Metadata is a key aspect of data management. This paper describes the work of NFFA project on the design of a metadata standard for nanoscience community. The methodology and the resulting high-level metadata model are presented. The paper explains and illustrates the principles of metadata design for data-intensive research. This is value to data management practitioners in all branches of research and technology that imply a socalled “visitor science” model where multiple researchers apply for a share of a certain resource on large facilities (instruments).
The Nanostructures Foundries and Fine Analysis (NFFAEurope) project www.nffa.eu brings together European nanoscience research laboratories that aim to provide researchers with seamless access to equipment and computation. This will support a single entry point for research proposals supported by the project, and a common platform to support the access and integration of the resulting experimental data. Both physical and computational experiments are in scope, with a vision that they complement each other and can be mixed in the same identifiable piece of research.
The project requires setting up the IT infrastructure for managing research proposals and substantial amounts of data resulted from physical and computational experiments. A common metadata model that supports different stages of the nanoscience research lifecycle is essential to unified researchers’ experience across locations, and also for the design and operation of IT infrastructure components.
Metadata design is a part of a joint research activity within NFFA that takes empirical input from the project participants, also takes into account state-of-the art standards and practices. Metadata design is an incremental effort of the project; this work presents the first stage resulting in a high-level metadata model that is agnostic to the actual data management situation in participating organizations yet is able to capture significant features of nanoscience physical and computational experiments.
The major purpose of any metadata is satisfying information needs of a certain community. “Community” should be understood in broad terms and includes machine agents, to ensure human-to-human, human-tomachine and machine-to-machine interoperability.
The information needs may be generic (common with other communities) or specific for a particular community. From the project perspective, the information needs should be expressed as clearly formulated Use Cases for the existing or proposed information and data management systems (IT platforms). A good metadata design should take into account user requirements and IT architecture, and in turn should feed considerations for the IT architecture.
The IT architecture, the use cases and practices, and the metadata design can be considered pillars of enterprise architecture that includes both technological and organizational aspects of a loosely coupled virtual enterprise that the NFFA project is going to deliver for the European nanoscience community.
The main purpose of metadata design effort in NFFA project can be formulated then as giving the adequate support for that widely defined enterprise architecture for nanoscience. This has an implication of metadata design from “first principles”, i.e. by pondering over existing best practices of information management, use cases for nanoscience and information technology opportunities (and limitations) rather than adopting any existent metadata standard.
The case for metadata collection and use can be specific to nanoscience, yet there are general information needs that are typical for a wide variety of users and that have been developed in other branches of science and information management.
One of the mature information design frameworks is
Functional Requirements for Bibliographic Records [
Another mature information design framework of
relevance is the Reference Model for an Open Archival
Information System [
Overall, the OAIS framework should be able to provide a good coverage of what NFFA needs to consider for sensible data collection, archiving and provision towards the end users (researchers in nanoscience), and the FRBR framework should be able to cover the end user needs for information retrieval. The respective areas of coverage and user categories relevant to NFFA are illustrated by the following table: Table 1 Information management frameworks and their coverage of NFFA scope.
management and dissemination
(nanoscience researchers) Information Ingest data Find data needs (user Manage data Identify data tasks) Disseminate Select data data Obtain data Being general in nature, OAIS and FRBR are still able to provide good recommendations for NFFA practices of information and data management. In particular, OAIS emphasizes the need of having a clear agreement between the data producer and the archive, and a clearly defined format for data exchange between them – so called Submission Information Package, whilst FRBR emphasizes the importance of having a clear identity for data assets.
A questionnaire was used to collect the NFFA partners’ responses about their data management practices and most popular data management solutions. The questionnaire inquired on the following aspects of data management in nano-facilities: Intensity of experiments and of resulting data flow Popular data formats Data catalogue software Data catalogue openness Data management policy Metadata standards for data catalogue Persistent identifiers for data User management platform Popular third-party databases and information systems
In total, seventeen responses out of the 20 project partners were received and reviewed. They showed very different levels of data management maturity. From the responses, the following priorities the metadata design were identified: One experiment to many samples and one sample to many data files relationships should be supported. A common set of metadata fields for data discoverability should be agreed upon, possibly based on an existing popular standards or recommendation for data discovery. User roles with different permissions for access to metadata should be developed. This means the metadata model will need to represent users as well as data. It is reasonable to develop a common data management policy for NFFA, or a set of policies with different flavours of access to data. Having links to external reference databases is valuable to ensure the high quality of metadata yet this will mean additional effort so should be de-scoped from the initial design of metadata.
In addition to the questionnaire where responses
were collected from research offices or relevant research
programme representatives, a common vocabulary of
terms and definitions relevant to nanoscience data
management was compiled and then refined by the IT
teams of participating NFFA organizations ([
A particularly important use case to be supported by the metadata model should be the situation when the same researcher (or a research group) applies for experimental time on more than one facility – as the nature of experiment may require this – yet the researcher wants a seamless experience across nanoscience facilities, with a single entry point for data management.
Another conclusion based on responses to the questionnaire is that computational experiments in nanoscience become common and can be mixed up with physical experiments, so there should not be an artificial division between the two.
As an additional consideration for principal metadata design, we used the draft NFFA Data System Architecture that defines the outline design of the NFFA portal, which considered the generic use case of the same user performing a measurement on two different facilities. Generic use cases when one user wants to access data produced by another user, or wants to release data into the public domain are currently not being considered. These may be considered in future, so should be taken into account within an extensible metadata design.
The draft architecture suggests that data should be
harvested from individual facilities in a suitable
“packaged” format, with METS [
After considering the draft architecture, the conclusion was that we could take METS as “the role model” metadata standard for data packaging that corresponds to a specific entity in the NFFA generic metadata model – Data Asset. As to particular elements of metadata suggested by the IT architecture draft, the fields for capturing intellectual property information and provenance are easily most important ones as they affect the data assets reusability that should be one of the important outcomes of the NFFA project.
The top-down, bottom up and side requirements resulted in the basic structure of the proposed metadata model that is illustrated by Figure 1.
The suggested metadata elements are presented as a matrix in Table 2 to make explicit the coverage of identified information entities (common vocabulary terms) and of earlier identified information needs categories of them, see Section 2.2).
Certain elements are in common with the Core Scientific
Metadata Model ([
The metadata elements suggested are not all we need for having a successful metadata framework in NFFA. In addition, there should be established metadata management practices, ideally assisted by clear recommendations for NFFA partner organizations of how to assign and curate metadata.
For example, there are choices of how you aggregate data: let us say all data files for all samples measured in a particular Experiment can be assembled in one package, and then the package is given common descriptions like Facility name, research User name, Data Policy etc. However, this may not suit actual data management practices or policies of certain Facilities, e.g. they may want to make a Sample rather than an Experiment a focal point of their metadata descriptions.
These operational aspects of NFFA metadata implementation will require further engagement and discussions with data practitioners in NFFA.
The NFFA metadata development so far has produced an
agreed common approach with its mapping to the existing
metadata frameworks and best practices. It has defined a
common vocabulary, the provisional list of mandatory
and optional attributes, and the ER diagram that can be
used both in metadata design and in IT architecture
design. The high-level metadata model will be further
refined through project work in NFFA and through
discussions in the wider nanoscience community. Also
the state-of-the-art metadata development for
nanoscience that may cover specific entities in our
generic metadata model, e.g. CODATA UDS [
This work is supported by Horizon 2020 NFFA-Europe project www.nffa.eu under grant agreement no. 654360.