What granularity to choose for dataset DOIs ?

There is no definitive answer to that question; it is a choice that will depend on data and their usage by scientists. The DOIs main objective is to facilitate and improve the reliability of data citation in scientific publications: a strategy should be defined based on the level of aggregation of data that are generally used in a publication.

If a DOI is assigned at a measurement level and if the related publication manages thousands of these measurements, it will be difficult to cite properly the thousand associated DOIs. In that case, DOIs are probably not properly assigned.

TIt is likely that for scientific publications exploiting different levels of the same bunch of data, DOIs should be assigned at these specific distinct levels. In the Landing Page of such a DOI it is possible to provide links between the different DOIs so that the reader will navigate from one level to another.

How to manage different versions of the same dataset?

There are no strong rules on the management of versions of a dataset. If different versions of a dataset should continue to be accessible independently and specifically cited (e.g. with a list of different authors), a specific DOI should be assigned for each version.

It is also possible to provide access to different versions of a dataset with the same DOI. In this case, two choices are possible :

• All versions of the dataset are available freely (ex : https://doi.org/10.17882/42182)
• Access to old versions is restricted (ex : https://doi.org/10.17882/41888): previous versions are accessible via a request to authors. It is then possible to ensure the reproducibility of an experiment based on an obsolete version, but access to an obsolete version of the data set is intentional and not accidental

Setting a unique DOI to a dynamic dataset has many advantages :

• Citation instructions are easier to understand. Users only have one DOI to figure out.
• To improve the visibility of the DOI in search engines (eg Google), it is preferable to distribute a single Web Landing Page which concentrates a maximum of citations and therefore backlinks rather than multiple Landing Pages of almost identical content.
• If a specific DOI is set to each versions, search engines will not necessarily present the most recent version first in their results lists. A user can therefore discover an obsolete version of the dataset and not realize that a more recent version is available

Is that possible to update a dataset ?

Yes, it is possible to update metadata and data files even after the publication of the record. To update a record, please send an email to sismer@ifremer.

How to cite a specific subset with it’s fragment (#) ?

Some datasets allow the citation of a subset (eg: a version, a geographical area, etc.) using a DOI fragment (#) (e.g : https://doi.org/10.17882/42182#95967). The fragment’s key are available in the key column of the data table.

To cite a specific subset, click on its key and copy the specific citation that will be suggested in the How to cite section.