Introduce global attributes "license" and "copyright"?

[rjhogan]

I am planning to release some of the static fields used by the ECMWF model (e.g. aerosol and albedo climatologies) in netCDF format, and would like to express the copyright and the license in the global attributes of the file. This would seem to me good practice, just as it is would be when releasing open-source software, and indeed for data Zenodo requires the license to be stated. But I can't find any reference to "license" or "copyright" global attributes in the CF conventions - are there plans to add these? I guess we could have license="CC-BY-4.0", and copyright="2024 ECMWF" but it would be nice to follow a convention if there is one. Note that the CC-BY-4.0 license requires the licensee to acknowledge the creator and copyright holder of the data (if provided) in any downstream products, so the ability to state the creator/copyright-holder in netCDF metadata would go together with the ability to state the license.
Robin Hogan (ECMWF)

[ajelenak]

The Attribute Convention for Data Discovery has the license attribute and is designed to fit nicely within the CF conventions. Also has several attributes for data creator.

[taylor13]

@ajelenak I was not aware of these conventions. Do you know whether they have been adopted by any large projects? What advantages would there be in the next phase of CMIP to changing the names of some of the global attributes used in past CMIP phases to be consistent with the Data Discovery conventions.

[ajelenak]

It is strongly recommended for all NASA satellite data, but soon it will be required. It is also recommended for NOAA satellite data (included in the NCEI's file templates).

[rmendels]

Most gridded datasets in the many ERDDAP servers have these conventions in their metadata, and that is a lot of datasets. I believe, but could be wrong, that most NOAA Coastwatch and NOAA IOOS datasets use these conventions.

[taylor13]

thanks, @ajelenak and @rmendels . Does anyone else have any experience with datasets conforming to the Attribute Convention for Data Discovery? Seems like CMIP should probably work toward consistency with them.

[rmendels]

I should add that ERDDAP allows you to "add" metadata to existing datasets without having to modify existing files. That is why many ERDDAP datasets have these metadata even though the underlying files may not.

[rjhogan]

Thank you for your comments! It is surprising that no-one has thought to add a "copyright" global attribute when most data licenses (e.g. CC-BY-4.0, clause 3.a.1.A.ii) explicitly require that copyright statements are retained when sharing modified versions of a dataset. The copyright holder is usually the institution rather than the individual creator. Perhaps I should contact the ACDD people, but it is less clear how to do this.

[ajelenak]

Perhaps I should contact the ACDD people, but it is less clear how to do this.

I'm one of them. @ethanrd is another, and the creator of initial ACDD.

[bzah]

@taylor13

Does anyone else have any experience with datasets conforming to the Attribute Convention for Data Discovery?

Eumetsat's MTG products also comply with the ACDD convention.

[MathewBiddle]

NCEI's netCDF templates build off CF and ACDD. https://www.ncei.noaa.gov/netcdf-templates