MartinaSt
commented
9 months ago Dear @jesusff
DRS changes is effort for the ESGF especially if the data with two different DRS should be searchable together. The function of the DRS or the DRS_id used by ESGF is to ensure uniqueness of a dataset. If you do not need the additional DRS element (experiment_id) for the uniqueness, you could add it as global attribute in the netCDF header. ESGF can index it, which makes it discoverable using the search field even if no facet is added.
If you need the information on experiment_id for the uniqueness of datasets, you could either combine the information or have it separate. Both are not ideal if you want to search the data of this FPS together with other CORDEX data. The combination requires a splitting of this DRS component in case the search facet should only include the part used by the other CORDEX experiments (without experiment_id). Adding the experiment_id as DRS component requires a different publication configuration and it might have implications on other automated processes like the data replication. A combined DRS component would make less problems or in other words using the same DRS for all experiments of a certain project_id is preferable. The discussion within CMIP6 for the DCPP initialization also resulted in this solution of a combined "rifs" element.
The separation of the FPS-URB-RCC from the other CORDEX experiments in a specific ESGF project is always possible but might not be what you want.
Dear @MartinaSt
In the new DRS for CORDEX-CMIP6, we are considering different activities under the same project_id (CORDEX). For the initial activities (dynmical and statistical downscaling at continental scale), the DRS is consistent and both activities can be described with the same elements. We are leaving ground for other activities to develop their own DRS, based on the general CORDEX-CMIP6 one, but I'm not sure how free we are to add new elements to the DRS.
One such additional activity is CORDEX flagship pilot studies (FPS). In particular, here we are trying to develop a DRS for the FPS-URB-RCC. In this case, we have one additional element (experiment_id), which is not in the general CORDEX-CMIP6 DRS. Considering that all CORDEX activities might end up under the same search interface on ESGF, could you guide us on how free are we to add elements to the filename or to the directory hierarchy for a particular activity?
I remember from the CORDEX-FPSCONV, that we packed some extra information in the version element (e.g. fpsconv-x2yn2-v1) in order not to alter the existing CORDEX DRS. Should we resort to such "tricks" to keep the consistency across activities? e.g. in CMIP6 DCPP the initialization date for decadal prediction was also packed into the "ripf" element (e.g. s1960-r2i1p1f1).
Some more backgroud info in #2 and https://github.com/WCRP-CORDEX/cordex-cmip6-cv/issues/20 , but I don't want to bother you with too much details.
Thanks in advance!