ODISEES: A New Paradigm in Data Access

As part of its ongoing efforts to improve access to data, the Atmospheric Science Data Center
has developed a high-precision Earth Science domain ontology (the “ES Ontology”)
implemented in a graph database (“the Semantic Metadata Repository”) that is used to store
detailed, semantically-enhanced, parameter-level metadata for ASDC data products. The ES
Ontology provides the semantic infrastructure needed to drive the ASDC’s Ontology-Driven
Interactive Search Environment for Earth Science (“ODISEES”), a data discovery and access
tool, and will support additional data services such as analytics and visualization.
The ES ontology is designed on the premise that naming conventions alone are not adequate to
provide the information needed by prospective data consumers to assess the suitability of a given
dataset for their research requirements; nor are current metadata conventions adequate to support
seamless machine-to-machine interactions between file servers and end-user applications. Data
consumers need information not only about what two data elements have in common, but also
about how they are different. End-user applications need consistent, detailed metadata to support
real-time data interoperability.

The ES ontology is a highly precise, bottom-up, queriable model of the Earth Science domain
that focuses on critical details about the measurable phenomena, instrument techniques, data
processing methods, and data file structures. Earth Science parameters are described in detail in
the ES Ontology and mapped to the corresponding variables that occur in ASDC datasets.
Variables are in turn mapped to well-annotated representations of the datasets that they occur in,
the instrument(s) used to create them, the instrument platforms, the processing methods, etc.,
creating a linked-data structure that allows both human and machine users to access a wealth of
information critical to understanding and manipulating the data. The mappings are recorded in
the Semantic Metadata Repository as RDF-triples.

An off-the-shelf Ontology Development Environment and a custom Metadata Conversion Tool
comprise a human-machine/machine-machine hybrid tool that partially automates the creation of
metadata as RDF-triples by interfacing with existing metadata repositories and providing a user
interface that solicits input from a human user, when needed. RDF-triples are pushed to the
Ontology Development Environment, where a reasoning engine executes a series of inference
rules whose antecedent conditions can be satisfied by the initial set of RDF-triples, thereby
generating the additional detailed metadata that is missing in existing repositories.

A SPARQL Endpoint, a web-based query service and a Graphical User Interface allow
prospective data consumers – even those with no familiarity with NASA data products - to
search the metadata repository to find and order data products that meet their exact
specifications. A web-based API will provide an interface for machine-to-machine transactions.