Remove the eomg job

[DavidHuber-NOAA]

What new functionality do you need?

The eomg job is no longer supported for the GSI and the GDASApp has its own jobs to handle observational diagnostics (i.e. when lobsdiag=".true."): atmensanlobs and atmensanlsol. Thus, the eomg job should be removed from the global workflow without affecting the new atmensanlobs and atmensanlsol jobs.

What are the requirements for the new functionality?

All references to eomg are removed and job-specific scripts are removed from the global workflow without affecting the new GDASApp jobs. Also, eomg-specific archiving is removed and/or replaced with GDASApp-specific archiving instructions.

[aerorahul]

@CatherineThomas-NOAA Is the functionality of eomg.sh and its use needed to be supported?

[CatherineThomas-NOAA]

@aerorahul I don't think so, especially given that we didn't notice the job was missing. Let me ask around a bit though.

[aerorahul]

Thanks @CatherineThomas-NOAA. In that case we will eliminate the configurable option for lobsdiag_for_enkf and simplify the scripting.

[RussTreadon-NOAA]

g-w issue #2415 was opened to add JEDI ATM equivalents to to eomg and eobs. These jobs perform observation quality control and h(x) for ensemble member states, eomg, and ensemble mean states, eobs. While we may not need eomg.sh for GSI ATM DA. We will likely be adding it for JEDI ATM DA.

[DavidHuber-NOAA]

Adding blocked while waiting for the GDASApp-equivalent job to replace eomg.

[RussTreadon-NOAA]

While we do not have a separate eomg job for GDASApp, the functionality of eomg currently exists in g-w. When GDASApp is run with lobsdiag_forenkf=".false.", we run a single job, atmensanlletkf, which combines eomg and eupd into a single executation of gdas.x. The wall time and memory requirements for this combined approach make it unpractical for all but coarse resolutions with few ensemble members.

When GDASApp is run lobsdiag_forenkf=".true." we have two g-w jobs: atmensanlobs and atmensanlsol. Job atmensanlobs is the GDASApp equivalent of GSI-based eobs. The split approach is more efficient in terms of memory and wall time.

[DavidHuber-NOAA]

OK, so then it sound like this issue is OBE. Agreed @aerorahul @WalterKolczynski-NOAA?

[DavidHuber-NOAA]

I updated the description of this issue to reflect that the eomg job should be removed from the global workflow without affecting the new GDASApp equivalent jobs. I am also removing the blocked label.