AN OBJECT MODEL FOR GEOLOGIC MAP INFORMATION

Please download to get full document.

View again

of 14
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Journals

Published:

Views: 0 | Pages: 14

Extension: PDF | Download: 0

Share
Related documents
Description
AN OBJECT MODEL FOR GEOLOGIC MAP INFORMATION Boyan Brodaric 1 and Jordan Hastings 2 1 GeoVISTA Center, Pennsylvania State University and Geological Survey of Canada, 234B-615 Booth St., Ottawa, ON K1A
Transcript
AN OBJECT MODEL FOR GEOLOGIC MAP INFORMATION Boyan Brodaric 1 and Jordan Hastings 2 1 GeoVISTA Center, Pennsylvania State University and Geological Survey of Canada, 234B-615 Booth St., Ottawa, ON K1A 0E9, 2 Department of Geography, University of California Santa Barbara, Ellison Hall 3611, Santa Barbara, CA 93106, Abstract National geologic map databases are presently being constructed in the U.S. and Canada, as well as in several other countries. Here, we describe an object-based model for geologic map information, specifically designed to represent digital geologic maps and related geoscientific information. Although oriented to geoscience, several fundamental issues in representing geospatial information are explored in this design, including the philosophic and cognitive basis of mapping in general, and the overall framework in which map-related information can be represented. Thus we take an ontologic approach to geospatial representation, supplemented by an epistemic view of the scientific process, which culminates in a very general model a meta-model for map information. Practical as well as theoretical considerations motivate this approach. Primarily, however, we describe the theoretical foundations of our meta-model, specifically semiotics, category theory, and ontology in geospatial information. Finally, we report briefly on a specific prototype data model derived from the meta-model and implemented in a commercial object-oriented GIS. Keywords: geologic map information, data models, ontology, classification. 1. Introduction The public agencies that provide geologic maps and related information in the U.S. and Canada are just now beginning to actively develop on-line systems for their work. These systems simultaneously address two basic needs: improving the efficiency of routine information handling within an agency, and promoting the non-traditional usage of geologic information within and outside the agency. 1 Though conflicts in information policy do exist among agencies, the need to cater to on-line science is cautiously overriding such concerns as several multi-agency systems are emerging. In North America, for example, three major systems exemplify this approach: (1) the Canadian Geoscience Knowledge Network (CGKN; a cooperative initiative to link the public geoscience data providers in Canada; (2) the U.S. National Geologic Map Database project (NGMDB; a Congressionally-mandated repository for map information from many state agencies as well as the USGS; and (3) GeoInformatics, a proposed network of U.S. academic geoscience databases (GEON; In this paper we discuss a project undertaken for NGMDB in particular 1 : an object-based geologic map information meta-model derived from the North American Digital Geologic Map Data Model standards effort (NADM; Johnson et al., 1999; The design of our meta-model rests on two ambitious objectives regarding the underlying geospatial 2 map information: 1. To supplement feature-based geospatial information with ontologic context, explicitly represented; and further 2. To supplement the ontologic aspects of geospatial information with epistemic considerations, drawn from geoscience directly. The first objective expresses our view that scientific meta-modeling, i.e. abstracting the many types of models within geoscience, is critical for the effective representation of geoscientific information (Bennet, 1997; Gahegan & Brodaric, 2002; Raper & Livingstone, 1995; Langley, 2000). Specifically, in our design we designate and encode four fundamental types of models, symbolic, conceptual, occurrence, and descriptive, that underpin the structure of geoscientific information and its visual presentation. The second objective reinforces the point that our meta-modeling emphasis proceeds from epistemology, the fundamental knowledge acquisition and evaluation task (Clancey, 1993). In particular, geologists evolution of environmental concepts from their interpretation of observed occurrences is often incremental, complex, uncertain and dynamic, and frequently results in multiple valid models for a geographic region (Brodaric & Gahegan, 2001). This behavior contrasts with many non-scientific domains in which concept development is a minor component, and where indeed, concepts are often known a priori rather than dynamically obtained and revised, and where feature occurrences are distinct. In general, ontologically-driven geospatial information frameworks are well suited to relatively fixed domains (e.g. land cadastre, urban utility networks), where they serve to partition geospatial concepts into perspectives and to instantiate relatively certain and unambiguously described occurrences within them (e.g. Benslimane, 2000; Fonseca et al., 2000). However, geospatial information frameworks for dynamic and exploratory work, in science 1 Variants of this architecture are also being implemented within CGKN 2 We use geospatial to subsume the terms geographic, geologic, and geoscientific uniformly 2 particularly, also need to be epistemologically-driven to account for the creation and evolution of concepts and occurrences, and to engage the uncertainties and ambiguities consequently introduced. In this paper we begin tackling the central representational issues inherent in geoscience by exploring the complex of relationships between concepts and occurrences, and by modeling them. This requires a general structure for modeling concepts and their relation to phenomena, which is a level of abstraction higher than traditional knowledge engineering approaches that directly model domain concepts rather than the structure organizing them. Thus we arrive at a general object-based structure, a meta-model for geoscientific information that embeds very broad ontologic and epistemological factors and can be refined into domainspecific data models. This design complements related work on geological maps that considers relational data model design (Baines & Giles, 1997; Johnson at al., 1999), web-based and multi-media applications (Brodaric et al., 1999; Heyn, et al., 2000), user interfaces (Kubler & Voisard, 1999), geometric representations (Balovnev, et al., 1997), reasoning (Voisard, 1998; 1999), and cognition (Brodaric & Gahegan, 2001). It is novel in terms of (1) integrating cartographic presentation with concept development and information ontology into (2) an object-based meta-model for geologic map information. The paper is organized as follows: Section 2 describes the meta-model developed; Section 3 describes a prototype implementation; Section 4 sketches future work; and Section 5 concludes with a short summary. 2. Theory and Design Geologic maps depict geologic occurrences within a specific geographic region and spanning particular periods of geologic time. Typical maps are 2D, representing the intersection of 3D occurrences with a surface, usually the Earth s. These maps often represent a causal-process history that explains the evolution in time of the occurrences and their interactions. The history is typically located in accompanying texts and diagrams, and is supported by the occurrences themselves. Not all occurrences, or their parts, can be directly observed or measured, as some are historical and others inaccessible mainly due to sampling constraints, causing them to be inferred. Even observables can be variously described and identified according to the specific expertise of different investigators. Reasoning is thus affected by the amount and type of evidence, by the known theory, by the physical situation, and by the mapper s cognitive orientation. Geologic map representations are thus meta-models of geologic spacetime-process models, including observed and inferred occurrences, their causal history, contexts and other factors impacting reasoning (Voisard, 1998; 1999). In this section we describe our meta-model and indicate how individual geologic map reference models can be derived from it. To attain generality we ground the meta-model in semiotics, category theory, and geospatial information ontology. 3 2.1 Semiotics Semiotics, the study of signs (Noth, 1990), provides a useful initial framework for representing a scientific view of a geologic map. In cartographic semiotics the meaning of a map symbol derives from the relationship held between the symbol, the concept being symbolized per some interpreting agent, and the occurrence being referred to (MacEachren, 1995). We implement these semiotic primitives as foundational objects in our meta-model (Figure 1): time instant interval cycle process discrete continuous space network coverage Symbol theme geology geologic time geologic process geologic entity rock unit X formation rock type granodiorite monzogranite Concept 1..1 Occurrence 0..1 SpaceDesc Fig. 1. The semiotic triangle is shown using UML notation (Rumbaugh et al., 1999); some relations for concept, symbol and occurrence primitives are not depicted. Insets show example instances of spatial descriptions (geometry; right), symbols (top) and concepts (left); indents and arrows denote concept specialization. 1. Concept: refers to the abstract objects that constitute geoscientific vocabulary; these can instantiate specific occurrences (below) and relations, and serve as values for attribute domains: e.g. X formation, granodiorite, fault, intrudes, etc. Concepts may tier from very general and perhaps universal, such as Aristotle s substance (Sowa, 2000, 57), through generic, such as space, time, process, etc., to particular (Guarino, 1998), such as those for a specific discipline, theory or even a geographic region. 2. Occurrence: refers to the tangible objects observable in the field and/or on maps (geospatial features), and their relations in space, time, and otherwise. Note that an occurrence is an instance of exactly one concept and optionally possesses one spatial description, its geometry. Thus a geospatial feature possessing constant identity might be represented by several distinct occurrences distinguished by conceptual, spatial, temporal or other variation this notion of geospatial feature is implicit and not explicitly modeled. The meta-model does allow multiple concepts to be linked to a spatial object. 4 3. Symbol: refers to the (carto)graphic objects in the visual display, presented as areas, line and point markers, charts, tables, and so on. Independent treatment of symbols enables cartographic behavior related to scale dependencies, symbol overlap, etc., to be associated with both concepts and occurrences. This permits the establishment of cartographic norms via concept symbolizations, and allows for cartographic exceptions in specific occurrences; e.g. though the default designation for all occurrences of the X formation concept is blue, a particular pluton occurrence might be showcased in red. 2.2 Category Theory The human interpretive process involves categorization where concepts and occurrences are obtained in parallel and are mutually affective. The meaning of concepts is thought to become increasingly contextual in lower conceptual tiers, affected by epistemological factors such as methodologies, actions, physical situations, and other factors related to how humans categorize, construct/recognize concepts (Smith & Mark, 2001; Frank, 2001). This insight would seem to hold for geologic mapping, suggesting more complex relations between concepts and occurrences than typically modeled in databases (Baker, 1999; Brodaric & Gahegan, 2001; Dehn et al., 2001). Specifically, in the mapping process, occurrences interact with concepts in two ways: as instances of established concepts, such as various map polygons labeled according to a geoscientific concept, e.g. X formation ; and also as evidence leading to the development of these concepts, such as the field observations employed in determining X formation (Figure 2). In information science these interactions are typically referred to as instantiation and classification, respectively; however, we use the term categorization, borrowed from cognitive science, to denote concept development from evidence. Thus, instances are unique identifications, whereas categorizations are prototypical statements based on evidence of varying typicality in which concept meaning is closely related to either (1) an average summary of evidence, or (2) an ideal description (Lynch et al., 2000) such as a geologic type locality. Occurrences are therefore distinguished by their role as instance or evidence in relation to concepts. Identifying and modeling these roles (Figure 2) then permits them to be described so that, for example, human or machine categorization methods can attribute evidence, increasing the context modeled in the database. A consequence of varying typicality of evidence is that evidence descriptions, consisting both of structure, such as attributes and behavior signatures, and of content, such as attribute values and behavior functions, will be variously similar to concept descriptions. In particular, concept descriptions will vary dramatically with descriptions for outlying evidence, but concepts will share descriptions with ideal evidence. Descriptions may also be shared between occurrences, as understandings converge within and between individual mappers in a specific mapped area (Brodaric & Gahegan, 2001). The meta-model therefore possesses a free-standing data store for descriptions (Figure 2), enabling them to be shared 5 among and between concepts and occurrences. For example, in Figure 2 the concept X formation shares a granodiorite rock type description with ideal evidence at a site; however, a single polygonal instance of the X formation concept is described by monzogranite instead. Thus, descriptions are the central information repository in the meta-model, containing the bulk of traditional data and information. Descriptions may be subclassed and also inter-related among themselves as needed (not depicted in Figure 2). The remainder of the meta-model can be seen as a knowledge superstructure adding conceptual, cartographic and occurrence-driven context to the descriptions. Reference models can be generated by loading specific concepts and description types into the meta-model. theme rock unit X formation rock type granodiorite monzogranite Concept Symbol Instance 1..1 Occurrence Evidence 0..1 SpaceDesc Descript ion ThemeDesc ProcessDesc TimeDesc etc. etc. Lithology granodiorite porphyritic foliated monzogranite recrystallized gneissic Fig. 2. Insets show example instances of concepts (left) and occurrences (right) with related spatial (right) and thematic (bottom) descriptions. The dashed lines illustrate concepts evolving from evidence. See text for details. 6 2.3 Information Ontology The knowledge structure described above emphasizes the relationships among and between symbols, concepts, occurrences, and descriptions. Specific knowledge representations can be achieved by grouping these primitives into arrangements, called models. We designate four types of models, conceptual, occurrence, cartographic and descriptive, and denote legends and maps as combinations of these basic types. In information science, conceptual models are formally known as ontologies (Guarino, 1998); in the practice of geoscience these are commonly expressed as vocabularies, taxonomies, or classification schemes, such as those for geologic time (e.g. Precambrian ), rock units (e.g. X formation ), rock types (e.g. granodiorite ), or as more complex models such as those for petrography, stratigraphy, genesis, among others. (Heyn, et al., 2000). By extension, we consider occurrence models as part of epistemologies, formalizing how we know/evaluate geologic realities (Raper, 1999). Epistemologies and ontologies are fundamentally linked: an ontology provides a set of concepts and logic for how occurrences might be arranged in space and time, such as typically described in a map legend and in accompanying notes, whereas the epistemology provides a specific arrangement of geologic occurrences and their causal explanations emplaced in space and time, thus demonstrating on the map the validity of the concepts and related logic. A map in which the conceptual and occurrence models are inter-consistent is deemed to work, expressing this connection. Ontologies (conceptual models) are represented in the meta-model as logically consistent collections of concepts, their relations, descriptions, and evidence (Figure 3). Similarly, epistemologies (occurrence models) bind together logically consistent occurrences, their relations, descriptions and instances. Conceptual models enable concepts to be bound in different arrangements, thereby providing various conceptual perspectives on, say, a single occurrence model (Fonseca, et al, 2000). Likewise, multiple occurrence models can derive from a single conceptual model, recognizing that, for example, a common taxonomy may lead to different maps for the same area. The meta-model also accommodates cartographic models for symbols and descriptive models for descriptions. For example, a palette is a model of symbols from a specific symbol library or agency-approved cartographic standard. Similarly, a geospatial model formalizes the spatial descriptions and relations (geometry and topology) that denotes a dataset/layer in a GIS. A map legend may then be thought of as a symbolized conceptual model devoid of occurrences. Applying a legend to a valid set of occurrences enables their instantiation, categorization and visualization; furthermore, applying an alternate legend to the same set of occurrences effectively generates a reconceptualized (derivative) visualization for the area. In conclusion, we define a geologic map view as a complex model consisting of a legend (symbolized perspective) manifesting a conceptual model (ontology) that in turn exemplifies an occurrence model (epistemology) for the region of interest. 7 This map view draws objects from one or more geospatial models (GIS layers), and/or utilizes descriptions from some aspatial models (attribute databases). models MapView conceptual cartographic occurrence descriptive ConceptModel Legend Symbol Palette OccurrenceModel S patia lmode l DescriptionModel Concept Instance 1..1 Occurrence Evidence 0..1 SpaceDesc Descript ion ThemeDesc ProcessDesc TimeDesc etc. Lithology Fig. 3. The definition of a map in the meta-model. See text for details. Note, for reasons of clarity and space some details are omitted 3. Implementation Testing of the meta-model is proceeding on several fronts: within NGMDB (Hastings & Brodaric, 2001), in a web-based geologic map database project (Davenport et al., 2001) and in a prototype digital library for sustainable development (Journeay, et al., 2000). Testing within NGMDB has thus far 8 concentrated on evaluating (1) the suitability of the meta-model to geological data and (2) the ease of execution of some common operations, putting aside for the moment issues of efficiency and scalability. The NGMDB test data consisted of four 1:24K geologic maps provided by the Kentucky Geological Survey (KGS), who partnered with the USGS and GE Smallworld (GESW) for the evaluation. As a preparatory step, the meta-model was configured to suit the maps: concepts, description types, and their relations were specified for the geologic formations, faulted structures, coal beds, etc., found in the region. Once configured, the meta-model was implemented virtually as is inside GESW s case tool, requiring minor custom programming to implement specialization relations such as those between the description object and its subclasses (Figure 3). Custom methods were also added to some metamodel objects to achieve the desired mapping functionality (see below). Testing involved loading data and performing some typical visualization and analysis operations. 3.1 Loading Data The Kentucky data were supplied in ESRI shapefile and MS Excel spreadsheet format
Recommended
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks