Professional Documents
Culture Documents
Geology Icws 2007
Geology Icws 2007
Zaki Malik, Abdelmounaam Rezgui, A. Krishna Sinha*, Kai Lin**, and Athman Bouguettaya Department of Computer Science, Virginia Tech, VA. USA *Department of Geosciences, Virginia Tech, VA. USA ** San Diego Supercomputer Center, UCSD, CA. USA {zaki, rezgui, *pitlab, athman}@vt.edu, **[email protected] Abstract
We present DIA, a Web services-based infrastructure for the Discovery, Integration, and Analysis of geoscience data, tools, and services. DIA provides a collaborative environment where scientists can share their resources (e.g., geochemical data, filtering services, etc.) by registering them through well-defined ontologies. We have developed a planetary materials ontology in OWL for this purpose. The ontology is used by different geoscientists (using Web services) to explore, extract, and integrate information from different heterogeneous data sets. The DIA system is now in its final pre-release phase. It is currently made accessible to a few geoscientists for conducting usability analyses, and it will eventually be made available to the community at large through the geoscience portal (GEON) at the San Diego Supercomputer Center. few have also been implemented. Up till now, most tool implementations have been private and similar to the data sharing problem, tool sharing has also been limited. This has resulted in redundant efforts for tool deployment. Geoscientists have been defining and using stand-alone tools, in segregated environments. The tools that are defined in one system are seldom used in the other. Even if the tools are shared in a public-setting, the interoperability of the different systems has been problematic. Therefore, geoscientists have mostly been re-inventing the wheel in terms of tool implementations. This has been a major impediment in taking geoinformatics forward. In many cases, geoscientists may neither be able to deploy, nor implement the necessary tools, due to lack of resources, knowledge, etc. Thus, many geoscientific queries remain unanswered. We believe that the Web services initiative undertaken by the World Wide Web Consortium (W3C) can resolve the problem of tool sharing across different systems. Web services can provide a platform that facilitates cooperation among researchers and allows integration of tools. The participants of a Web services system do not have to worry about the operating system, development language environment or the component model used to create or access the services. The tools developed by geoscientists can simply be wrapped as Web services and made accessible to the scientific community at large. In Web services, only the input and the output parameters are defined and there is no need to know the exact details of the involved processes. This encourages application reusability and reduces development time, effort, and cost. In this paper, we describe our approach towards organizing geoscientific data through well-formed ontologies, and the need for defining different tools as Web services. Specifically, we describe the DIA engine: a system for the Discovery, Integration, and Analysis of geoscience data. The DIA engine will be made available to the geoscience community through the GEON portal. GEON is a federally funded, multi-pronged project. As a collaboration among several institutions, projects, and institutions, GEONs main objective is to provide
1. Introduction
Several research initiatives by geoscientists over the years have produced large amounts of data. However, the ability to find, access, and properly interpret these large data repositories has been very limited. Two main reasons for this lack of data sharing are the absence of publicly advertising research data, and the adoption of personal acronyms, notations, conventions, units, etc. by each research group, when producing data. This makes it difficult for other scientists to correctly understand the semantics of the produced data, and makes the automatic interpretation and integration of data simply infeasible. This leads to a poor ability in answering complex geoscientific questions. We believe that to enable the sharing and integration of geosciences data on a global scale, ontology-based data registration and discovery is required. Hence, we have defined a planetary materials ontology for efficient, reliable, and accurate data sharing among geoscientists. The ontologically registered data can be used through different geoscientific tools to answer complex user queries. Thousands of such tools have been made available in the geoscientific literature, and a
cyberinfrastructure support for integrative geoscience research. The paper is organized as follows. In Section 2, we define geoinformatics and show how geoscientists can benefit from this approach. In Section 3, we provide the details about the ontologies that we have defined. In Section 4, we introduce the DIA engine and show how Web services are used in the system. Section 5 provides some concluding remarks and direction for future work.
we have built a prototype system (DIA) as proof of concept, that shares and uses a variety of tools to answer complex geoscience questions. In the following, we provide details of the tools and technologies used in constructing the DIA engine.
2. Geoinformatics
Man has been studying the earth and the solar system for centuries. Over the years, the study of several phenomena as eclipses, tides, volcanism or earthquakes, etc. has proven to be challenging. The inability of scientists to create frameworks that provide integrative solutions has been the prime impediment in this regard. Nowadays, the need for multidisciplinary observations and their integration is well-understood. However, we still conduct our science in similar ways to our forefathers. For instance, we make our observations both on the ground as well as through remote sensing techniques and store our information in computers, but still find it difficult to achieve an integrative understanding of our objectives. It poses a fundamental problem in seeking a more robust knowledge about the earth and the solar system. The prominent reasons for this are: (1) vast volume of data which individuals cannot assimilate, (2) inability to discover heterogeneous data, (3) variability in format of data even within a single discipline, (4) focus on narrowly defined science questions that cannot be easily expanded to a higher and integrative level of understanding of how and why the earth works as it does, and (5) changing the societal needs that influence support for research and education. To make novel geoscience discoveries, we will require the rapid availability of data and information that encompass a variety of temporal and spatial scales, as well as the capability to integrate heterogeneous data sets and tools to analyze them. Geoinformatics can be thought as the field in which geoscientists and computer scientists are working together to provide the means to address a variety of complex scientific questions using advanced information technologies. Thus, as a concept, geoinformatics is rooted in providing rapid access to both academic and agency hosted databases, with tools and services, so that the best integrative approaches in (earth) science research and education can take place. As an initial step towards realizing the goal of geoinformatics, we have defined the planetary materials ontology that covers a variety of concepts, minerals, elements, isotopes, etc. for geoscience data. Moreover,
defining ontologies is to provide an organizational structure for classifying data that can be discovered automatically by computers. These high-level ontologies and data level ontologies contained within, allow geologists to discover databases as well as datasets within databases using geoscience related concepts instead of simple keyword-based search. This is made possible as ontologies allow the registration of databases at different levels of granularity. For example, the hierarchical concept for rocks: [Rock -> Igneous Rock -> Geochemistry of Igneous Rock -> element abundance] provides the conceptual framework for registering and indexing a database at increasing levels of detail. This forms the basis for developing the ontologies that represent more explicit statements about the data. As mentioned earlier, the use of these ontologies, especially at the most detailed level, facilitates semantic integration of heterogeneous geologic datasets. We favor ontology based integration as it systematically resolves both syntactic and semantic heterogeneity, thus allowing integration of multiple distributed databases into a single virtual database. Unlike integration based on merging multiple schemas, ontologic registration relates the data to concepts, rather than the structure within a database [12]. Figure 1 shows the high-level representation of the Planetary Ontology. For example, the package Planetary Material can be used to represent the nature (physical, chemical) of substances and their properties. This figure also shows the utilization of imported and inherited properties from additional packages, e.g., Physical Properties, Location, and Planetary Structure, to more fully define the concept of Planetary Materials. Existing ontologies available from SWEET ontology library [4] such as Numeric, Time, and Units are also used as they provide common concepts which are useful for the development of data level ontologies.
Figure 1. The planetary ontology OWL DL, the description logic variant of OWL [5]. Apart from browsing options, the ontology repository at GEON, provides an option to combine, parameterize, and switch between different ontologies. This is a useful practice as mappings can be defined that translate between classes and properties in one classification and corresponding classes and properties in another classification [6, 7]. In describing the GEON ontology mappings below, we assume that all ontologies are formalized in the same logic, i.e., we consider mappings between different ontologies (not between different logics). An ontology mapping from an ontology to is defined as having a class mapping f and a property mapping g. Class mapping is a partial function from the class set of to the set of all derived classes in . A derived class is defined from other classes, e.g. an intersection of two classes. The mapping should preserve the subclass (is a) relation, i.e., 1 and 2 are classes in and 1 is a subclass of 2, then f(1) must be a subclass of f(2) in :
If f does not preserve the subclasses relation, then a query about individuals in f(1) may incorrectly return some individuals in f(2) [12]. The property mapping g is a partial mapping from the property set of to the set of all derived properties in , and should satisfy the following condition: If p is a property between the classes 1 and 2 in , then g(p) is a property between the classes f(1) and f(2):
Note that a class mapping and a property mapping induce a natural translation from the constraints of to the constraints of . For instance, if has a constraint
|| < 30, i.e., the cardinality of is less than 30, then this constraint can be translated into an constraint |f()| < 30. An ontology mapping should satisfy the following condition: if a constraint ca in can be naturally translated into a constraint cb in by its class mapping and property mapping, then cb is implied by . In other words, this condition says that an ontology mapping should not introduce any new constraints into the target ontology. In fact, the requirement that a class mapping should preserve the subclass relation is a special case of this condition [12]. If there is an ontology mapping from to , then from any model M of , a sub-model can be extracted and naturally transformed into a sub-model of based on this ontology mapping. An ontology mapping is more general than a simple equivalence relation on classes and properties. In fact, it defines a structural translation from one ontology to another. For example, let contain a class Person and a data property hasName and contain a class Employee and two data properties name and id, then a mapping sending Person to Employee and hasName to name is a valid general ontology mapping from to , so any model of can be naturally transformed into a model of via this mapping, that is, an employee is a person, but a person may be not an employee. In general, combining several ontologies can not be done by a simple union operation, because these ontologies may contain similar but different definitions for the same concept. For example, suppose ontology has a property p defined between the classes 1 and 2 (i.e., 1 2) and has a property q between 1 and 2, and assume that 1 and 1 are conceptually the same. Combining and should give an ontology O containing a property p between a new concept (of O, representing 1 and 1) and 2, and a property q between and 2. In OWL, this can be done by using equivalentClass or sameAs tags. Theoretically this result is equivalent to the pushout [8] of the following:
p
ontology mapping from the formal ontology to OB must be provided. The instantiation result is the pushout of the diagram above, i.e., the ontology O in the diagram. It is unique up to ontological isomorphisms. The system can infer implied facts in the selected ontologies. Ontology mappings are used in navigation and query processing. If a data set is registered to an ontology OA, and there is an ontology mapping from OB to OA, then users can choose both, ontology OA and ontology OB to query the data set.
where is an ontology containing only one class . The ontology mapping 1 sends to 1, and 2 sends to 1. Note that parameterized ontologies and their instantiation can be implemented based on the same principle. A parameterized ontology is a pair (, OA) where the parameter ontology is included in the body ontology OA, which means that there is an inclusion ontology mapping from to OA. To instantiate this parameterized ontology with an actual ontology OB, an
definitely produced as long as the data required to answer the query is present, i.e., empty result sets are only returned in case of missing data. The user clicks through the different menus to build an exact query. Using our running example of distribution of A-Type plutons, Figure 2 shows what menus are involved in getting to the A-type filtering (Four top-level menus are navigated to get to the A-Type menu). Similarly, the Tool Selection layer in Figure 3 shows how menus are navigated and how Web services are used.
Figure 3. A layered bottom-up depiction of DIA phases from data to data products and extensive query processing, but enable far richer possibilities for scientific discovery. There are two main classes of integration: Intra-class integration is a process whose input is two or more homogeneous data sets, i.e., registered to the same ontology. This process uses the common ontology to interpret all data sets and generate an integrated data product. Inter-class integration is a process whose input is two or more heterogeneous data sets (i.e., registered to different ontologies.) This process uses the appropriate ontology to interpret each data set. It uses an integration class to generate a data product out of two or more data sets. Figure 3 shows the integration between geospatial, geotemporal, and gravity griding data products that are obtained after data filtering through a bounding box (used to limit the area of interest). evaluation, i.e., to analyze the results. In Figure 3, using a space-time analysis and studying the generated 3-D model, a geoscientist can verify the hypothesis on AType bodies. Our running example query (A-type rocks) has major implications on tectonic models such as (1) a failed rift associated with a triple junction [13], (2) gravitational collapse of crystal regions over thickened by Grenville orogenesis [14], (3) flanks of an active within plate rift zone similar to the Red Sea region [15], and (4) as a continental plume track. Thus, facilitating rapid data /tool access clearly is a requirement as geoscientists engage in more complex queries.
4.5. Analysis
Geoscientists can use the data products generated as a results of the above mentioned phases in hypothesis
Figure 4. The DIA engines architecture and workflow graphical interface (Figure 4). The engine translates these queries into a sequence of tasks that include: accessing map servers, discovering and accessing data sources, invoking Web services, filtering features, joining layers, and the graphic rendering of query results for visualization. The DIA engine also enables users to save their query history as well as export data products for future references. Since the DIA engine is developed along a service-oriented approach, key code modules are wrapped as Web services. This approach has two advantages. First, it makes the system readily extensible. As the geoscience community introduces new services, these could be integrated in the DIA engine as new functionalities. Second, services developed for the DIA may be used as building blocks to produce other systems. Figure 4 depicts the service-oriented architecture of the DIA engine and also describes how this architecture supports the discovery, integration, and analysis of Earth science data. The DIA engine supports several querying modes (Geological map-based queries, Region-based queries, etc.) To answer the running example A-type query, the user first selects the option Geological map-based queries in the DIAs main menu. The system then accesses a geological map server, gets a (default) geological map, and displays it to the user (Figure 4). This map enables the user to select the area of interest (i.e., Virginia). This may be done by selecting a bounding box or by selecting the entire state. In the latter case, the DIA engine accesses a gazetteer to determine the selected states latitude-longitude coordinate. The user then uses DIAs drop down menus to identify a computational filter (A-type filter in this case) to be applied to the data samples located in the selected area. The DIA system invokes a GEON Web service called GEONResources that provides functions for searching and getting the metadata information for GEON resources including all the registered databases that have samples located in the area of interest. In the case of the A-type query, the DIA invokes the service GEONResources with the keyword GeoChemistry as well as Virginias latitude-longitude coordinates. GEONResources then returns a set of database identifiers that correspond to (registered) data sets containing geochemical data for samples located in Virginia. For each returned database, the DIA system executes a two-step process. First, it builds a virtual query (expressed in SOQL, a language developed by GEONs researchers at SDSC) that requests all the data (i.e., columns) that are necessary to apply the filter specified by the user. The DIA system then invokes a GEON Web service called SoqlToSql that translates this SOQL query into a SQL query. In the second step, DIA submits the SQL query to the GEON server that interacts with the actual database server, gets a record set containing the relevant data samples, and returns the data to the DIA engine. The DIA engine then inputs the received data to the A-type filter and displays the query results to the user. Note that the A-type filter tool is also a Web service that may be hosted other geoscientists. This facilitates reusability, as new filters do not need to be written by each individual group. Similarly, gravity information etc. is gathered and a data product is generated. The A-type bodies data product and gravity data product are then integrated to deliver the desired query result in a visual manner. The DIA Engine is now in its final pre-release phase. Currently, the DIA engine is accessible through http://mapserver.geos.vt.edu/ATypeDemo at Virginia Tech. It will eventually be made available as a portlet in the GEON portal.
5. Conclusion
We have presented our approach for the semantic integration of data and tools for geoinformatics. The resulting system termed as the DIA engine, uses ontologies and Web services to organize, annotate, and define datasets and geoscience tools. We have defined several geoscience ontologies and have wrapped several geoscience tools as Web services. These tools and ontologies are available for public use through the GEON portal at the San Diego Supercomputer Center. The DIA engine is in its final pre-release phase, and will soon be made accessible to the geoscience community at large through GEON. We expect that as the Semantic Web matures, and more geoscientists adopt this paradigm, a number of geoscience tools will be made accessible as Web services. This would require that similar to data management, Web services are also ontologically registered. Annotating Web services with semantics would ensure that appropriate tools (in form of Web services) are selected in an efficient and automatic manner for answering geoscience queries. In the future, we intend to define a service ontology for geosciences.
[6] T. J. Bench-Capon, and G. Malcolm, Formalising ontologies and their relations, Database and Expert Systems Applications, 250259. 1999. [7] Y. Kalfoglou, and M. Schorlemmer, Information flow based ontology mapping, 1st International Conference on Ontologies,Databases and Applications of Semantics, Springer, Irvine, CA, 2002. [8] M. Barr, and C. Wells, Category Theory for Computing Science, Prentice-Hall, 1990. [9] U. Nambiar, B. Ludaescher, K. Lin, and C. Baru, The GEON portal: accelerating knowledge discovery in the geosciences, The Eighth ACM international Workshop on Web information and Data Management, WIDM 06, VA, USA, November 10, 2006. [10] A. K. Sinha, Z. Malik, A. Rezgui and A. Dalton, Developing the Ontologic Framework and Tools for the Discovery and Integration of Earth Science Data, Cyberinfrastructure Research at Virginia Tech, Annual Report, June 2006. [11] U. Nambiar, B. Ludaescher, G. Memon, and D. Seber, GEONSearch: From Searching to Recommending, GeoInformatics, May 10-12, 2006. [12] K. Lin, and B. Ludscher, A system for semantic integration of geologic maps via ontologies, Semantic web technologies for searching and retrieving scientific data (SCISW), 2006. [13] D. W. Rankin, Appalachian Salients and Recesses: Late Precambrian Continental Breakup and the Opening of the Iapetus Ocean, Journal of Geophysical Research, 81(32), pg. 5605-5619, 1976. [14] A. K. Sinha, The Latest Precambrian Magmatic Record of Extensional Tectonics in the Central and Southern Appalachians: Response to Grenville Compressions, Geological Society of America Abstracts with Programs, Southeastern Section, v.24, n. 2, p. 65, 1992. [15] R. P. Tollo, and J. N. Aleinikoff, Petrology and UPb Geochronology of the Robertson River Igneous Suite, Blue Ridge Province, Virginia - Evidence for Multistage Magmatism Associated with an Early Episode of Laurentian Rifting, American Journal of Science, v. 296, p. 1045-1090, 1996.
6. Acknowledgement
This work is supported by the US National Science Foundation award EAR 0225588 to A. K. Sinha and National Science Foundation grant number 0627469 to A. Bouguettaya.
7. References
[1] A. Sheth, Changing focus on interoperability in information systems: From system, syntax, structure to semantics, M. Goodchild, M. Egenhofer, R. Fegeas, and C. Kottman, eds., Interoperating Geographic Information Systems, Kluwer, 530, 1998. [2] A. Bouguettaya, Z. Malik, A. Rezgui, and L. Korff, A scalable middleware for Web databases, Journal of Database Management, 17(4), pg. 20-47, 2006. [3] Y. Ding, D. Fensel, B. Omelayenko, and M. Klein, The Semantic Web: Yet Another Hip?, Data and Knowledge Engineering, 41(3), 205-227, May 2002. [4] NASA SWEET, Semantic Web for E&E Terminology, http://sweet.jpl.nasa.gov/ontology/, February 2006. [5] W3C-Consortium, Owl web ontology language reference, www.w3.org/TR/owl-ref/, 2006.