In earthquake science and geoscience, scientists typically have their own interpretations and analyses of the raw data from different sources (observations, simulations, etc.). Individual databases are distributed and different data schemas are designed. Therefore, the semantic representation of the data and those interpretations are essential to enable to access to such federated databases and the integration of those data. The goal of this project is to support interoperability for heterogeneous geoscience data and experiments/simulations using that data. In order to break the barriers and useful to access the tremendous amount of heterogeneous geoscience data, a semantic metadata management system and wrappers for web services are required to support effective information retrieval and web-based search for data of interest to specific scientists. In this work, we design and build a semantic based system to provide interoperability for heterogeneous data, different applications and databases systems, and user-defined packages ? for the earthquake science domain. We also employ domain ontologies and ontology-interconnection capabilities to support interoperability. Such ontologies are also used to support information discovery and recommendation for users (scientists, students, etc.).
The underlying technologies are

? Semantic Metadata Management System accommodates data-mining, learning and updating capabilities for metadata management.
? Web Services are used for light-weighted protocols and universal data formats.
? Topic mining is being developed to perform event-based detection and tracking.
? The integration and representation of multi-modal data is accommodated by agent-based combination/merge routines.
? Data assimilation and 3-dimensional graphic based simulation technologies are employed to convey information to and from scientists and users

Figure 1. System Architecture

Semantic Metadata Management System

The semantic metadata management system we propose has five main functional capabilities:

1. Facilitate domain ontology creation and update,
2. Associate the ontology metadata with observational and hypothetical data,
3. Learn new concepts, relationships, and patterns among the metadata and data,
4. Support user (scientist) data and meta-data discovery/search and,
5. Provide the base for the semantic wrapping of information sources.

Rule-based reasoning agents and a simple metadata authoring/editing tool will accommodate data-mining, learning and updating capabilities for metadata management. We intend to ensure that the structure and format of metadata are compatible with RDF [1], DAML+OIL, and XML with limited ¡°process¡± [2, 3]. This will make the ontology/metadata portable. All communication of data will utilize XML that is compatible with and to an extent based upon Geography Markup Language (GML) [5]. GML helps to describe the format and transmission of geographic information and ensures that both spatial and non-spatial data can be integrated.

Web Services
Web services use an XML-based protocol and schema of interface definitions to invoke the applications among servers and clients. The protocol provides the information required by the remote services. One of the common web services, SOAP, is the method message procedure and deployed as an application in a web server. Web services cannot be completed without the method interface in the Web Services Description Language (WSDL) [4]. The interface to services is implemented with a web-friendly programming language (such as Java or Python). Web services allow information exchange among different platforms and applications and make remote application invocation possible [6].

Topic Mining

Topic mining is to find (new) concepts and events in a collection or stream of data [7]. Topic mining is able to perform thematic and/or the pattern-oriented trend detection and tracking. Unlike traditional keyword-based search, topic mining provides information upon an event-based point of view and helps to adjust the various interpretations of data for geo-science. ¡°Event¡± means a certain thing that happens at a certain point of time. For example, if a user searches for ¡°earthquake in Southern California¡±, a typical web search engine would provide the links of general descriptions or the research center information of earthquake in Southern California. However, an event-based search would return facts, i.e. ¡°earthquake on San Andreas fault in May 2002¡±. Furthermore, as shown in Figure 2, topic mining can hierarchically cluster collections of web pages describing earthquake occurrences. It also associates topical terms (from specific to abstract) with each cluster according to a topic.

Figure 2. An Example of Topic Mining

For additional information, please visit ¡°http://quakesim.jpl.nasa.gov¡±.