7. Metadata

7.1 Policy

• All GTOS data and information must have directory level metadata in accordance with GTOS guidelines

• All GTOS datasets must have adequate metadata at the dataset level.

7.2 Discussion

Metadata^[1] are "data about data", describing such things as the location, sources, general content, condition, format, etc. of existing datasets. They constitute documentation covering all aspects of the end-to-end data management process. Metadatabase systems are systems specifically designed to manage metadata i.e. to provide facilities for input, update, retrieval and reporting of data about data. Such systems may be used within a single institution to organise and maintain their own data holdings. They are also used on a broader level and can then provide a mechanism through which data producers can ensure that potential users are made aware of existing data, their nature, and how they might be obtained.

In general, metadata are at two levels. The first, referred to as "directory level", identifies the dataset through such items as a general description (subject, geographic coverage, dates, collection methods, processing done,...), details of availability (access conditions, costs,...), contact point (for further information and/or ordering). These are items which are essentially common to all types of dataset, regardless of the subject matter. The second or "dataset level", is subject matter specific, for instance, instrument settings, calibration data, adjustment factors, classification systems and legends, reference standards, taxonomies, etc.

In summary, directory level metadata enables a potential user to judge whether a dataset might be useful for the intended purpose and how to obtain it; the dataset level metadata allows the data to be used correctly, once obtained.

Metadata must exist to enable GTOS to deliver the desired high-quality data and information products to users. They are used specifically to facilitate data access and release (see 6.2), to enable quality assessments to be made (see 8.2), and in archiving (10.2).

Considerable work has been done relating to metadata at the directory level and, as in other areas, GTOS should build on what is currently in use. There is no single accepted global standard although there is commonality across some in wide use. Metadata systems have also been developed, specifically designed to manage metadata i.e. to provide facilities for input, update, retrieval and reporting of the directory level items. GTOS must have guidelines for directory level metadata that will be required for all GTOS products. These should be developed from existing standards in use. Appendix 4 offers a starting point for this, outlining some examples of metadata and identifying issues to be addressed.

GTOS Data Centres will almost inevitably have metadata specifications and systems in use and these will need to be in consistent with the guidelines. As the network of Centres is established, metadata content between the different institutions will need to be harmonised.

Specific mention should be made of JDIMP activities in this area. The Panel view metadata as key to "registering" G3OS datasets (see 8.2) and, in that context, are developing guidelines for metadata at the directory level. This is building on work done at the Australian Oceanographic Data Centre. One element, common to metadatabases, is the use of theme- or key-words to describe the content of datasets (with the aim of facilitating retrieval). Each observing system would be expected to provide a list of keywords.

As noted above, the requirements for dataset level metadata are dependent upon the particular dataset. GTOS-wide standards for this are impractical. Custodians should be encouraged to use standards or practices common in the particular subject matter field (see section 9).

7.3 Actions

• Identify metadata requirements at the directory level (with JDIMP) and produce guidelines

• Build metadata harmonisation methods (with DCs)

• Identify minimum generic metadata requirements at the dataset level for different data types