Intern:Versioning

When can a dataset be corrected, edited or rewritten without a new version?
Datasets without a registered DOI
 * Any change of data or metadata including deletion is possible.
 * Exception: datasets connected to articles with a public review (e.g. ESSDD), which link to the dataset in review (=> keep a copy of in-review-version and link it as a reference "Original version" in the updated version).
 * The changes are not registered in the dataset history.

Datasets with a registered DOI

Overview
This is an overview table of what can and should be done when a registered dataset (DOI active) needs to be changed:

OLD:{
 * These data sets can never be deleted.
 * Small changes, especially those intended for metadata enrichment, leading to an improved FAIRness, are possible:
 * Typos in authors, title, abstract or comments. Document the changes in the.
 * Small changes (e.d. adding sentences to make the description clearer) to abstract or comments. Document the changes in the.
 * Adding/correcting information to Events. If this information can have impact on interpretation, e.g. correcting Latitude, Longitude, Date/Time,..., document the changes in the.
 * Adding/correcting keywords is possible. Usually no need to document it in the.
 * Correcting parameters: If this information can have impact on interpretation, e.g. units, document the changes in the.
 * Adding/correcting references is possible. Usually no need to document it in the.
 * Adding projects & awards is possible. Usually no need to document it in the.


 * If the data were password protected, also changes to datasets can be done.
 * Corrections in data series (e.g. recalculation of series due to wrong factor (cm->m, a ->ka, ...) or other "small" and traceable changes if necessary, are possible. Always document such changes in the .}

When to archive a new version?
For other cases (major change in DOI-registered dataset), archive a new version.

Archiving a new version
Old dataset:
 * Add the new dataset (new version) as a reference with relation "replaced by"
 * This way the dataset cannot be found anymore in PANGAEA search '''
 * In most cases: change the restriction to "access rights needed"
 * This prevents the dataset being downloaded and used although potentially wrong
 * The old version is not included in the Data Warehouse
 * Refer to the new version in dataset : “A new version of the data has been uploaded, and can be found here: doi:yyyyyy.”
 * Configure the record out of the parent

New dataset:
 * The same title can be used.
 * Add comment with Version number, date and reason why a new version was added, e.g. "Version 2, 2020-02-29: Previous version doi:yyyyyy contains errors due to data processing." or "This version replaces ... doi:yyyyyy."
 * Optional: add link to old version, relation "original version".

Dataset History
The data set table in 4D (tab Details) contains a data set history record comprising of:
 * Date/Time of the change record (automatic entry),
 * Editor name responsible for the change (automatic entry) and
 * Comment, describing the change.

In this, any relevant changes to the data are documented. Date/Time and Comment are stored in metadata and shown in the data set web view as "Change history:".

Examples of dataset history comments:
 * Latitude and Longitude for station XXX (wrongly reported in expedition report) were corrected
 * "Depth, water" values were corrected for samples XXX and YYY (correction submitted by authors)
 * List of datasets containing history record: https://www.pangaea.de/?q=history%3A*

Retraction of the data
A dataset with a registered DOI is considered published and cannot be withdrawn. A legal basis for publishing data and not deleting it after registration is provided by our Terms of Use.

If, however, a registered data record absolutely has to be withdrawn, then the possibility exists as follows: Example:
 * Replace dataset with a "stable link" dataset, link to https://doi.pangaea.de/10.1594/PANGAEA.823081
 * Set Author/PI "Anonymous"
 * Set Dataset Title: "object deleted on request of the author, download disabled"
 * Set dataset protection to "access rights needed"
 * Delete anything else, e.g., project, keywords, events, reference, ...