December 5-6, 1997, Santa Barbara, California

Michael F. Goodchild, University of California, Santa Barbara

Max J. Egenhofer, University of Maine

Robin Fegeas, U.S. Geological Survey

Geographic information systems have been adopted widely over the past two decades in support of planning, forestry, agriculture, infrastructure maintenance, and many other fields. Each software product developed essentially independently, with little in the way of overarching theory or common terminology. As a result, it is very difficult for different systems to share data, for users trained on one system to make use of another, or for users to share procedures developed on different systems. The term ‘interoperability’ suggests an ideal world in which these problems would disappear, or at least diminish significantly, as a result of fundamental changes in design, approach, and philosophy.

The Varenius project is an effort by the U.S. National Center for Geographic Information and Analysis, with funding from the National Science Foundation, to stimulate advances in certain key strategic areas of geographic information science. This document reports on a specialist meeting held under the auspices of the project, in Santa Barbara, California in December 1997, to assess the state of research in GIS interoperability, define research needs, and develop a research agenda. The workshop was held immediately after Interop ‘97, an international conference on interoperating GISs—this juxtaposition of a conference and a workshop on the same topic allowed many of the ideas to be presented and discussed before the workshop began, and led to greater focus.

The report begins with a general discussion of the nature of interoperability, and the consequences of progress toward its goals. This is followed by sections that review the meeting’s efforts to define the conceptual framework of interoperability, and appropriate theory; identify efforts that could help build the organizational infrastructure for research; outline research needs; and identify specific research topics. The report also includes a selection of position papers provided by the meeting participants in an appendix.

Interoperability means many things to people. It means openness in the software industry, because open publication of internal data structures allows GIS users to build applications that integrate software components from different developers, and it allows new vendors to enter the market with competing products that are interchangeable with existing components, just as the concept of interchangeable parts helps competition in the automobile industry. In the past few years the Open GIS Consortium (OGC) has emerged as a major force in the trend to openness, as a consortium of GIS vendors, agencies, and academic institutions (http://www.opengis.org). Interoperability also means the ability to exchange data freely between systems, because each system would have knowledge of other systems’ formats. Exchange standards such as the Spatial Data Transfer Standard (also known as Federal Information Processing Standard 173; Morrison 1992) have had a significant impact on the ease with which data can be transferred between systems. They allow a user of one vendor's products to make use of data prepared using another vendor's products, because data can be transferred in a standard format. Interoperability also means commonality in user interaction, as system designers build interfaces that can be customized to a ‘look and feel’ familiar to the user. Thus one of the important tasks of the specialist meeting was to achieve some degree of agreement on the precise significance of interoperability, and on a conceptual framework that could be used to bring all of these disparate meanings into some degree of cohesion and uniformity.

Simplification is a common theme in discussions of interoperability—simplification in the complex collections of formats and standards in the industry, simplification in the interaction between user and system, simplification in the knowledge a user requires to be effective. In an interoperable world the user would have to know less in order to achieve the same outcome. Training on ARC/INFO would not be wasted if the user transferred to an Intergraph platform, and there would be no need to master the complex details of data formats in order to assemble a project database from different sources. From an educational perspective, progress in interoperability would be measured by what it was no longer necessary to teach.

The term transparency is used when user no longer needs to be aware of the details of a computer implementation to use it effectively. A database management system offers transparency to its users, who need to know nothing about the actual implementation of a data model, or about the physical locations of data and software, but can work instead at a conceptual level. Transparency implies that certain things are no longer important to the user, and no longer intrude upon the user’s conceptualization of the problem. It implies a uniform view of multiple, heterogeneous, distributed, and autonomous participating systems.

Another term with particular relevance to interoperability is similarity, a measure of the degree to which two data sets, software systems, disciplines, or agencies use the same vocabulary, follow the same conventions, and thus find it easy to interoperate. Currently, interoperation is possible only over the narrowest of domains. The effort to achieve interoperability is thus an effort to extend domains, or to raise the threshold of similarity below which interoperability is possible.

The current architecture of GISs requires its users to be specialists, who must learn a terminology that is largely system-specific, a user interface that is similarly dominated by details of implementation, and a world of data that is riddled with convention. In order to make use of today’s GIS one must be a spatially-aware professional (SAP). The ability to decode acronyms is one of the tests of an SAP, who must know, for example, the meaning of all of the Dxx acronyms—DEM, DTM, DLG, DRG, DOQ, DCW (digital elevation model, digital terrain model, digital line graph, digital raster graphic, digital ortho-photo quadrangle, digital chart of the world respectively)—and their general characteristics. SAPs hold much of the metadata of the common data sets in their heads, and thus are able to locate necessary data and assess its fitness for use without use of the apparatus normally required to support information retrieval, such as directories, catalogs, and libraries. SAPs will have taken courses in GIS, or may have acquired their awareness through the use of software, attendance at conferences, or from the published literature.

SAPs know the conventions of the geographic information community, and its language. They know, for example, the conventions that allow the producers of DOQs to assert that the representative fraction of their product is 1:12,000 (Goodchild and Proctor 1997). This has no relationship to representative fraction as normally defined for paper maps, since there is no ‘distance’ in a digital database that can be compared to distance on the ground. Rather, a DOQ has a ‘scale’ of 1:12,000 because its positional accuracy, which is well-defined, matches that of a map at that scale, according to national map accuracy standards (http://mapping.usgs.gov/www/ti/DOQ/doqpt1.html).

Perhaps most importanty, SAPs know the conventions of GIS discretization, which maps real-world objects and fields to their digital equivalents. The object conceptualized by a user as a continuous line is discretized as a polyline, consisting of mathematically straight connections between discrete points. Similarly, by convention an area is discretized as a polygon, and may even be referred to as such by an SAP. A field is discretized in many different ways that are embedded within distinct suites of software. Thus the same concept, a continuous surface of elevation, that is widely understood across many disciplines and professional cultures, may be represented in GIS as a TIN (triangulated irregular network, or triangular mesh), the digitized contours of a DLG, or the regular grid of a DEM. While the conceptual schema is the same, the implementations are entirely different, and are never hidden from the GIS user, ensuring that GIS is essentially inaccessible as a tool to anyone other than an SAP. Six distinct implementations of fields exist in GIS (Goodchild, 1992), and several others are commonly used in finite element models (Segerlind 1976). Kemp (1997a,b) and Vckovski (1997) have argued that an interoperable world would have just one conceptualization of a field, and that many aspects of the actual implementation can be made transparent to the user.

This view of an SAP is deliberately narrow, and overlooks the understanding that any user of GIS needs of the basic underlying principles of geography and geographic representation. Any user of GIS needs to be aware of the concept of scale, for example, and this need will never disappear however much progress is made on interoperability and ease of use. If the term "spatially-aware professional" connotes someone who is aware of the basic underlying concepts of geographic information, instead of someone trained in the lower-level details of GIS implementation, as suggested above, then the trend towards interoperability, easier use, and wider adoption will create greater demand, rather than less. In a world of ubiquitous GIS, everyone will need an understanding of the basic principles by which the real world is represented in digital form.

Interoperability conceived in this way is clearly relevant at many levels, and in many different aspects of GIS. Many different types of detail can be made transparent to the user, and many aspects of implementations can be hidden. Efforts are needed on many fronts, and many conceptual and technical problems will have to be solved, before much progress can be made towards the goal of interoperability. That progress might be measured by ease of use, represented by the amount of training needed to accomplish a certain task, or by the number of user actions required. It might be measured less directly in terms of redundancy of instruction, as items in the GIS curriculum that now must be covered before students can make effective use of GIS become redundant, or at least relegated to classes that focus on the technical details of GIS, rather than its applications. Other suitable metrics might be based on the transferability of knowledge, measuring the effort required by someone trained on System A to achieve the same productivity on System B. Progress might also be measured by comparing across disciplines, or application domains.

Many recent developments in information technology and GIS are immediately and obviously relevant to interoperability, either by motivating interest in achieving its objectives, or by providing the technical means to do so. The Internet and its applications, particularly the World Wide Web (WWW), are driving much of the interest in interoperability, because they make transfer of data and software possible, but fail to resolve many of the more difficult issues that transfer raises. Developments in distributed systems, client/server architectures, digital libraries, and other related areas are also high on the technical agenda at this time. Thus it was clear at the meeting that any research agenda in interoperating GISs would necessarily include much of the research agenda of information technology, or be strongly related to it; and that a conference and workshop focusing on interoperating GISs would attract substantial interest.

The following sections of this report are organized as follows. The next section addresses conceptual frameworks for research, with particular emphasis on layer models. This is followed by a section on semantics, which the workshop agreed would be among the hardest of the research problems presented by interoperability. Section 4 presents a vision for interoperating GISs, by detailing some of the properties the workshop agreed an interoperating GIS should have, and that might be achievable in a given number of years. Section 5 addresses the infrastructure of research, and mechanisms that might foster collaboration between the academic research sector and the GIS vendor community. Section 6 examines education, and the implications of interoperation for advancing the cause of GIS education on an international basis. Section 7 discusses the measurement of progress, and metrics of the difficulty of achieving interoperability in specific contexts. Finally, Section 8 presents the workshop’s ideas for specific research topics that could be examined within the next few years by the research community, and which if addressed could result in significant progress towards the goal of general GIS interoperability.

Issues of interoperating GISs can be assigned to three distinct layers: technical, semantic, and institutional. At the technical level, where interoperability is easiest to achieve, they address issues of format compatibility, the removal of details of implementation from the user’s conceptualization of a problem, and the development of languages or user interfaces that are common across different vendor systems. At a more abstract level, the ability to transfer data from one system to another does not guarantee that the data have meaning to the new user; interoperability also requires the sharing of meaning, so that the bits that are now acceptable to the new system are also meaningful to its users, and furthermore that the two sets of meanings are identical. A given set of bits might mean a set of coordinates with respect to the North American Datum of 1927 (NAD 27) to one user, but interoperation would fail if a new user interpreted them incorrectly as being with respect to the North American Datum of 1983 (NAD 83). In essence the term set of coordinates is not sufficiently well-defined, so that both interpretations—NAD 27 and NAD 83—are consistent with it. In another example, the term wetland is not sufficiently precise to imply a rigorous definition, allowing several different agencies to claim to be mapping wetland using different procedures. The semantic level of interoperability addresses these issues of shared meaning, and is clearly more problematic than the technical level.

Finally, interoperability poses issues at the institutional level, which may be the most problematic of all. Although interoperability may appear to be well-motivated, it is not at all clear that it is always desirable. The willingness to achieve interoperability may depend on many factors, including:

• behavioral factors, including willingness to be open and to share, and incentives that may be offered to modify behavior, pitted against traditional concepts of ‘turf’ and the desire to protect interests;
• economic factors, including the added cost of achieving interoperability, particularly across very different domains, compared to the benefits and value added by interoperability, which may or may not be expressible in economic terms;
• legal issues, including the legal right to know, which is enshrined in law with respect to some aspects of the citizen’s relationship with government, versus the right to privacy and the protection of intellectual property, which is also enshrined in law;
• organizational issues, including the impacts of technological change on institutions that have been designed to achieve certain ends.

The willingness to interoperate and to share clearly varies widely between agencies, organizations, and individuals, and appears to be of special interest and complexity in the case of geographic data (Onsrud and Rushton 1995). Geographic data often form the common framework on which other activities rely, ensuring that many departments in a given organization will require access to the same basic information. Agencies whose geographic jurisdictions overlap, such as counties within a state, or states within a nation, will have reason to share geographic data, as will agencies whose jurisdictions share a common boundary.

The willingness to share may be strongest at the highest levels of government, where principles of Jeffersonian democracy have a powerful legacy. In the private sector it conflicts in part with the principles of competition, although by and large the movement to interoperability is supported strongly by the GIS vendor community, and seen to be conducive to healthy competition rather than to conflict with it.

Interoperation is an attribute of two systems, measured by the degree to which users, data, software, and other commodities can be transferred between them. A subset S of all systems is interoperable if all pairs of systems within the subset are interoperable. But the relationship between S and the set of all systems S might have several interesting aspects. For example, S might be geographically defined, if all of the systems within some region were interoperable; interoperability thus can range from local to global. It might be hierarchically defined, if systems at some level of the administrative hierarchy were capable of interoperating. For example, interoperability might exist at the level of the U.S. Federal government, or within the government agencies of a county.

The workshop devoted considerable discussion to possible conceptual frameworks that might help the group build a useful research agenda. A framework of layers was found to be the most appropriate option. In one scheme (Table 1) layers were ordered by levels of abstraction, or the degree to which users interacting at that level need to be aware of details of implementation. At the lowest level, users interact with and are aware of aspects of the engineering, including details of the communication network and hardware. At the next level, users are aware of the digital technology, including the platforms on which the systems run, but details of the engineering are hidden. At the next level, users interact through computational architectures, and details of the technology and engineering do not intrude. This is the level, for example, of the user of the Netscape browser, who knows that the software will run on virtually any platform and operating system. At the fourth level, interaction with users is at the conceptual level, and no aspect of the technology intrudes; in essence, the user is unaware of the technology underlying the application. Finally, at the highest level the user interacts at the level of the enterprise, and individual conceptualizations are transparent. At this level the user in Department A can have his or her conceptualization, which is supported by the same system that supports users in Department B, with a quite different conceptualization.

The scheme shown in Table 1 is organized by levels of abstraction, but these correspond closely to levels of scale. The top level of the enterprise is the most abstract, but also the most extensive; the bottom level of enginering is the most concrete, but also the most detailed.

After extensive discussion the group arrived at consensus on the layer scheme shown in Table 2, which combines several organizing dimensions, including scale and abstraction, into a single coherent whole, and also includes the nature of the items that must be exchanged between systems at each level, and the services provided by the systems.

Enterprise

Information, conceptual data modeling

Computational architectures, software

Technology, platforms

Engineering, networks

Table 1: A five-layer schema based on levels of abstraction

Table 2: The final 8-layer schema adopted by the workshop (inspired by Buehler and McKee 1996; Voisard and Schweppe 1998).

For the two lowest layers of Table 2, it was felt that interoperability was already complete, because these services are interoperable by definition. But the remaining six layers all require some advance to achieve full interoperability. In the cases of tools, middleware, and data the need is clearly for technical advances; at higher levels the need switches to semantics, and finally to the resolution of institutional and social issues related to policy, values, and culture. Thus the highest levels are those at which interoperability will also be the most difficult to achieve.

3 Semantics

Formal systems allow meaning to be established and made consistent over widely-scattered communities. There is universal agreement, for example, about the definition, meaning, and numerical value of the symbol p . This conveys enormous advantages, and supports extremely impressive levels of compression. In order to send the 10¹⁰ known decimal digits of p it is necessary only to send a single character in an agreed alphabet. Assuming order 10³ such characters in a word processor, the effective compression achieved by this example of shared meaning is on the order of one to one billion. To take a geographical example, the shared meaning inherent in the standard system of latitude, longitude, and reference ellipsoid allows the location of any point on the Earth’s surface to be defined to an accuracy of about 30m by specifying a pair of degrees, minutes, and seconds.

There was agreement at the workshop that geographic information in general lacks a suitable formal system or theory capable of dealing with all of its many aspects or forms. Formal systems exist for defining point locations, but not for more complex objects, and not for surfaces or connected networks. Instead, there are numerous more-or-less formal systems implemented in the wide variety of GIS products available today. Thus the language of latitude and longitude is universal, but the general language for describing all phenomena distributed over the surface of the Earth is not, and has not yet been fully defined. Instead, meaning tends to be common only within certain disciplines or subdisciplines, within individual textbooks or classes, and within agencies or groups. Semantic interoperability addresses the need to extend these common meanings more generally, through the adoption of general specifications, standards, languages, vocabularies, and formalisms. The OGC specification (http://www.opengis.org) is clearly a major step in this direction.

Formal languages exist for many purposes, some of which are directly relevant to GIS. They have been defined for:

• creating models and representations (UML, EXPRESS)
• programming, or formalizing processing steps (C, C++, Java)
• accessing and retrieving data (SQL)
• simulating systems (SIMULA)
• creating visualizations
• creating metadata and other abstractions
• annotating and marking up documents.

Several attempts have been made to define general languages for GIS processing, though not for the entire domain of geographic information. These include the various versions of Tomlin’s map algebra (Tomlin 1990, 1991), the work of Takeyama and Couclelis (1997), the dynamic simulation language of van Deursen (1995), the computational modeling system of Smith et al. (1995), and various efforts to extend SQL to handle spatial data (e.g., Egenhofer 1994). Any of these could provide a basis for interoperability, by allowing users to interact with many systems using a common, consistent language, and several GIS products have adopted versions or dialects of Tomlin’s map algebra. However, all of these efforts fall somewhat short of the objectives of a comprehensive theory or language of all geographic information that could be adopted as the basis of a general interoperability.

Kuhn (1997) discusses a framework for examining shared meaning. One difficulty with comparing information-processing systems is that the same information can appear in many different forms, separated by a potentially automatic processing step. For example, a location can be expressed with equal accuracy in UTM coordinates or in latitude and longitude, because a standard mapping exists between the two systems. Thus a database containing a coordinate in latitude/longitude is essentially the same as one containing the same location in UTM, even though the contents may appear very different. Any two systems can be said to be identical with respect to some information product X if both are able to return the product by some well-defined process operating on data, even though both the data and the process may be different. The differences between the two systems are essentially moot to a user working at the conceptual level and requiring X.

A system based on a formal language is able to support computation expressed in symbolic form. It is also able to support translation between formats, provided the mappings between these formats and the formal system is well-defined. It can support internal verification and quality control, because the formal system forces certain logical rules to be followed. If it is known, for example, that a network of lines represent a partitioning of the plane into a set of non-overlapping areas that collectively exhaust the plane, then it follows that the network can have no nodes of valency 1 (the dangling segments and undershoots of the familiar task of building ‘topology’ from digitized ‘spaghetti’). Finally, a formal language can support a human interface that is consistent across systems, requiring no retraining provided the system designer allows the user to customize the interface to a familiar ‘look and feel’.

The workshop agreed that semantic problems will persist, and impede interoperability, long after the technical problems are solved. Semantic problems are clearly the ‘hard’ problems of interoperating GISs, especially since several trends seem to be working against the kind of universal sharing of meaning that semantic interoperability would require:

• Previous efforts to specify uniform standards in the GIS community have not met with great success, except where their adoption can be mandated. Even within a single public agency there are difficulties in forcing compliance with standards.
• The fragmentation of the GIS software industry, and increasing overlap with other forms of software such as CAD and DBMS, has made it more difficult to promote uniformity.
• After many years of concerted effort, the failure to arrive at a consensus on a unifying theory of geographic information is frustrating. Because so much has been invested in data and software already, it seems unlikely that a uniform theory could be successfully disseminated and accepted even if one could be found—if one were found, would the community acknowledge it? Older systems of knowledge dissemination are becoming confused by the comparative ease of access to information on the Internet, the high prices of books and journals, and the budget problems faced by traditional libraries.

4 A Vision for Interoperating GISs

The introduction laid out some of the characteristics of an interoperable GIS world, and argued that interoperability has many aspects. Many of these were identified in the previous section, in the discussion of layer models. This section presents the views of the workshop on the implications of interoperability, in the form of a vision for the future of GIS. The time horizon varied during the discussion, some participants feeling that a horizon of 2002 was most realistic, and others opting for a longer horizon of 2010. Although there was agreement on the nature of the vision, no attempt will be made here to present a consensus on the timetable.

GIS is becoming more and more ubiquitous, as a larger proportion of the population becomes aware of it, or trained in its use, as GIS becomes easier to use, and as the price for an entry-level GIS installation of hardware and software continues to fall. Thus the workshop participants were willing to consider the realistic possibility that in future a large proportion of society, if not society as a whole, will be GIS-enabled, if not geographically-enabled.

Ubiquitous GIS implies a pervasiveness of spatial thinking and awareness—that people commonly think of activities in terms of location, proximity, adjacency, and other basic spatial properties. Problems are solved in their spatial context, by considering the relationships between activities that take place in spatial proximity, and the impacts of one co-located activity on another. Pervasive spatial thinking implies a much richer spatial grammar, allowing people to express geographic relationships and patterns that are now difficult because of the weakness of the language in certain areas, such as the description of continuous spatial change. Ubiquitous GIS also implies that the tools for acquisition of spatial data are readily available at low cost, empowering everyone to be a creator and publisher. Simple, low-cost GPS receivers are already making this a possibility, along with the kinds of sensors now regularly deployed in support of precision agriculture, for example. A GPS receiver deployed in a mail delivery truck vastly reduces the cost of building a geocoded database of street addresses, an un-manned aircraft can provide low-altitude aerial imagery much more cheaply than traditional methods, and ‘soft’ photogrammetry has had enormous impact in lowering the cost of constructing DEMs.

The workshop agreed that in the future GIS will become:

• distributed, with processing, storage of data, and user interaction occurring at locations that are potentially widely scattered; a user at location A might send data from location B to a server at location C to be processed, and have the results returned to location A for display;
• disaggregated, as monolithic systems developed by single vendors are replaced by ‘plug and play’ components from different vendors that are designed to interoperate through conformity with industry-wide standards;
• decoupled, with the disaggregated components needed to complete a given complex of tasks being distributed over many networked systems;
• interoperable, a requirement that is clearly a precondition for all of the previous three.

The group began to define a wish list of objectives for this new world of interoperable GIS. some of its key elements are:

1. Workflow models for domain applications. If GIS software is to be decoupled, then there needs to be a much clearer concept than exists today of the specific needs of given applications. What, for example, are the standard workflows in an application like infrastructure maintenance? A general model will be needed so that vendors can provide its individual components as interchangeable parts.
2. One benefit of interoperability will be the existence of clear principles for the internal logical consistency of GIS databases. For example, if there is general agreement that city streets partition the plane into blocks, then tests can be applied to any database purporting to represent streets, to see if this condition holds. Indeed, this was the basis for some of the first procedures for checking internal logical consistency, in the case of the U.S. Bureau of the Census DIME (Dual Independent Map Encoding) database prepared for the 1970 census.
3. Software packages that are interoperable are also likely to be stable, since the same basic principles will apply to the design of each succeeding version. The general theory of geographic data that provides the basis for interoperability will also ensure stability of software through time; in contrast, the current lack of such a theory is one reason for persistent backward incompatibility.
4. Just as (1) above argued for standard workflow models, with standard processes being applied in specific applications, there will be a need for standard essential data models to underlie typical GIS applications. Interoperability implies a much higher level of agreement on basic data models than exists in GIS today. Models in an interoperable world will be self-describing, allowing them to be transferred readily between systems. They will reflect the decoupled world of ‘plug and play’ GIS software modules. Transformations between systems will be largely transparent to the user, who may even be aware of the format in which data are actually stored. These workflow models will be rich in semantics, reflecting an increasing trend towards universally-understood vocabularies. Finally, they will extend to dynamic data models, based on a comprehensive theory of geographic data types that includes time-dependence.
5. In the interoperable world of the future it will be possible to search for data using intelligent engines that are far more powerful than today’s search engines, which are largely dependent on recognition of text. Geographic data are not rich in text, and recognizable words may be entirely absent from a GIS database; if present, they may be in coding schemes other than ASCII, that are therefore essentially invisible to agents designed to scan and catalog text. Intelligent search engines will recognize geographic data sets, and be powerful enough to open and examine the contents of certain standard types. They will be able to recognize certain key concepts, such as level of detail and accuracy, that are essential to users searching for data to satisfy specific needs. They will be context-based, capable of modifying concepts to suit the context defined by the data. Finally, they will be intelligent enough to make assessments of the fitness of specific data sets for use in given user-defined applications.
6. Interoperability creates the potential for much more sophisticated strategies for management of distributed spatial data. Custodianship of data can be decoupled from issues of location, allowing owners of data to retain control despite being physically removed from the data’s actual location. Consistency across distributed databases will permit much more effective version and change management control.

5 Mechanisms

The layer model shown in Table 2 includes interactions that are purely technical, such as the exchange of data between systems. But at the higher levels, interactions are between more abstract entities, including communities, and the object exchanged is similarly abstract. The workshop participants recognized that interoperability can refer to a range of human activities, from exchanges between systems to the kinds of compatibilities that are needed for collaborative research. One can think of interoperability in the context of education, and address the problems that need to be overcome if teaching materials are to be exchanged across the barriers that exist between individuals, institutions, departments, disciplines, or countries. One can think of interoperability in research, and address the problems of interoperation between individuals, groups, and teams, through the sharing of knowledge, equipment, data, or methods. The problems of collaboration between disciplines, exemplified by the impediments to joint work involving atmospheric and ocean scientists on the transfers of energy and matter at the air–ocean interface, are similar in many ways to the problems of interoperating GISs, especially if the latter are examined at all of the levels of Table 2.

Science is going through something approaching a paradigm shift as it becomes clear that the complex problems of today require the joint effort of specialists in many disciplines. The old concepts of scientists as rugged individuals are being replaced by newer ideas of collaboration and sharing. This new kind of science faces numerous impediments, and traditional scientific culture has few ways of addressing them.

The participants raised an interesting question: if one agrees that collaborative research requires an interoperability in science, then does it follow that research on interoperating GISs must necessarily be itself collaborative, or interoperating? Can a research agenda be designed to address impediments to interoperating GISs that is at the same time useful in addressing the needs of interoperating science generally? Clearly the case is easier to make in the higher levels of Table 2 than in the lower?

With these thoughts in mind, some attention was devoted at the workshop to appropriate mechanisms that might move the research agenda of interoperating GISs forward. Discussions began with three premises:

1. That the groundwork for interoperability had been laid through academic research efforts over the past decade, many of them sponsored by NCGIA (e.g., NCGIA's Research Initiative 2 on Languages of Spatial Relations; see in particular Mark and Frank 1991).
2. That the Open GIS Consortium had made substantial progress in implementing the results of this research by constructing an appropriate framework and language for interoperating GISs, with broad support from the industrial community.
3. That many fundamental problems remain to be resolved; that they would require much greater attention from the academic community, with strong links to industry; and that a suitable infrastructure was needed.

There are very significant cultural differences between the GIS software industry and the academic GIS research community, despite the fact that regular exchanges occur, in the form of students hired, interactions at conferences, networks of acquaintances, etc. Academic research is rewarded largely at the individual level, as professors progress through the ranks of the professoriate, and through key stages such as tenure. Academics build personal reputations, publish as individuals, and are treated largely as individuals by the community. By contrast, software engineers in industry are encouraged and rewarded as members of teams, ideas belong to corporations, and success is measured by the corporate, not the individual bottom line.

This led to a strong belief that any mechanism for building the research infrastructure for interoperating GISs in the academic community would have to be very different from the mechanisms employed by OGC. Academics would not ‘play’ in a structure oriented to building consensus, where no rewards accrued to the individuals involved. A separate but parallel structure was needed, with strong links between it and OGC.

The discussion led to a conceptual design for an International Interoperability Institute (I³), a mechanism that would promote close, synergistic collaboration between academia and industry, to promote and facilitate research on GIS interoperability. The institute would provide a parallel structure to OGC, with strong linkages, but designed specifically to support academic research, following the cultural norms of the academic community. Figure 1 shows how this parallel structure might be implemented.

At the head of I³ is a Director, reporting to an Advisory Board, with a small staff. The Institute itself would comprise a set of individuals, distributed across a number of institutions, and committed to advancing research in interoperating GISs. The equivalent of OGC’s Working Groups would be a number of research teams, perhaps centered within academic departments or universities, and carrying out the research with various types of funding, from industry, foundations, or government sources.

The Institute’s programs would be designed to promote research and education, and to foster strong linkages with industry through the parallel structure of OGC. They would include a mix of traditional and novel approaches to academic infrastructure, such as:

• Collaborative publication. The participants discussed the Bourbaki model, a device used in mathematics to foster collective research through publication under a collective pseudonym. This device might work well in some countries, though in others, including the U.S., the loss of identity for the participants could be perceived as harmful to careers.
• Conventional publications. The Institute would promote and facilitate publication by conventional means, through books, conference proceedings, and journal articles.
• Training and outreach. The Institute would provide and organize internships and sabbaticals, to foster collaboration between academics and with industry.
• Academic input to the Technical Committees. The Institute and OGC would together foster linkages between the two structures, with particular emphasis on links at the level of the Technical Committees and the Institute membership.
• Industrial input to academic research. The Institute and OGC would work together to facilitate participation by industry in research centered in academe.
• Individual contributions. Unlike the Technical Committees, where the work is entirely directed at building consensus, the Institute would accommodate individual contributions, and build mechanisms to recognize their significance. Its culture would celebrate the creative work of individuals, and foster originality and creativity at the individual level.
• Research funding. The Institute would broker funding, particularly that originating in industry and destined to support academic research.

6 Education

Karen K. Kemp, NCGIA

As has been noted above, the advent of interoperating GISs has many implications for education. Many of the measures of the success of interoperation are specified as measurable changes in the content of GIS courses. This suggests that GIS education may become an unwitting accomplice in the move to interoperation. However, an alternate view may be that GIS education will become a fortunate beneficiary. The vision of interoperating GISs foresees ubiquitous GIS and the corresponding necessary pervasive spatial thinking and awareness. The same vision also acknowledges that success in interoperability means that there are many things which will no longer need to be learned. How will GIS education change with interoperability? There are two perspectives to consider in this context: 1) Interoperability and GIS education, and 2) Interoperability for GIS education.

If we consider some of the themes discussed at the workshop which characterize interoperability, the impact of this technology on GIS education becomes apparent. The vision implies that because interoperability exists, GIS becomes ubiquitous, embedded in many everyday activities across a wide spectrum of enterprises. Clearly, a large portion of today’s curricula will be superceded by new priorities brought about by the need to learn more about abstract concepts and less about technical details. Some of the changes in GIS education which we can anticipate include:

• Current tools will migrate to new disciplines as the concepts are incorporated into more enterprises. This is already happening, without interoperating GISs. The assumption is that the incorporation of geographic information and related functionality into a broad range of enterprises will continue and expand as the tools become more accessible, as less knowledge is needed to use them.
• Some technical foundations and old tools will no longer need to be taught. What goes from the curriculum will depend upon what is standardized (thus removing current contenders from further consideration) or formalized through new theories. For example, if fields can be formalized as a comprehensive concept, the various digital representations of fields and their associated procedures for determining values at any location no longer need to be a fundamental education component. Tools for converting geographic coordinates between the various systems will no longer need to be taught if there is a universal concept of location.
• New tools incorporating concepts expressed at a higher conceptual level will appear and new introductory materials will be needed. For example, consider what might need to be taught if semantic interpreters appear as a new tool. Students will need to learn how to itemize the characteristics of a geographic object so that the interpreters can determine precisely what should be represented and how it should behave. As we once learned about interpolation operators, we will now need to learn about the different modes of behavior of various geographic objects.
• More spatial theory, when and if it does materialize, will be taught as foundation material. As spatial theory begins to provide the basis for interoperability, it will provide the basis for the fundamentals of geographical analysis and spatial thinking.
• New technical specialties will arise to support the training of programmers and technicians for interoperating GISs. While many technical aspects will disappear from common knowledge, even amongst SAPs, many new technical aspects will arise as a result of the new technologies. Thus, where we once had GIS courses for the education of SAPs, we will now have as many specialized courses for the technicians and programmers who will be needed to support the new technologies. While the vision implies that SAPs themselves will become ubiquitous, technicians will continue to receive similarly specialized education.

In order to achieve this vision of interoperability, some conditions will need to be met. In particular, if we are to see spatial thinking and awareness become ubiquitous, a common spatial language or grammar will be needed. At some time in the distant past, the concept of latitude and longitude was not as universally understood as it is today, even within learned circles. Just as the concepts of latitude and longitude can today be clearly and comprehensively described, we must someday have widely accepted comprehensive definitions for field, network, area, etc. A formal spatial grammar will also allow some stability to be achieved so that what we teach to everyone about spatial concepts can have a lifetime longer than a single software version. This will make it much easier to embed spatial thinking across the curriculum. And, if we can have a spatial grammar, then can we also have a grammar checker which will ensure that our semantics are interoperable?

How are we to achieve this commonly accepted spatial grammar? It must have taken decades if not centuries for the current definition of latitude and longitude to be generally acknowledged; however, given the rate of change today, we don’t have that luxury of time. In the near term, can the OGC specifications act as models for this spatial grammar? Can we speed the adoption of this grammar by allowing OGC to certify educational materials as "OpenGIS compliant"? Unfortunately, the global nature of today’s economies add a further complexity to the development of a common grammar. A common grammar may be desirable, but many of us think of it as a subset of English. Can it be in English only, or must it sustain a multilingual and multicultural context?

• What is the social and economic value of the spatial revolution in education?
• What educational infrastructure and superstructure is required to support ubiquitous GIS?
• What needs to be taught about spatial thinking and analysis at different education levels in different disciplines?
• Can we develop a standard spatial grammar incrementally? Can we use OpenGIS specifications as a model for building this grammar?
• If a grammar is created, can we also create a spatial grammar checker?

The second perspective to consider is the impact of interoperable technology in general on the enterprise of education itself. There are many challenges facing today’s educators, including:

• Individual class sizes in higher education are growing with no trend to the contrary in sight. In lower education, class sizes have increased dramatically in recent years, though there is now a trend in some states, such as California, to reduce these numbers.
• An awareness of the need for integration of knowledge from many different disciplines is becoming generally accepted pedagogy. Funding agencies such as NSF support integrated research as well as education activities which support an integrated approach (such as the IGERT program—Integrated Graduate Education and Research Training)
• Computing skills among students have increased though spatial skills have not.
• The resource base of institutions is getting smaller, leading to growing demands on instructors’ time and productivity.
• Browser technology and the internet are making it much easier to distribute educational materials and databases, though the ease with which they can be integrated and incorporated into individual activities is limited.

Within GIS education, many of these challenges are amplified since it is a technologically dependent discipline in which the rate of change in the technology on which curricula are based is faster than the annual school year cycle. It is impossible for an individual GIS instructor to stay on the leading edge of the technology where students (and administrators) need them to be.

Based on the interoperability model, a vision of an interoperable resource base for teaching GIS (or any other subject) was envisioned at the workshop (see Figure 2). This resource base would contain a wide range of interoperable education objects, be globally distributed, and supported by one or more services such as that proposed by the Instructional Management System (IMS) project (see http://www.imsproject.org) or provided by WebCT (see http://homebrew.cs.ubc.ca/webct). Here, when preparing a course, lecture, or weekly module, an instructor chooses a number of different education objects from the distributed resource base. Each of these objects is dependently developed and maintained. This means that the materials provided by the software vendors can be based on the most current versions available, case studies from local agencies can include current projects, and the knowledge base can evolve gradually to reflect changes in current theory. In an interoperable world, all of these objects fit together seamlessly and can be incorporated into an individual instructor’s education module easily and quickly.

• What levels of interoperation are needed to make the resource base work?
• What is the proper level of interoperation between educators and resource providers?
• What services and products might an Education SIG in OGC produce?
• What new models of education institutions need to exist to support an interoperating education structure?
• Can an international resource base be developed? What are the cultural and linguistic requirements of such a global resource?
• Who pays for interoperability? How are the contributors rewarded?

Appendix 1 contains a call for participation in a workshop on interoperability for GIScience education; it was planned during and immediately after the specialist meeting, and represents an immediate follow-on activity.

Figure 2: An interoperable resource base for teaching

If research is needed to remove impediments that currently stand in the way of interoperating GISs, then how does one know that progress has been made, and how does one finally know that all impediments have been removed, and interoperation has been achieved? Workshop participants felt that some time should be spent discussing suitable metrics of progress, and of the inherent difficulty of achieving interoperability.

Some space was devoted in the first section of this report to the relationship between interoperability and the concepts of transparency and ease of use. That discussion suggests that progress towards interoperability might be measured by:

• Reductions in the complexity of the user interface, and the number of instructions needed to complete a given task.
• Reductions in the complexity of GIS curricula, or what it is necessary for a user to know to complete a given task.
• Increases in the population able to make use of a GIS, through reduced need for specialized training.

Each of these is a possible basis for measuring progress, but not for identifying the stage at which full interoperability has been achieved. That stage by definition is reached when training on one GIS is simultaneously training on all GISs; when no effort is required to transfer any data set from any system to any other; when one single language is sufficient to define all spatial problems; and when there are no difficulties of communication between GIS users embedded in different disciplines or cultures. Full interoperability need not mean a single language, since the same effect can be achieved if perfect translators allow two people to communicate despite differences of language between them.

The group discussed various possible case studies, and the characteristics that made one case study more difficult than another. The following six examples possess a wide range of characteristics, and it would be interesting to identify the important issues impeding interoperability in each case:

1. Land use/land cover. The problems of mapping of land use or land cover illustrate many of the problems of interoperability, in the specific context of data semantics. Any scheme for classifying land must be to some degree subjective, since it is impossible to write down a rigorous set of rules that are sufficiently objective to satisfy the criteria of reproducibility—two people set the same task will almost certainly produce different results, however much effort is made to ensure consistency between them. Several projects, including an extensive one centered on Wicomico County, Maryland, involving several agencies whose responsibilities include the mapping of wetland, have provided persuasive demonstration of the difficulties.
2. Roads. The VITAL project at the University of California, Santa Barbara (http://www.ncgia.ucsb.edu/vital), has demonstrated the difficulties of achieving interoperability between street centerline databases. Although several vendors provide such databases for all roads and streets in the U.S., there are significant differences between them in terms of road positions, names of streets, existence of streets, and other properties.
3. Images. Interoperability issues abound in the image domain. Impediments include problems of registration, differences between sensors, the effects of cloud, growth stage, solar illumination, and season, and many more.
4. Projection and datum. The technical issues associated with geodetic datum and projection are among the most obvious ways of distinguishing SAPs from the general public. Yet in principle there seems little reason why these issues could not be made fully transparent; why the average user of GIS should be any more concerned about such issues than about the more technical issues of computer design.
5. Cross-cultural issues. The group speculated about the problems faced by an American driving in a foreign country, such as the United Kingdom, and using an in-vehicle navigation system. Impediments to interoperability in such cases center on the degree of understanding of the user, and the ability of the system to support significant differences in language.
6. Urban infrastructure. Another potential case study might take two or more utility operations, such as a telephone and a water utility, and examine the impediments to interoperation, due for example to basic differences in software.

The group identified five basic criteria for progress and success in interoperating GISs:

• Ease of use. Metrics associated with the difficulties faced by users in achieving given tasks.
• Cost. Metrics of the costs associated with overcoming difficulties due to lack of interoperability.
• Speed of response. Elapsed time can also provide the basis for measuring the problems associated with lack of interoperability.
• Robustness. Interoperable systems are more difficult to misuse, so metrics of robustness might also be suitable.
• Impediments. One might measure progress by enumerating the impediments to interoperation, and recording success in overcoming them.
• New opportunities. Finally, progress could be measured through the new opportunities that it creates. This is more consistent with contemporary notions of performance measurement through outcomes. The most compelling case for true progress would be made, for example, by being able to point to new scientific insights that resulted from interoperability, or new collaborative policies and plans, or new organizational forms.

This final section reviews the research agenda devised by the workshop participants. It includes a selection of ideas for research, but is not intended to be exhaustive. The first part discusses general issues, and this is followed by a list of specific suggestions.

In the short and medium term, there is great potential for overcoming impediments to interoperability at the technical level. The group anticipates that decoupling and disaggregation of GIS software will proceed rapidly, as the industry develops a new approach to software engineering based on component-ware. In the next few years it is anticipated that most spatial functions will become available in this form, as modular software components operating within a standardized and open ‘plug and play’ environment. The software industry will implement ORB-style protocols, and data providers will move to significant encapsulation of processes with their products.

In the medium to long term, efforts will focus on semantics, and the achievement of uniform vocabularies and interoperability of meaning. At this stage it is difficult to see how this will occur, given the skepticism expressed earlier about the community’s ability to achieve consensus. But competitive pressures in the industry clearly point in this direction.

The following items were suggested by participants as suitable topics for research, that could be executed within a reasonable timetable, and had the potential to advance knowledge and understanding in the area of interoperating GISs. The order in which they are presented is not significant:

• How must GIS education change in a world of interoperable GIS?
• How can we formalize a spatial grammar for educational purposes?
• How can OGC and interoperability in general be used as a model for collaborative development of education materials?
• Who will teach GIS in an interoperable world?
• How can GIS curricula be kept in sync with OpenGIS specifications and related innovations?
• Study the impact of inconsistences in interoperating GISs on the design of new user interfaces
• How do information communities work together?
• Specify user requirements for interoperability
• Develop a taxonomy of spatial concepts/data models
• Identify domains for research—e.g., disasters, cadaster
• Use OMT to express a domain-specific problem
• Formalize a spatial concept in any modeling language, and compare formalisms
• Develop a language for e.g. SimCity 2010
• What information is needed for a geoservice to work with another geoservice?
• Case studies of how two agencies work together—e.g., wetlands
• Case study of the horizontal datum
• What are the interoperability questions in risk assessment?
• Develop, prototype, and test measures of the (geometric, attribute) difference between two data sets covering the same area
• Test whether component-ware solves the interoperability problem
• What is the appropriate granularity for tasks and component-ware?
• How to build workflows from components?
• Compare product models: data supply vs service supply
• Representation of spatial data quality in an interoperable environment
• Interoperability with respect to scale and error
• What new business opportunities are created by interoperability?

Acknowledgment

The Varnenius project is supported by the National Science Foundation under Cooperative Agreement SBR 96-00465.

Amsterdam, The Netherlands

18-20 May 1998

The National Center for Geographic Information and Analysis (NCGIA), The Free University of Amsterdam, and UNIGIS International are seeking interested GIS educators and others to participate in a 3-day workshop entitled "Interoperability for GIScience Education". The purpose of this workshop is to draft a discussion document outlining the need for action and a possible research and development agenda with action plans. Results of the workshop will be presented for review and discussion to the international GIS education community and others at various meetings in 1998, possibly leading to the initiation of an international cooperative project in 1999. The structure of the workshop will be a combination of plenary discussions and small group sessions on specific topics.

The motivation for this meeting comes from a recognition that GIS educators in the private and public sectors are faced with both an opportunity and a dilemma. As the GIS vendors move to open systems which can be integrated with many traditional operations, the use of spatial data and analysis will become widespread throughout business, government and education. Hence the need for GIScience education is expanding rapidly. However, at the same time, rapid changes are occurring in both GIS technology and the structure of higher education. These shifting foundations make it impossible for individual GIS educators to stay on the leading technological edge where their students need them to be. Collaboration in education is now essential. The aim of this workshop is to explore how the GI community might work together to develop an interoperable or open environment in which educators can exchange resources and add value to these resources for use in their own unique educational settings while at the same time retaining intellectual (and commercial) copyright. Several related activities are now underway and the GIS education community should play a role in these. Background information about this workshop is available at http://www.ncgia.ucsb.edu/ige98. Updates on the meeting planning will be posted there.

Participation in the workshop is limited to 15–20 people and is by invitation only. The organizers are hoping to bring together a diverse group of participants, particularly people involved in a) the creation of educational materials for GIS Higher Education, b) the distribution of educational materials for higher education, or c) the development of open on-line distribution systems appropriate for these types of digital materials.

Proposals to participate in the workshop are invited. Applications for participation should consist of three parts:

1. a brief indication of why you want to participate in the meeting, why you are interested, and what you would contribute (max. 500 words);
2. a position statement or research abstract, describing a particular element of or perspective on the topic (max 1000 words);
3. a brief curriculum vitae with up to five selected publications most relevant to the topic.

Applications for participation must be submitted by email to kemp@ncgia.ucsb.edu. The documents requested above may be delivered via email as included ASCII text, local URLs, or attached word processing format documents. Delivery by ftp is also possible; please send email to request details. Proposals must be received by 20 March 1998 to ensure consideration. All submissions will be reviewed by the Workshop co-Leaders and by the Steering Committee.

Accommodation for the event is still being finalized, costs will be posted shortly. A limited amount of funding may be available to cover local expenses for academic participants. Most participants will be required to fund their own travel to Amsterdam. Assistance in finding travel support may be available for qualified participants.

May Yuan
University of Oklahoma
100E. Boyd Street
Sarkey Energy Center 684
Norman OK 73019
USA
myuan@ou.edu
405-325-4293
405-325-6090