How to specify a PROJ string in a CF-netCDF file?

[neumannd]

One might want add a PROJ string to a grid mapping description as additional (redundant) information. Currently, no attribute name (e.g. proj_string) is specified for this purpose in the CF Conventions.

There exists a specific crs_wkt attribute in Table F.1 of the CF Conventions. However, no similar attribute for a PROJ string is provided although the Notes of most grid mapping descriptions in Appendix F point to the PROJ software package.

Does there exist a best practice where/how to specify a PROJ string?

By the way: PROJ4 is called PROJ, now, and recently advanced to versions 5, 6 and 7 after two decades of version 4. Therefore, several URLs in Appendix F don't work anymore. I will create a issue and a pull request next week for this (issue cf-convention/cf-conventions#253; pr cf-convention/cf-conventions#254).

[JonathanGregory]

I do not recall that this question has been discussed before and it's a good point. I would support the definition of such an attribute of the grid_mapping variable, provided that (like the WKT) it was regarded as subordinate to the attributes specifying the projection.

[TomLav]

We systematically add a :proj_string attribute to our crs variable.

Having a standardized name for such an attribute could help and avoid the zoo of :projstr, :proj4string, :proj4_string, etc... etc...

[JimBiardCics]

How about proj_definition or proj?

[TomLav]

Hei Jim,

I think "proj string" is very much accepted in that sense. For example this is how the "quick start" of the PROJ software begins:

Coordinate transformations are defined by, what in PROJ terminology is known as, “proj-strings”. A proj-string describes any transformation ...

So it would makes sense (at least to me!) to stay close to the existing PROJ terminology. But of course, there are many options, including "crs_proj" (that would relate to the existing "crs_wkt").