Data Working Group

Ongoing updates to a time series do change the content of the data set, but they do not typically constitute a new version or edition of a data set. New versions typically reflect changes in sampling protocols, algorithms, quality control processes, etc. Both a new version and an update may be reflected in the release date.

Locator, Identifier, or Distribution Medium

Then it is necessary to include a persistant reference to the location of the data.

This may be the most challenging aspect of data citation. It is necessary to enable "micro-citation" or the ability to refer to the specific data used--the exact files, granules, records, etc.

Data stewards should suggest how to reference subsets of their data. With Earth science data, subsets can often be identified by referring to a temporal and spatial range.

A particular data set may be part of a compilation, in which case it is appropriate to cite the data set somewhat like a chapter in an edited volume.

Increasingly, publishers are allowing data supplements to be published along with peer-reviewed research papers. When using the data supplement one need only cite the parent reference. F

Confusingly, a Digital Object Identifier is a locator. It is a Handle based scheme whereby the steward of the digital object registers a location (typically a URL) for the object. There is no guarantee that the object at the registered location will remain unchanged. Consider a continually updated data time series, for example.

While it is desirable to uniquely identify the cited object, it has proven extremely challenging to identify whether two data sets or data files are scientifically identical.

At this point, we must rely on location information combined with other information such as author, title, and version to uniquely identify data used in a study.

The key to making registered locators, such as DOIs, ARKS, or Handles, work unambiguously to identify and locate data sets is through careful tracking and documentation of versions.

how to handle different data set versions relative to an assigned locator.

Track major_version.minor_version.[archive_version].

Typically, something that affects the whole data set like a reprocessing would be considered a major version.

Assign unique locators to major versions.

Old locators for retired versions should be maintained and point to some appropriate web site that explains what happened to the old data if they were not archived.

A new major version leads to the creation of a new collection-level metadata record that is distributed to appropriate registries. The older metadata record should remain with a pointer to the new version and with explanation of the status of the older version data.

Major and minor version should be listed in the recommended citation.

inor versions should be explained in documentation

Ongoing additions to an existing time series need not constitute a new version. This is one reason for capturing the date accessed when citing the data.

we believe it is currently impossible to fully satisfy the requirement of scientific reproducibility in all situations

To aid scientific reproducibility through direct, unambiguous reference to the precise data used in a particular study. (This is the paramount purpose and also the hardest to achieve). To provide fair credit for data creators or authors, data stewards, and other critical people in the data production and curation process. To ensure scientific transparency and reasonable accountability for authors and stewards. To aid in tracking the impact of data set and the associated data center through reference in scientific literature. To help data authors verify how their data are being used. To help future data users identify how others have used the data.

The ESIP Preservation and Stewardship cluster has examined these and other current approaches and has found that they are generally compatible and useful, but they do not entirely meet all the purposes of Earth science data citation.

In general, data sets should be cited like books.

hey need to use the style dictated by their publishers, but by providing an example, data stewards can give users all the important elements that should be included in their citations of data sets

Access Date and Time--because data can be dynamic and changeable in ways that are not always reflected in release dates and versions, it is important to indicate when on-line data were accessed.

Additionally, it is important to provide a scheme for users to indicate the precise subset of data that were used. This could be the temporal and spatial range of the data, the types of files used, a specific query id, or other ways of describing how the data were subsetted.