GEOGRAPHICAL INFORMATION SYSTEMS: A PERSONAL HISTORICAL PERSPECTIVE, THE FRAMEWORK FOR A RECENT PROJECT, AND SOME QUESTIONS FOR THE FUTURE Carl Steinitz Victoria and Alexander Wiley Professor of Landscape Architecture and Planning Harvard University Graduate School of Design 48 Quincy Street, Gund Hall 409 Cambridge, MA 02138 USA Presented at The European Conference on Geographic Information Systems Genoa, Italy March 30, 1993 ABSTRACT The presentation has three major parts. The first describes some of the pioneering GIS research of the Laboratory for Computer Graphics founded by Howard Fisher at The Graduate School of Design, Harvard University, in 1965. Included are some early experiments for which I share responsibility. The second part describes the "framework for theory" which now organizes my teaching and projects. It is illustrated by a recent GIS-using landscape planning study. The third part compares "then and now" and uses the framework to pose some important questions for the future of GIS. INTRODUCTION Thank you for inviting me to join this conference and to offer this plenary paper. It is particularly satisfying to see the important work related to geographic information systems (GIS) that is being done throughout Europe; some of which is by people whom I have known for a long time. My presentation will be in three major parts. First, and as I was asked to do, I will describe some of the pioneering GIS work of the Laboratory for Computer Graphics and Spatial Analysis at the Harvard Graduate School of Design between 1966 and 1968. Then, I will discuss the framework which now organizes my teaching and projects and show a recent landscape planning study. Finally, and most importantly, I will compare the "then and now" and use the framework to link this comparison to some important questions for the future of GIS. 1965 My first contact with GIS activity came in 1965 at a lunch at the Harvard-MIT Joint Center for Urban Studies during which, by chance, I was seated next to Howard Fisher. Howard Fisher was visiting Harvard while considering a move from the University of Chicago. He had recently invented SYMAP, a computer mapping program which was based on line-printer technology, and which had not yet been applied to a substantive problem. I immediately seized upon the relationship between the capabilities that he described and the needs of my doctoral thesis and convinced him to let me try some experiments with his basic program. With Fisher as my tutor, I gave SYMAP its first applied test in my study of the perceptual geography of central Boston.1 In part because of this work, I obtained my first teaching appointment as Assistant Professor at the Harvard University Graduate School of Design and as an initial research appointee to the then new Laboratory for Computer Graphics. The Laboratory for Computer Graphics was established in 1965 with a grant from the Ford Foundation to the Graduate School of Design of Harvard University. Under Howard Fisher's direction, the Laboratory assembled a group of bright, energetic, and experimentally oriented people. In a very short time, they developed several innovative methods of high-speed electronic digital computer mapping and new techniques for graphic display. These made full and efficient use of the accuracy, speed, and cost of the computers of the time. The Laboratory's research activities essentially consisted of two types. The first was investigation into the uses of graphical representation, and computer graphics in particular, and was built largely upon Fisher's SYMAP, which became, in its time, the world's most widely used computer mapping program. The second type of research undertaken by the Laboratory was pure research in spatial analysis related to city and regional planning, landscape architecture, and architecture, with emphasis on the roles of computers in programming, design, simulation, and evaluation. In addition, a strong research effort in theoretical geography was organized and directed by William Warntz and related to the theory of surfaces, the macrogeography of social and economic phenomena, and central place theory. In the fall of 1966, I taught my first course at Harvard, a collaborative regional-scale studio, and used SYMAP in a landscape planning study of DELMARVA, the Delaware, Maryland, and Virginia peninsula. To my knowledge (and certainly so, at the time), this was the first application of GIS to a large geographical region.2 Even in this first study, some rather sophisticated analytic steps were undertaken. These included a gravity model, the analysis of the effect of one map pattern upon another, and quantitatively weighted indexes, such as the relative attractiveness of the area for vegetable or grain agriculture. I cannot understate the importance of the initial academic decision made by Charles Harris to enable me to introduce GIS in a problem-oriented studio rather than in a specialized "technical" course. This would prove crucial to the future development of GIS at Harvard. As early as 1967, I designed a GIS for the State of New York at the request of the State Office of Planning.5 A pilot project was undertaken by the Center for Aerial Photographic Interpretation at Cornell University under the direction of Donald Belcher (who had seen the DELMARVA study). Belcher developed a data-gathering process which linked aerial photographic interpretation with GIS techniques6 in a data sampling and generalization approach conceptually similar to today's raster image scanning. This GIS was used for various planning projects in the Adirondack and Syracuse areas of New York. In 1967 our research group, which included Peter Rogers, Doug Way, and Richard Toth, began a series of GIS-based plan-generation and evaluation experiments. The "Honey Hill" study,3 named after its location in New Hampshire, involved a large proposed flood control reservoir and new park. GIS-based evaluation models were made of the attractiveness of this large area for recreation and other uses and of the vulnerability of the site's natural systems to harmful impacts. Then, each of us proposed his "best" plan for the new lake and park facilities. In addition, Peter Rogers used a linear-programming algorithm to produce a fiscally "optimum" plan. These alternatives were all compared in yet another model which simulated varied user- demand patterns over the site. The optimizing program performed best (my alternative was fourth). This study provided important insight into the potential power of using GIS to link different model types to make better plans; it would shape our work for years. This research concept was the inspiration for a series of studies focusing on the Boston region in the late 1960s and a major research program supported by the U.S. National Science Foundation in the early 1970s in which GIS methods were integrated with sectoral models of the processes of urbanization and change.4 Many of today's leaders in GIS were members of this generation of graduate students at Harvard. Among them were David Sinton (of Intergraph), Jack Dangermond (founder of ESRI and producer of ARC/Info), Lawrie Jordan and Bruce Rado (co-founders of ERDAS), Hans Koeppel, and Nicos Polydorides. One additional early experiment may be of interest. In 1968, I designed a series of programs which automated the process of relating a raster terrain model and a landcover map to a series of prepackaged visual simulation forms for trees, houses, etc. This program allowed one to specify the location and azimuth for view, and the program would painstakingly draw, via a pen plotter, a series of perspectives in that GIS "landscape." The system was configured so that changes in the GIS landcover map would automatically trigger changes in the landscape view. This was a successful but inefficient and uneconomical technique.7 Only recently have we efficiently linked GIS to animated visualization. Between 1966 and 1968 the Laboratory developed rapidly as a service organization to Harvard and to many "outside" projects which were used to test or demonstrate the new GIS and computer-mapping capabilities. An important policy of the Laboratory was to disseminate its programs for the widest possible use. Correspondence training in the use of SYMAP and other programs was offered, and several intensive instructional sessions were held at various universities and other organizations. The Laboratory held annual conferences (the first at Harvard in the Spring of 1967) and weekly luncheon seminars dealing with many aspects of the Laboratory's research interests for example, pattern recognition strategies. The Laboratory also produced many publications, most notably the Harvard Papers in Theoretical Geography. Many computer programs were developed, tested, and widely distributed. Among them were: SYMAP, a general purpose line printer mapping program; CALFORM, a conformal mapping program for use with a pen plotter; SYMVU, a pen plotter surface-perspective mapping program; POLYVRT, a cartographic data base manipulation program; and the beginnings of the ideas that led to ODYSSEY, a geographic data management, analysis, and display system. Some of these programs have a traceable genealogy. For example, it was clear from the onset that there were inefficiencies in the way SYMAP operated and that it was extremely difficult for people to use. The teaching of this program and its use required people to learn a computer language, FORTRAN. Many of its GIS concepts were reorganized several times. The first, GRID, was a data management program which could more efficiently structure raster-based cartographic analyses for SYMAP. It became obvious to David Sinton (then a graduate student) that SYMAP's repetitive operations approach could be much more simply organized in a keyword structure. David then conceived and developed IMGRID which we used in a wide variety of applications during the early 1970s. In the mid-1970s and while a student, Dana Tomlin saw ways in which the keyword approach of IMGRID could be changed to an English-using command language. He conducted a series of experiments which became a basis of his doctoral thesis at Yale and ultimately resulted in the Map Analysis package (MAP). Dana joined our faculty, and another group of students was prepared. These included David Hulse, who wrote MACGIS. Today, Stephen Ervin and our students continue to develop GIS applications. We are now part of a world- wide network of colleagues; we hope that our contribution to GIS will continue. In 1968, the Laboratory for Computer Graphics was reorganized on as the Laboratory for Computer Graphics and Spatial Analysis. This name change recognized the expansion of the Laboratory's intellectual goals, and William Warntz became director. I think that many of you know that the Laboratory grew rapidly in size and influence during the 1970s under the directorship of Allan Schmidt and that it ceased to exist-- for many complex reasons--in 1981. By then, 165 people had served on the Laboratory staff at one time or another. These people and their ideas, computer programs, publications, and students have been instrumental in the development of GIS to what it is today. Much of this credit should go to Howard Fisher who died in 1974 and who was a remarkable person of uncommon energy and foresight. TODAY In 1990, after almost 25 years of applying GIS to many projects, I came to the realization that there was a common structure to this work, and I wrote a short paper entitled "A Framework for Theory."8 Over the past three years, this framework has become the primary organizational basis of my teaching, projects, and research. In this second part of my lecture, I would like to give a brief description of this framework and show how it was applied to a recent GIS-using project. As a teacher, I have always believed that we must foster an integrative and adaptable approach to theory and practice. I reject the "top-down" concept of a universally applicable landscape planning model or method. Rather, I believe that an appropriate strategy results first from understanding what the questions are, and then from the "building up" of an appropriate project methodology. My search for an over-arching framework within which to organize this process derives from my experience that there is an overwhelming (and perhaps necessary) structural similarity among the questions asked by and of landscape planners and other environmental design professionals. Amos Rapoport has provided a useful definition of theories, models, and frameworks. In short, he states, "a theory explains, a model predicts, and a framework organizes. A framework can be judged on its reasonableness and its utility, but claims no exclusivity vis-a-vis other frameworks."9 My proposed framework identifies six types of questions. Each can be considered a level of inquiry relating to a theory-driven modeling type. The models on which we rely must be based in usable and valid (or presumed to be valid) theory. They each require the management of information, and GIS can be used--albeit differently--in each type of model. The framework is "passed through" at least three times in any project: first, in defining the context and scope of the project; second, in specifying the project methodology; and third, in carrying the project forward to its conclusion. The six questions with their associated modeling types are listed in the order in which they are usually considered when initially defining the context of a landscape planning study. I How should the state of the landscape be described; in content, boundaries, space, and time? This level of inquiry leads to representation models. [Note that to date, representation has been the major emphasis of GIS.] II How does the landscape operate? What are the functional and structural relationships among its elements? This level of inquiry leads to process models. III Is the current landscape functioning well? The metrics of judgment--whether health, beauty, cost, nutrient flow, or user satisfaction--lead to evaluation models. IV How might the landscape be altered; by what actions, where, and when? This is directly related to I, above in that both are data; vocabulary and syntax. This fourth level of inquiry leads to change models. At least two important types of change should be considered: change by current projected trends, and change by implementable actions, such as plans, investments, and regulations. V What predictable differences might the changes cause? This is directly related to II, above, in that both are based on information; on predictive theory. This fifth level of inquiry shapes impact models, in which the process models (II) are used to simulate change. VI Should the landscape be changed? This is directly related to III, above, in that both are based on knowledge; on cultural values. How is a comparative evaluation among the impacts of alternative changes to be made? This sixth level of inquiry leads to decision models. [Implementation could be considered another level, but this framework considers it as a forward-in-time feedback to level I, the creation of a changed representation model.] Note that the six levels have been presented in the order in which they are normally recognized. However, I believe that it is more important to consider them in reverse order, both as a more effective way of organizing a landscape planning study and specifying its method (which I consider the key strategic phase) and as a more effective educational approach. The methods for a landscape planning study should be organized and specified upwards through the levels of inquiry, with each level defining its necessary contributing products from the models next above in the framework. VI To be able to decide to propose or to make a change (or not) one needs to know how to compare alternatives. V To be able to compare alternatives, one needs to predict their impacts from having simulated changes. IV To be able to simulate change, one needs to specify (or design) the changes to be simulated. III To be able to specify potential changes (if any), one needs to evaluate the current conditions. II To be able to evaluate the landscape, one needs to understand how it works; and I To understand how it works, one needs representational schema to describe it. [Again, This has been the major GIS role.] Then, in order to be effective and efficient, a landscape planning project should progress downward at least once through each level of inquiry, applying the appropriate modeling types: I representation, II process, III evaluation, IV change, V impact, and VI decision. At the extreme, two decisions present themselves: "no" and "yes." A "no" implies a backward feedback loop and the need to alter a prior level. All six levels can be the focus of feedback; (IV), "redesign" and sensitivity analysis, is a frequently applied feedback strategy. A "contingent yes" decision (still a "no") may also trigger a shift in the scale or size or time of the study. (An example is a highway corridor location decision made on the basis of a more detailed alignment analysis). In a scale shift, the study will again proceed through the six levels of the framework, as previously described. A project should normally continue until it achieves a positive, "yes," decision. (In my area of application, a "do not build" conclusion can be a positive decision). A "yes" decision implies implementation, and (one assumes), a forward-in-time change to new representation models. While the framework looks orderly and sequential, it frequently is not so in its application. The line through any project is not a smooth path: it has false starts, dead ends, serendipitous discoveries--but the line does pass through the questions and models of the framework as I have described it, before a "yes" can be achieved. When repeated and linked over scale and time, the framework may be the organizing basis of a very complex study. Regardless of its complexity, the same questions are posed again and again. However, the models, their methods, and their answers vary according to the context. The following is an application of the framework to a recently completed and complex study entitled "Alternative Futures for the Synderville Basin, Summit County, Utah, U.S.A."10 This multi-faceted study, with many links across scale and time, was a collaboration among Harvard University Graduate School of Design (HGSD), Utah State University (USU), and The University of Toronto (UT). I The Synderville Basin, 180 square kilometers in area, is a 30-minute drive through the Wasatch Mountains from Salt Lake City, Utah. It is the first "open" and developable landscape east of the highly built-up Great Salt Lake Valley. II Today, approximately 10,000 persons live in the County, whose primary economic bases are agriculture and recreation. III Park City and Deer Valley are firmly established as World Class ski resorts. IV Salt Lake City is projected to grow from a population of 1 million to 1.5 million during the next 20 to 30 years; the County population is expected to grow by 40,000 persons during that time. There is already land speculation, and almost all private landowners in the Synderville Basin have studied and proposed options for development. V Knowledgeable persons, including the Summit County Commissioners, sense that this rapid growth will generate harmful impacts on the county. VI Despite an historically strong private property ethic and broad-based public interest in controlling growth and change is increasing. The key decision criteria would be to minimize public action, expenditures on infrastructure, and interference with the private housing market, while maintaining a high environmental quality and sense of "openness." VI-I After a site visit and orientation, my framework was used to specify the three time phases, spatial scales, and the methods which were to be used. Several technical requisites for the study were organized for sharing and exchange. Coordinated GIS data formats were agreed upon, and an efficient data acquisition was completed. Computer software and GIS data were transferred between Harvard and Utah State and Toronto both via personal exchange and via INTERNET, the computer link among most of the world's universities. Considerable use was also made of fax technology to exchange diagrams, drawings, lists, etc., and video-taped project reviews were also exchanged. I The study area representation then was completed. The GIS was specified to meet the needs of the requisite models and to be as "lean" as possible, both in categories and scale. It consisted of digital terrain, satellite-derived landcover, ownership, and another half-dozen plan-related map layers. There was a substantial non-GIS "data base" including interviews, video and photographic records, historical records, and reports. II The key study area processes were then evaluated for their vulnerability to impacts of change. III The current local building code requires evaluations for avalanche zones and flood hazard areas. Additional evaluations included potential sources of water, primary wildlife winter habitat and movement corridors, wetlands, and visual "openness" which, in this landscape, is an undeveloped foreground viewed against a mountain skyline. According to several surveys, this is the type of view that people of the Snyderville Basin value most. These and other vulnerable landscape processes are not protected by the current building regulations. IV Five alternatives for future change were developed by the studio to reflect increasing public involvement. Each was driven by market forces guided by public investment and regulation. Each also included several areas for which more detailed design studies were carried out. The existing conditions provided the base for comparison. The first alternative assumes the Development Build-out of the actual proposals of current landowners and was found to accommodate an estimated population of 80,000 persons, twice forseeable demand. The second alternative shows the Development Projection trend at 40,000 persons under the current code. Given the process by which privately owned land is subdivided, one can expect extensive coverage but low density. The Cluster Alternative uses conservation-oriented regulations and development-density bonuses to cluster development in and around the meadow area which is the heart of the Snyderville Basin. The Village Alternative is implemented by transfers of development rights among current landholdings creating a series of villages around a conserved and restored meadow area. The New Town Alternative concentrates public infrastructure investment development on one large private landholding to the east of highway U.S. 40. When the alternatives were configured, the POLYTRIMS program developed by John Danahy and others at the Centre for Landscape Research at the University of Toronto was used for producing computer-generated perspective views and animations of various viewing sequences within the future landscapes. At one point during this simulation stage, the animation programs were controlled at Harvard and simultaneously run on computers in Toronto with the results returned via INTERNET for video image capture at Harvard. V The impacts of the change alternatives were then compared according to the GIS-based environmental, social, and economic models. There was a direct relationship: the alternatives performed better as public action increased. VI In summary, there were three reasons to publicly manage growth and change in the Snyderville Basin: 1) to lessen infrastructure investment; 2) to lessen specific environmental impacts; and 3) to preserve the sense of visual openness that this landscape still offers. In the 1992 county election and influenced in part by the public presentation of our study, this decision was made, and a version of the cluster alternative was subsequently implemented. This fall, the American Planning Association gave an Award of Merit to Harvard University for the planning, preparation, publication, and implementation of "Alternative Futures for the Snyderville Basin, Summit County, Utah." 1965-1993 What has changed between 1965 and 1993? Certainly GIS computer technology has changed (several times), as have the "styles" of software, the diversity of analytic methods, the range and quality of products, and the costs of use. The contrast between some of our early studies and more recent ones is enormous. But what has not yet changed may be far more important. My view is that the next phase of GIS research and development must--refocus on the ever-present questions of theory and models, and I would like to use my framework as a guide to some of these questions. I How should the landscape be described? I began my GIS work using Holerith cards and a line printer to make paper maps in black and white. My first large-area GIS map cost $35 (in "1965 dollars") for computing time on a $2,000,000 machine, the only one at Harvard. I doubt that you could know how happy I was to produce it-- finally--after 30 days of effort. But it was graphically primitive. Today, the same data might be mapped in less than a minute by pointing to English language or graphic commands on a computer notebook, in a perspective view, textured in color and on shaded relief. It would certainly look better, but it would have absolutely no more information. Data quantity and quality have always been crucial issues to GIS. Early studies tended to be limited to previously mapped data sources which were remapped for computer use in coarse grids or coarse polygons. The data representation, a map, was presumed to provide an understanding of its underlying process. Now we routinely combine data from several sources and of differing structures. And it is clear that the ability to acquire data by remote sensing in real time has a great impact on GIS. But do we know more about the world because we can represent more data? I do not believe that research on representation models is where major emphasis should be placed.11 We now access multi-media systems and have the ability to generate perspectives in real time using combinations of computer and video technology, and as a result we can "walk through" and look at landscapes. Surely, we will see continued innovation, technical development, and increased efficiency in data capture, storage, and display, etc., and this is likely to be hardware-driven. But how much better should representations be when still only relying on the visual acuity and individually varied interpretations of the users? If this relationship is seen as a communications-theory problem, then research on representation models should focus on the end-users. II How does the landscape function? Most early GIS process models were gross simplifications; typically classifications, sieve maps, or overlay combinations--all of which could have been produced using hand-drawn methods. Later, fixed-format quantitative and spatial analysis languages were introduced that could be applied to mapped data. Then, we saw more "packaged" models which were used to conduct sectoral analyses--air quality models, water quality models, erosion models, etc. In recent years, we have seen the development of much more easily used and adaptable software packages. Now, it is perhaps too easy for users to specify their own models. It thus becomes increasingly more important to know whether the models are reliable. How do we know? How will GIS reflect uncertainty? GIS operations will need to accommodate spatially and temporally variable processes. The real world has many examples, but how much of today's GIS analysis is still simple query or sieve-mapping? How will GIS help us to understand complex process interactions over space and time? III Is the landscape working well? Early GIS studies required predefined evaluation models which were presumed to be consistent across a pre-specified geographical (or other data) subset. Frequently, these were no more than re-mapped maps or abstract indices. Then, the linking of sectoral process models, each having its own metrics of evaluation, led to the recognition of multi-dimensionality in evaluation, the "apples and oranges" problem. We have not progressed far from these early stages. Indeed, we may be faced with "evaluative overload" unless GIS methods can be developed to make multiple evaluation models comprehendible without relying on a single (and still abstract) common evaluation metric. IV How might the landscape be altered? GIS is very limiting when considering change models. It has always been applied to known and stable systems. The continued emphasis on the map as "product" reflects this assumption. In the early days of GIS, there was great tension between the "map-makers," who demanded cartographic excellence, and the "planners," who needed adequate representations. Fortunately, this is now a moot issue. But that argument dealt with representations of known data. Now, there is widespread research on the development of "smart" software which can "learn" and apply that "knowledge" to manage change. We can now represent change in real time, and we can (in part, at least) simulate the future. The technology of virtual reality exists; the future can be represented. But what future do we want to visualize? And how will GIS integrate innovation and invention, which are part of change and are not necessarily reflected in prior representation and process models? V What differences might the changes cause? Much of the funding for early GIS research and application was driven by the need to develop impact models for major capital expenditures such ashighways, airports, and urbanization. Note that the Laboratory for Computer Graphics and the (American) National Environmental Policy Act (NEPA) were established during the same period, the late 1960s. We now recognize new complexities. For example, how will GIS interact with process models to affect an understanding of the interactive, compound, and cumulative impacts of change, over time and space and possibly beyond the edges of the GIS? Think of Chernobyl as an example. VI Should the landscape be changed? How is this decision to be made? From the beginnings of GIS, there existed the belief that information could influence decisions. There was the assumption that critical professional roles would involve organizing that information, having it available and adaptable to "questions," and thus aid decision-makers by providing understandable "answers" relevant to the decisions at hand. The focus on "aiding," rather than "making" decisions was related to both professional interest and public acceptance. As early as the 1967 DELMARVA study, different "political" criteria-sets were mapped to compare conflicts and agreements regarding conservation and land development. Also among our earliest studies was a comparative study of the influences of different interest-groups on a highway location. Yet in the time of these early studies, public political (and peer-professional) acceptance was low regarding anything related to "computers." Today, in universities and among the population-at-large we see the first computer- comfortable generation. Computer use is no longer regarded as something special. Indeed, GIS use is now widely perceived to be a prerequisite for successful practice in all spatial planning and management professions. Increasingly, we think of computers as our "partners," in professional decision-making. Yet there are complexities. What decision models will interact with GIS in the future? We have decentralized, interactive, and personal "networked" computer access and increased technical freedom. But how personalized can GIS become? GIS does not deal very well with individual decisions. Qualitative judgments normally expressed in common evaluative language do not translate well. We cannot express feelings, indecisiveness, or inconsistency in GIS. Yet these are real issues that must be addressed in making decisions. How will GIS capture change at the scale of individually varied decisions without a "police-state bureaucracy"? Will GIS serve the public at large or only government, the military, and other major corporate users? Should GIS become a world-wide public utility? CONCLUSION The GIS community faces questions in all aspects of its work: in representation, process, evaluation, change, impact and decision models. Thirty years of research and innovation have provided us with a wide menu from which we can choose. The questions which I have raised are not new and they are not without past and present research. But they are basic to the future of GIS and its many user-professions and especially for those integrative, multi-professional, public activities which inspired the Laboratory for Computer Graphics and which today are still the focus of our energies. REFERENCES 1. Steinitz, C. "The Congruence of Urban Form and Activity." Journal of the American Institute of Planners, July 1968. 2. Steinitz, C. "Regional Planning and Computer Graphics." In Computer Graphics in Architecture and Design, M. Milne (ed.), Yale University, 1969. 3. Steinitz, C.; Murray, T.; Rogers, P., Sinton, D.; Toth, R., and Way, D. Honey Hill: A Systems Analysis for Planning the Multiple Use of Controlled Water Areas, U.S. Army Corps of Engineers, Harvard University Graduate School of Design, 1971. 4. Steinitz, C.; and Rogers, P. A Systems Analysis Model of Urbanization and Change: An Experiment in Inter-disciplinary Education. MIT Press, December 1970. 5. Steinitz, C.; Sinton, D.; and Belcher, D.; et. al. New York State Natural Resources Inventory, Office of Planning Coordination, Tthe State of New York, 1968. 6. Steinitz, C. "Landscape Analysis." Landscape Architecture, January 1970. 7. Steinitz, C. "The PERSBLOCK Programs," in Tomlinson, R.F. (ed.), Geographical Data Handling, UNESCO/International Geographical Union, Ottawa, Canada, 1972. 8. Steinitz, C. "A Framework for Theory Applicable to the Education of Landscape Architects (and Other Environmental Design Professionals)," Landscape Journal, October 1990. See also: Steinitz, C. "Toward A Sustainable Landscape Where Visual Preference and Ecological Integrity are Congruent: The Loop Road in Acadia National Park," Landscape Planning, Vol. 19, No. 1, 1990. 9. Riley, R. "Editorial," Landscape Journal, Spring 1991, p. 49. 10. Steinitz, C. (ed.), et. al. "Alternative Futures for the Snyderville Basin, Summit County, Utah," Harvard University Graduate School of Design, 1991. 11. Steinitz. C. "Some Words of Caution," Landscape and Urban Planning, 21, 1992. 12. Ervin, S. "Landscape Planning and Design: Computing Across Ranges of Scale and Levels of Abstraction." Journal of Architectural and Planning Research (JAPR), Special Issue [month unknown], 1993, Ed., M. Gross. (In press.) I thank Stephen Ervin, Irene Fairley, Marybeth Flaherty and TenBroeck Patterson for their advice and assistance on this paper and also many colleagues and students (too many to name herein) for their collaboration on its contents. Carl Steinitz Cambridge, MA