entire landscape by assuming that the landscape is homogeneous. In other words, most models in ecology have little, if any, spatial articulation. It is clear, however, that space needs to be more explicitly included if ecological models are to be truly useful tools for understanding and predicting the behavior of real ecosystems (Risser, et al., 1984)." Although this quote is 10 years old now, it has not lost any of its truth. Section 3 reviews the current state in GIS-based ecological modeling.
Section 4 tries then to apply the universal GIS operations described in section 2 in an environmental modeling context. Special emphasis is given to the phase of conceptual modeling as well as to an object-oriented application of individual modeling units. The conclusion (section 5) describes current limitations to the proto-type and discusses some rather fundamental ideas about the synthesis of GIS and environmental modeling.
One motivation for research on universal GIS operations simplifying that simplify GIS use was the observation, that in spite of its name, current GIS have little to offer to the scientist, who is interested in modeling spatial phenomena. Tomlin's (1990) cartographic modeling language (also known as 'Map Algebra') is the most sophisticated GIS modeling environment so far. This lack has been articulated and mourned by many in the modeling community and resulted in a conference series, devoted to overcome this discrepancy between the GIS and the modeling community (Goodchild et al., 1993; Goodchild et al., 1995).
Current Geographic Information Systems (GIS) are so difficult to use that it takes some expertise to handle them, and it is not unusual to assess a whole year until an operator masters a GIS. This is especially cumbersome for cursory users (such as environmental modelers) that employ GIS as one tool among many others. Coulsen, et al., (1991) expressed a similar argument when they wrote "GISs are complex computer programs. Proficient application in natural resource management and landscape ecology involves a commitment to training and practice by the user. None of the GISs would be considered user friendly by a human factors engineer." Similar comments may be found throughout (Medyckyj-Scott and Hearnshaw, 1993) and (Turk, 1992).
Although a number of GIS claim to be data structure-independent none of them really is; they all show their origin as so-called either raster (grid cell-based) or vector systems. This data structure distinction has dictated differences in analysis functionality. Even Goodchild (1991a, p. 45) in his often cited classification of spatial data analysis techniques groups them depending on the underlying data model. And while there are numerous efforts to standardize data models (SAIF, SDTS, DIGEST, GDF), so far none of these attempt to standardize the operations as well. The advent of the Open Geodata Interoperability Specification (OGIS) (Buehler, 1994; Buehler, 1995) opens for the first time a real opportunity to develop data structure-independent GIS applications (see also the contribution by Kenn Gardels in this volume). As far as is known so far, however, the Open GIS Consortium (OGC) does not attempt to define high level operations but restricts its specification to low level database (SQL-like) and topological operations based on the work by Egenhofer (1991, 1993).
The author tackled these deficiencies with a user survey to determine user expectations of a GIS' functionality (Albrecht, 1995b). The responses reveal a vast array of complexity ranging from elementary operations to compound tasks. A dissection of the latter into fundamental primitives leads towards a normalization of GIS operations. Since the analytic capabilities of GIS are the only ones that distinguishes them from other visualization software or database management systems (Burrough, 1986; Goodchild, 1987), further consideration of GIS functionality within this paper will concentrate on this ability.
By analyzing current GIS user interfaces and omitting all those operations that are due to either the historic development of the particular software package or are a result of the data model employed, the author derived a list of only 20 universal GIS operations that allow to build all but the most exotic GIS applications (see also Table 1). This small set of spatial analytical tools provides the means to perform environmental spatial modeling without having to learn about the intricacies of current GIS. A detailed description of how this particular set of GIS operations was derived is given in (Albrecht, 1996). There, the reader will also find a formalization of these operations based upon a simplified version of the OGIS data model. An implementation of these fundamental GIS operations within an interactive flow-charting environment (see section 4) reveals the window of opportunities opened by this approach.
Table 1. The 20 universal GIS operations
Search:
Interpolation Thematic Spatial Search (Re-)classifi
Search cation
Location Analysis:
Buffer Corridor Overlay Thiessen/Voro
noi
Terrain Analysis:
Slope/Aspect Catchment/Bas Drainage/Network Viewshed
ins Analysis
Distribution/
Neighborhood: Cost/Diffusion/ Proximity Nearest Neighbor
Spread
Spatial Analysis:
Multivariate Pattern/Dispe Centrality/Connect Shape
Analysis rsion edness
Measurements:
Measurements
"GIS can be used to seduce the user into an unrealistic
sense of model accuracy", using "a few poor, anemic
point measurements".
(Grayson, et al., 1993, p. 91)
One motivation for the search for the GIS usage simplifying universal GIS operations was the observation, that in spite of its name, current GIS have little to offer to the scientist, who is interested in modeling spatial phenomena. Tomlin's (1990) cartographic modeling language (also known as 'Map Algebra') is the most sophisticated GIS modeling environment so far. This lack has been articulated and mourned by many in the modeling community and resulted in a conference series, devoted to overcome this discrepancy between the GIS and the modeling community (Goodchild, et al., 1993 and Goodchild, et al., 1995).
The two probably most exhaustive overviews (Sklar and Constanza, 1991; Hunsaker, et al., 1993) describe numerous environmental tasks such as inventory, assessment, management, and predicting the fate of environmental resources supporting applications in atmospheric modeling, hydrological modeling, land surface-subsurface modeling, ecological systems modeling, plus integrated environmental models as well as policy considerations for risk/hazard assessment involving these models (see Table 2). All of these, however, use a GIS as an inventory for spatially referenced data and for presentation (map production) only.
Nyerges (1991) identified three primary modes of GIS use, namely map mode providing referential and browse information, query mode to address specific requests for information based either on field or object views, and model invocation which is the only mode that makes use of the analytical capabilities of a GIS. On a rather abstract level Nyerges describes the development of a typical modeling process, however, he fails to note that up to now there exists no GIS that actually supports such a procedure.
Two research projects focus on the problem of facilitating a computational modeling environment employing rather different approaches. The first one to be described here is the computational modeling system (CMS) developed at the Department of Computer Science at the University of California (Smith, et al., 1995). Their computational modeling language (CML) is supposed to support cooperative (geographic) modeling activities at all stages of the modeling process, i.e. data extraction, construction and evaluation of conceptual models, model refinement, and communication of the results (Alonso and Abbadi, 1994). The CML is based upon the concept of so-called representational (or -) structures and their transformations. These -structures contain specifications how to represent the same information using a different data model, so the user does not need to explicitly know about the data model that the source data is based on. Each -structures then also contains the operations that can be applied on it, e.g. a digital elevation model (DEM) may contain the transformations 'union' to combine several DEMs, 'compute-slope', or 'max-height'. Creating such schemas nevertheless requires to learn a new programming language which might be worth the effort for a some model builders, but renders it unlikely that CMS will become a wide-spread tool.
The other research project aimed at facilitating modeling procedures within a GIS is the Virtual GIS (VGIS) project described in Albrecht (1995a). VGIS is a shell that employs the universal GIS operations described in section 2 using a flow charting environment (see Figure 1). Flow charts are a the standard process-oriented tool in visual programming (Chang, 1990; Glinert, 1988; Monmonnier, 1989). As in the case of CMS, the system has yet to prove its usability in real world (meaning non-academic) applications. Proposed (and already granted, see section 6) extensions of this work promise a wider applicability, as one of the interpreters that is intended to be developed for VGIS will be an interface to the OGIS data model.
So far, integrated application of GIS and environmental modeling have been specialized modeling languages, such as PCRaster and DYNAMO in the hydrologic domain (van Deursen and Kwadijk, 1993; Wesseling and v. Deursen, 1995). These languages, however, do not really support the creative process of model building. Rather, they require an intricate knowledge of the model and the language, and are harnessed to fine-tune a fixed model run. The Virtual GIS (VGIS) on the other hand, attempts to be a prototyping tool and development platform similar to STELLA® (HPS, 1994), but working with real GIS data and thereby graphically extending 'Map Algebra' according to the concepts presented in (Kirby and Pazner, 1990). More realistic (real world) data with a locational character have a significant impact on the model results. In addition, geographical displays interactively depicting the nature of the sensitivity of certain parameters can be useful in support of model parameterization. examining scale effects can be accomplished by changing interactively the nature of the data aggregation. The model brings together the locational, temporal, and thematic aspects of phenomena in a geographic process characterization.
Such a visual programming example is depicted in Figure 2, where the modeling flow chart allows the user to "play" with the data flow. Figure 2 represents an intermediate step in the conceptual modeling of erosion. The four input files (rounded boxes) are geology, landcover, precipitation, and elevation. It is possible to model this complex system with only five of the universal GIS operations. Within VGIS, it is easy to test the result of new routing paths within the flow chart. The hypothesis that a certain region buffered around drainage channels has a different water retention capacity can easily be tested by adding one connection to the flow chart. A similar reconfiguration of a conceptual model would require substantial GIS expertise if it were attempted in a vendor GIS. The similarity with the previously mentioned prototyping tool extends to another feature as well: each data object might explode into another model that can either be treated as a black box or specified in a similar manner as its parent level.
Probably the main advantage of the VGIS environment lies in the first time-ever possibility to easily include feedback loops within a GIS. "Landscapes are never static; their elements are in constant temporal and spatial flux" (Merriam, et al., 1991). Sklar and Constanza (1991) therefore consider the incorporation of space as well as time as the most prominent issue in ecosystem research. This needs to be done at all levels of resolution that are meaningful to the myriad ecosystem management problems we now face. It is this explicitly spatial aspect is what motivates landscape ecology.
Another major advantage of the VGIS environment is the inclusion of spatial statistics as GIS functionality. "To consider something a system, it is necessary to describe its boundary and its interaction with the environment" (Frank, et al., 1994). Therefore, one of the first tasks within a modeling environment should be to delineate the spatial boundaries. This can be readily accomplished by calling the 'Pattern/Dispersion' analysis operation. Although this is one of the core functions of a GIS, the author has yet to come across a reference for an environmental modeling application where a GIS is actually used for this purpose.
VGIS is implemented as an interpreter to the public domain GIS GRASS. Current work focuses on the development of another interpreter for Arc/Info®. The only handicap for a free distribution of the prototype is the utilization of the flow charting tool WiT®. A detailed technical description is given in (Brösamle, et al., 1996). The VGIS environment does not yet include conditional and iteration operators as they are used in formal programming languages. Therefore, in its current state, the VGIS can not be called a geographic modeling language yet.
One of the main impediments to the application of VGIS in real world applications has been the difficulty to environmental modeling tasks that require the full scope of analytical GIS functionality. Most examples would do well with a minimal set of 'overlay', 'slope calculation' and 'diffusion/spread'. The latter two are often implemented in cell-based modeling systems as well. This experience comes very much as a surprise to a geographer who started out with conviction that space is such an important feature in any environmental model that the full power of GIS will be required to satisfy even minimal scientific deeds. Now the far more modest author recognizes that there is still a long way to go, before we will understand the rules governing landscape ecology before we can implement a modeling system on a truly large scale.
This research is supported by the German Science Foundation which provides the funding for the Virtual GIS (VGIS) project under grant number IIC5-Eh 85/3-2.
Albrecht, J., 1995a. Virtual Geographic Information System (VGIS). In Nunamake, J. and R. Sprague (Eds): Proceedings 28th Hawaii International Conference on System Sciences, GIS Minitrack, Volume IV, pp. 141-150. IEEE Computer Society Press, Los Alamitos, CA.
Albrecht, J., 1995b. Semantic Net of Universal Elementary GIS Functions. Proceedings ACSM/ASPRS Annual Convention and Exposition Technical Papers, Vol. 4 (Auto-Carto 12), pp. 235-244. Bethesda, MD.
Albrecht, J., 1996. Universal GIS Operations. Dissertation, published electronically and accessible as http://www.ncgia.ucsb.edu/~jochen/diss/dissabst.html.
Alonso, G. and A. Abbadi, 1994. Cooperative Modeling in Applied Geographic Research. International Journal of Intelligent and Cooperative Information Systems, 3(1):83-102.
Aronoff, S., 1991. Geographic Information Systems: A Management Perspective. WDL Publications, Ottawa, 2°, 294 pp.
Brösamle, H., Albrecht, J. and M. Ehlers, 1996. GIS Functionality and Interface Design: The Virtual GIS (VGIS) Example. Invited paper, ISPRS Commission II Working Group II/2, Workshop on New Developments in Geographic Information Systems, Milan, Italy.
Buehler, K., 1994. OGIS Project Document 94-019. OGIS Geodata Model Overview. The Open GIS Foundation, Cambridge, MA.
Buehler, K., 1995. Open Geodata Interoperability Specification (OGIS). Presentation given by Kurt Buehler at the OGIS workshop in Zürich, Switzerland, 30 November 1995.
Burrough, P., 1986. Principles of Geographic Information Systems for Land Resources Assessment. Monograph on Soil and Resources Survey, No. 12. Clarendon Press, Oxford. 194 pp.
Burrough, P., 1992. Development of Intelligent Geographical Information Systems. Internatinoal Journal of Geographic Information Systems, 6(1):1-11.
Chang, S., 1990. Principles of Visual Programming Systems. Prentice-Hall, Englewood Cliffs, NJ.
Constanza, Sklar, 1985. Articulation, Accuracy, and Effectiveness of Mathematical Models: a review of freshwater wetlands applications. Ecological Modeling, 27(1):45-68.
Coulsen, R., Lovelady, C., Flamm, R., Spradling, S. and M. Saunders, 1991. Intelligent Geographic Information Systems for Natural Resource Management. In Turner, M. and R. Gardner (Eds.): Quantitative Methods in Landscape Ecology, pp. 153-172, New York, Springer-Verlag.
de Man, E., 1988. Establishing a geographic Information System in Relation to its Use. International Journal of Geographic Information Systems, 2(3):257.
Egenhofer, M. and R. Franzosa, 1991. Point-Set Topological Spatial Relations. International Journal of Geographical Information Systems 5(2):161-174.
Egenhofer, M., Sharma, J. and D. Mark, 1993. A Critical Comparison of the 4-Intersection and the 9-Intersection Models for Spatial Relations: Formal Analysis. Proceedings, Eleventh International Symposium on Computer-Assisted Cartography (Auto Carto 11), pp. 1-11.
Frank, A., Egenhofer, M. and D. Hudson, 1994. The Design of Spatial Information Systems. Part 1: Formal Systems. File on ftp://grouse.spatial.maine.edu/pub/SurveyEng/sve451/451.ps.
Glinert, E. (Ed.), 1988. Visual Programming Environments: applications and issues. IEEE Computer Society Press, Los Alamitos.
Goodchild, M., 1991. Spatial Analysis with GIS: problems and prospects. Proceedings GIS/LIS'91, pp. 41-48. ASPRS/ACSM, Falls Church, VA.
Goodchild, M., 1992a. Geographical Information Science. International Journal of Geographic Information Systems, 6(1):31-46.
Goodchild, M., 1992b. Spatial Analysis Using GIS: a seminar workbook, pp. 40-48, 2°, National Center for Geographic Information and Analysis, University of California, Santa Barbara.
Goodchild, M., Parks, B. and L. Steyaert (Eds.), 1993. Environmental Modeling with GIS. Oxford University Press, New York/Oxford.
Goodchild, M., Steyaert, L., Parks, L., Crane, M., Johnston, C., Maidment, D. and S. Glendenning, 1995. GIS and Environmental Modeling: Progress and Research Issues. GIS World, Inc., Ft. Collins, CO.
Grayson, R., Blöschl, G., Barling, R. and I. Moore, 1993. Process, Scale and Constraints to Hydrological Modelling in GIS. In Kovar, K. and H. Nachtnebel (Eds.) Application of Geographic InformationSystems in Hydrology and Water Resources Management. Proceedings HydroGIS 93. Vol. 211, pp. 83-92, IAHS Press, Wallingford, England.
High Performance Systems (HPS), 1994. STELLA II: an introduction to systems thinking. Hanover, NH.
Hunsaker, C., Nisbet, R., Lam, D., Browder, J., Baker, W., Turner, M. and D. Botkin, 1993. Spatial Models of Ecological Systems and Processes: The Role of GIS. In Goodchild et al., (1993), pp. 248-264.
Kirby, K. and M. Pazner, 1990. Graphic Map Algebra. In Brassel/Kishimoto (Eds.): Proceedings of the 4th International Symposium on Spatial Data Handling, Vol. 1, pp. 413-422.
Medyckyj-Scott, D. and H. Hearnshaw, 1993. Human Factors in Geographic Information Systems. Belhaven Press, London, 266 pp.
Merriam, G, Henein, K. and K Stuart-Smith, 1991. Landscape Dynamics Models. In Turner, M. and R. Gardner (Eds.): Quantitative Methods in Landscape Ecology, pp. 399-416, Springer, New York, NY.
Monmonnier, M., 1989. Graphic Scripts for the Sequenced Visualization of Geographic Data. Proceedings GIS/LIS'89, pp. 381-389. ASPRS/ACSM, Falls Church, VA.
Nyerges, T., 1991. Analytical Map Use. Cartography and Geographic Information Systems, 18(1):11-22.
Rhind, D. and N. Green, 1988. Design of a Geographical Information System for a Heterogeneous Scientific Community. International Journal of Geographic Information Systems, 2(2):175.
Risser, P., Karr, J. and R. Forman, 1984. Landscape Ecology: Directions and Approaches. Special Publication No. 2. Illinois Natural History Survey, Champaign, IL (qf. Sklar and Constanza, 1991).
Schenkelaars, V., 1994. Query Classification, a First Step Towards a Graphical Interaction Language. In Molenaar, M. and S. de Hoop, (Eds.): Advanced Geographic Data Modeling. Netherlands Geodetic Commission, Vol. 40, pp. 53-65.
Sklar, F. and R. Constanza, 1991. The Development of Dynamic Spatial Models for Landscape Ecology: a review and prognosis. In Turner, M. and R. Gardner (Eds.): Quantitative Methods in Landscape Ecology, pp. 239-288, Springer, New York, NY.
Smith, T., Su, J., Abbadi, A., Agrawal, D., Alonso, G. and A. Saran, 1995. Computational Modeling Systems. Information Systems, 20(2):127-153.
Tomlin, D., 1990. Geographic Information Systems and Cartographic Modeling. Prentice-Hall, Englewood Cliffs, NJ.
Turk, A., 1992. GIS Cogency: cognitive ergonomics in geographic information systems. Unpublished dissertation, Department of Surveying and Land Information, University of Melbourne.
Unwin, D., 1990. A Syllabus for Teaching Geographical Information Systems. International Journal of Geographic Information Systems, 4(4):461-462.
van Deursen, W./Kwadijk, J., 1993. RHINEFLOW: an Integrated GIS Water Balance Model for the River Rhine. In Applications for Geographic Information Systems in Hydrology and Water resources. IAHS Publications No. 211, pp. 507-518.
Wesseling C., and W. van Deursen, 1995. A Spatial Modelling Language for Integrating Dynamic Environmental Simulations in GIS. Proceedings Joint European Conference and Exhibition on Geographical Information, Vol. 1, pp. 368-373.