We implemented a Mediator-based Federated Information System: a tightly-coupled information system based on a unified model at the conceptual level. This system provides integrated access to EOS datasets managed by disparate (i.e. heterogeneous and distributed) resources. Such solution was successfully experimented in the framework of a project funded by the Italian Space Agency. It set up a federation of resources, located in the central and southern regions of Italy. The experimentation concerned two main kinds of dataset archives -which are the most important ones for such federation purpose: Sensor imagery archives, and Case Study archives. Such application is described.

In order to guarantee data source autonomy, the developed FIS is a read-only system. Federation-shared data may not be freely available but, e.g. provided under commercial license. Therefore, according to EOS data providers, a sophisticated AAA (Authentication, Authorization, Accounting/Auditing) architecture was implemented, which allows the Single-Sign-On (SSO).

The developed federation Client Application presents a combination of browsing and querying services. It is completely driven by the federation unified view and integrates browsing and querying according to the tree-structured common domain model.

All developed components are written in Sun Java 2 language which allowed their easy deployment in multi-platform environments (i.e. Windows NT, Linux, and Unix). For component integration technology we used the Web Services enabling technology (i.e. SOAP/XML and HTTP protocols). We utilized the Java WebStart technology as reliable code delivery technology. Database access is performed with JDBC/ODBC technology.

-Rationale: The EOS community needs information integration, because of the large heterogeneity of both data types and methodologies used by scientists. When data already exists in a variety of highly heterogeneous and autonomous managing systems, simply integrating this data cannot be sufficient. Effectively, what is required is a federated distributed system where the individual participants (i.e. EOS data resources) are self-contained autonomous systems, but together form a consistent wider picture-the federation.

Our goal was to enable users to issue a single query to search multiple EOS information sources and, in return, receive a combined result incorporating data from across these sources.

-Architecture: Federation customer are masked from the data resources logical and semantic heterogeneity by means of a federation common schema.

Data source wrappers hide technical and data model heterogeneities. Our approach requires well-customised wrappers, since they deal with data model heterogeneity and implement the data resources autonomy. The system carries out meaningful semantic integrations at both the presentation and data levels. Two kinds of mediations are utilised: pre-canned query mediation -implemented by the mediator components- and dynamic service mediation -implemented by the facilitator components.

-Common Model: The introduced common model is at the conceptual level rather than at the schema level. In fact, the system mediates across complex sources whose data comes from possible disjoint EOS information communities (e.g. oceanography and atmospheric communities). The common conceptual model introduces a common ontology, according to which data source must be modeled. The common model is a profile of existing standard specifications for geo-data and metadata (i.e. ISO 19100 series, and OpenGIS specs).