Membership: Buttenfield (leader) Larsen, Reitsma, Kersky, Tsou, Rokoske, Rock, Smith, Kole
Mission Statement of Team The goal of the Colorado component of the Evaluation Team is to evaluate the effectiveness of ADL from the user perspective, and to establish needs and requirements of three classes of target users. Knowledge gained from these activities informs system design and implementation. A variety of evaluation methods are applied to evaluate system functionality and user perspectives. Part of the Team's mission is to compare results of these methods to assess their relative effectiveness.
Research Activities and Progress over the Past Year
Continued support by the Colorado Administration for ADL have been
particularly focused in the Library Administration and the College
of Arts and Sciences. The Dean of Libraries provides 10% of
Suzanne Larson's time to ADL activities; the Dean was co-Convener
of last year's Alexandria Design Review Workshop. The College has
(through the Geography Department) continued support of a physical
lab space, and made available a small amount of computer system
administrator's time to set up networking services. The ADL lab
is presently converting to Windows NT from UNIX.
Overviews of the Alexandria Digital Library architecture, interface design, and user evaluation were presented by Buttenfield (1997d) and in Buttenfield and Larson (1997). Overviews of as yet unsolved algorithmic and infrastructure issues for delivering geographic data via the Internet were presented in Buttenfield (1997b; 1997c). Results of a workshop on the same topic are reported on in a National Research Council Report (1997); Michael Goodchild and Barbara Buttenfield serve as members of the NRC Mapping Science Committee which hosted the workshop.
In other research, the diversification of the User Interface and Evaluation Teams (with Teams at both Santa Barbara and Colorado) has allowed both teams to work not only on mechanics of user evaluation, but also to pursue theoretical research associated with use of digital library technology. The first type of effort emphasizes creating an operational library testbed. The second emphasizes utilizing that testbed as a vehicle for questions and issues in fundamental science. The discussion below demonstrates that both types of activities are present at Colorado.
Our findings show that the pattern of using the Web interface has not changed significantly over the year in spite of many changes made to the Web interface (changing background wallpaper color to distinguish the catalog, gazetteer and browser, adding exit surveys, adding cognitive affect buttons, help buttons, etc.). This is not to say the interface refinements were unnecessary, but rather that the changes made to the ADL interface did not adequately address users' demands for a simpler interface. We have been passing along the insights gained from transaction logging to the UCSB teams, and many of our suggestions have been incorporated into a recently released JAVA interface. Comparing spatial interaction models for the new JAVA interface with the original Web interface can tell us if user pathways through the Library have changed, and what new interface components have had an impact on those changes. We feel that the research has progressed to the point where we can begin to publish results in the coming year.
The landscape has several interesting properties, including line-of-sight Intervisibility. Connecting any two points on the surface one can draw a straight line which connects a whole series of related items, much like cross-referencing a card catalog search in a physical library. Zooming in to the landscape modifies the details one can extract from the catalog: at higher zoom levels, there is more detail available about the metadata record. One can imagine a direct manipulation interface tool, where lasso-ing a portion of the landscape essentially refines the query, without the user necessarily knowing the keywords involved in the SQL refinement. It is important to discover (by empirical testing) whether ADL users understand these properties of the information landscape (Fabrikant, 1998; see also Dissertations Supervised). The Colorado Team will run subject testing on sample landscapes in the coming year.
Automation of the georeferencing will proceed by searching a gazetteer (GIPSY) and a landcover database (GIRAS) and comparing catalog descriptions with locations having a high probability of correspondence. We have implemented GIPSY, and have negotiated with INFORMIX to receive their relational database software, which will enable us to automatically georeference the photographs. Comparison of the output from GIPSY with the true locations (from GPS, topographic maps, etc.) will help to advise catalogers when their MARC descriptions need additional locative information. The intention is to implement a working solution at the Denver Public Library sometime in the coming two year. Preliminary reports on this project can be found in Smith and Buttenfield (1997); see also Master's Theses - ongoing.
Abstracts of Published or Submitted Papers
Both uncertainty and errors are inherent in spatial databases. The processes of observing, measuring, interpreting, classifying and analyzing data among other operations give rise to systematic and random errors. Some errors may be quite large (blunders) and easily detectable. Other errors and uncertainties in spatial data are more subtle and are not easily detected or evaluated. Casual users of GIS may not be aware of their presence or even the possibility of their existence. These are the most problematic and the ones we must try hardest to illuminate. Graphical methods when used in conjunction with error analysis provide a means for identifying both gross and subtle errors and evaluating the uncertainty in geographic data. The implications for spatial analysis and spatial decision making of error and uncertainty in geographic data can be identified in theoretical work (for example in spatial statistics), in domain-specific applications (for example in environmental resource management) and in empirical testing (for example in recent cartography and cognitive science research). This chapter outlines a rationale for the use of graphical methods, highlights several historic and recent examples, develops a framework for graphical methods, and points to research challenges for the future and the potential for new techniques arising from technical innovations.
This paper reconstructs a thread of research discussion dating back to the mid-seventies on the nature of cartographic representations that must vary in geometry, topology, and semantics with variations in the resolution of source data. This thread underlies much of the work in map generalization, some of the work in multiple representations and multi-scale databases, and forms a preliminary statement to address the progressive transmission of vector data across Internet channels. What relates these three areas of research is the cartographic necessity to link visual logic (representations) with geographic semantics (communication).
The obvious mechanism for efficient digital data delivery is the Internet. Environmental information that is available is either difficult to locate, or if locatable, difficult to establish as current, precise, and credible. What impedes the international flows of geographic information that social scientists, earth and space scientists would use if they could just get reliable access? One impediment is that policies on information access and use vary widely between nations. A second impediment is that privatization of digital data delivery limits access to those who can afford to pay. Technical impediments include the volume of data files to be delivered which continue to exceed transmission capabilities, the inability to transmit vector information efficiently. Institutional impediments may in the end prove the most difficult to overcome, as the nature of protecting primary data sources and inter-disciplinary tensions in funded science inevitably obstruct willingness to share information.
The need to coordinate the delivery of geospatial data has been recognized by the business community and by the federal government. Dramatic changes in the way geospatial data are produced and exchanged are underway. The impact of these changes can already be felt in daily life. This in turn creates new opportunities for research on technical, organizational, legal, economic and societal aspects of data delivery and access. This article reviews important emerging research themes and on one particular project (the Alexandria Digital Library) implementing a mechanism for delivering geospatial data on the Internet. Some background on implementing the spatial data infrastructure will put digital library technology into context.
The paper will present ADL components briefly, focusing on research problems that have been resolved and those that continue to challenge the development of a map library whose archives are distributed across the Internet. Specific cartographic issues to be discussed include interface design and evaluation for a target audience that is difficult to profile, limitations of current Web browsers for performing otherwise simple graphical tasks (e.g., dragging a rectangle across an index map, providing special formats such as animation and sound), metadata collection and reporting, and scaling digital map archives beyond the terrabyte level while still permitting reasonable search and query activity. A final section of the paper will consider implications of delivering digital cartographic data that reflect on mandates for national mapping services around the world.
A digital library should be more than a physical library in electronic form. In a digital library, traditional distinctions between books, digital spatial coordinates, maps and satellite imagery should become transparent to library patrons. It should be possible to retrieve maps and images and overlay them with digital attributes from another data source. The digital library catalog should include digital files that are archived in depositories distributed across the nation. It should be possible to browse spatial metadata prior to downloading files. Patrons should be able to visit the library without ever leaving their own offices. This paper is an overview of the Alexandria Digital Library project (ADL), providing comprehensive library services of a map and imagery library over the Internet. This paper describes the origins of ADL, and of merging maps and images into the library information mainstream. We will describe the development of the ADL prototypes, and focus on the features of the current implementation that distinguish ADL from other efforts. We present research issues raised by ADL and their likely impact on the accessibility of spatial data to earth system scientists.
We are overwhelmed by the vast amounts of data accumulating daily. The extraction of information from online data sources is becoming more and more difficult. For example, if a query to a large archive returns hundreds of ÒhitsÓ, the most effective presentation is probably not a list of items, but some other type of graphical display. The concept of spatialization offers a promising potential to overcome the current impediments of retrieving items from large volume archives. Spatialization involves effective combination of powerful scientific visualization techniques with spatial metaphors that represent data that are not necessarily spatial in nature. Familiar spatial concepts such as distance and direction, scale, arrangement etc. which are part of the human experience in everyday life, are applied to create lower-dimensional digital representations of complex digital data. Skupin and Buttenfield (1996;1997) have demonstrated how spatial metaphors can be constructed for abstract information spaces. However, as these authors (1996: 616) point out, there has not yet been any subject testing to determine the appropriateness of such methods for visualization. We are not certain how people comprehend spatialized views, or whether the components of distance, direction and so forth are understood by viewers. This paper presents an experimental design to explore how spatialized views are understood by users. Subject testing procedures on graphical displays are outlined, which include the collection of performance measures on information retrieval tasks. The experimental application will rely on data collected from the Alexandria Digital Library Project.
This paper reports final results of dissertation research to determine effective and efficient map depictions of metatdata. Empirical testing demonstrates that attribute accuracy is not perceived by map viewers to increase visual complexity of a map display. In a task requiring simple and difficult spatial decision-making, subjects reported correct answers proportionately more often and with a shorter response time for some but not all tested cartographic symbologies. The shortest response times were discovered when attribute accuracy was displayed using saturation, for the simple decision. Alternative symbologies included value and texture. Experimental results for the difficult decision were less clear, although subjects were uniformly confident of their decisions. This work indicates that cartographic guidelines for displaying metadata can in fact be embedded within a map display, and the application of such metadata into map displays could improve the efficiency of data delivery from a digital map library.
This paper summarizes the discussions and key findings of a three day workshop held in Gavle, Sweden immediately prior to the 18th ICA/ACI International Cartographic Conference in Stockholm. The workshop focus on multi-scale representations of digital cartographic data comprised demonstrations, presentations, and parallel sessions. Topics of relevance to digital libraries included sessions multi-scale data integration, distributed computing as applied to GIS modeling, and delivery of vector data on the Internet. Thirty nine delegates came from seventeen countries comprising 20% from National Mapping Agencies, 10% from the private sector and 70% from research institutions (geography, cartography, surveying, computer science, physics, planning and linguistics).
The challenge to automate cartographic design and generalization stems from the fact that much of cartographic expertise is understood only intuitively. Intuitive knowledge is ill-structured, and therefore difficult to formalize. Non-formal principles are difficult to exchange verbally or procedurally. Incorporation of sound cartographic principles into existing GIS software will ensure that the cartographic products derived from GIS processing are valid. In addition to the improvement in quality, formalization of cartographic knowledge may reduce the need for manual intervention in some parts of the cartographic process, thus reducing costs. Delivery of generalized data from distributed Internet archives (as for example in digital libraries) can free users to focus on their systematic research instead of timely pre-processing to reduce raw data into a usable form.
Government agencies and other organizations are committing resources concerning the collection, management, and use of geospatial data. Although these actions are influenced by current policies, priorities and opportunities, their ultimate success depends on future developments and trends. An examination of possible strategies and policies for managing and building the National Spatial data Infrastructure (NSDI) was the focus of a two day workshop convened by the National Research Councils Mapping Science Committee and the Federal Geographic Data Committee, in April 1996. The framing question was ``In the year 2010, how will societal needs and public policies affect the requirements for spatial information and services and their integration at the individual, community, national and global levels?''
Information spaces come in many different shapes. Each is characterized by the information it contains and by a certain organizing structure. It is often reported that users browsing very large information spaces become quickly disoriented, and many visual navigation tools have been proposed in the literature. One class of proposals is to spatialize the information space, that is, to construct a spatial metaphor allowing users to browse very large archives of possibly non-spatial content. We present a rigorous definition of spatialization and a case study completed with real non-geographic data. We present the mathematics underlying a spatialized coordinate system, and demonstrate several visual display methods that preserve relationships within the archive. Our research is based on the assumption that cartographic principles of scale and generalization can have a major clarifying impact on this emerging method of data organization.
The Denver Public Library (DPL) has begun an intensive photodigitization project to convert its collection of historical photographs into digital format, and to distribute them via an intranet to its rural library counterparts in Colorado. Public access to the collection via the Internet is a planned development and is the motivation for collaboration with the Alexandria Digital Library (ADL). The ADL provides a framework for putting collections such as those owned by the DPL on-line via the Internet, providing search and access to collections to broad classes of users, and allowing both collections and users to be distributed throughout the Internet. A pilot project reports georeferencing a small set of DPL photographs by various methods (GPS, verification with existing maps, and probabilistic georeferencing methods) and compare the labor and accuracy of results.
Several important problems relating scale and cognition of geographic information remain unsolved. The first mentioned in this paper is the issue of scale dependence of digital cartographic data. Because the resolution of a map representation will dictate what geographic processes are made apparent, delivery of digital data at some generic all-purpose resolution is not possible. Moreover, the preservation of geometry, topology and attribution can vary with scale. Thus incremental transmission of cartographic data in vector form cannot be solved by a single wavelet transformation. A second, related problem is that we have not yet defined the smallest set of critical scales to minimize the number of versions of vector data that must be transmitted to give a scale-comprehensive vector database. Finally, fusing multi-scale vector data is a problem that has not been addressed. Other issues mentioned in the paper relate to aspects of empirical testing that could shed light on user comprehension and cognition of scale characteristics in cartographic data.
Data products have advanced to a fine level of resolution that inhibits downloading. Without data, geographic information processing for any application cannot proceed. This paper addresses how to de-centralize the geospatial data processing in such a way as to minimize the need to download large volumes of data. This paper presents examples of encapsulating data and generalization operators in object-based and agent-based structures which are sent to a server (via Intra- or Internet channels) to process large datasets remotely. Instead of bringing the data to the processor, we send object-processors to very large data archives to perform data reduction and other data modeling tasks, and return subsets (called ``data kernels'' of the original archives encapsulated with the parameters used in the generalization tasks. The object structure additionally captures a chronology of the processing applied to combine and process data from multiple sources, essentially providing a snapshot of data lineage which can be transferred with the processed data.
Decision-making in natural resources and environmental sciences is increasingly based on data modeling and data exploration. With development of GIS - related technologies, scientists are able to display large volumes of data rapidly. Human interpretive capabilities remain to a large extent rooted in the ability to visualize accurately. Therefore, information on data uncertainty is important for effective use of GIS data. It impacts the credibility of data representation and the confidence that is attached to data interpretations. We know that uncertainty varies within data layers, and that some types of natural resource data are more uncertain than other types. We can predict some GIS operations that increase or decrease levels of uncertainty. Much remains to be learned about how to model uncertainty, and how to embed uncertainty information into maps. This paper reports on a number of recently developed graphical techniques for including uncertainty information on environmental maps. A few techniques are derived from cartographic theory. Others result from empirical work, including subject testing. Results of recent experiments will be presented using a case study on siting a natural resources preserve in a wetlands environment.
As the digital data flow rises exponentially, the need for techniques and methods to efficiently extract information from large on-line archives becomes crucial. This paper demonstrates how spatial metaphors might be embodied in a working interface, to visually explore the holdings of the Alexandria Digital Library. The interface design is based upon three spatial primitives, namely distance (similarity), scale (hierarchy of catalog terms) and regionalization (concentration within some catalog categories). The use of spatialized views to represent catalog holdings of a digital library has not yet been implemented. A proof-of-concept demonstration requires subject testing to determine whether cognitive associations exist between the spatial primitives and interface displays.
This paper establishes an architecture for global user interfaces by defining appropriate user tasks and designing an intelligent agent-based interface mechanism for distributed geographic information services. User tasks include query, display, data integration, and GIS processing functions. The use of agents is described and justified in the context of increasing geospatial data volume, increasing complexity of GIS modeling, and the diversity of GIS software resources available on local and distributed networks. The architecture is defined and placed within a context of a new generation of distributed component technology. GIS users need a global user interface, which is platform-independent, modularized, reusable, self-described, and able to access multiple, heterogeneous, and distributed geographic information services on the Internet.