Add soil increments to deterministic member

[ClaraDraper-NOAA]

What new functionality do you need?

The soil analysis (soil moisture, soil temperature) is done using the EnKF only, and we do not have a varitional soil analysis available for the deterministic member. Instead, we will apply the (upscaled) ensemble mean EnKF increment to the deterministic member.

This will require adding code for: 1) calculation of the ensemble mean increment [add to enkfgdaseupd job] 2) re-gridding the (Gaussian, ens res) mean increment to the deterministic native model resolution [in gdassfcanl?]

Initially, will use pre-existing sfc update executable (from UFS_UTILS) in gdassfcanl to add the deterministic increment (later, replace with IAU). This will require adding a dependency on enkfgdaseupd to gdassfcanl.

What are the requirements for the new functionality?

If GSI_SOILANAL option is selected, the ensemble mean soil increment is added to the deterministic member.

Acceptance Criteria

If GSI_SOILANAL option is selected, the ensemble mean soil increment is added to the deterministic member. If not, output should be zero diff.

Suggest a solution (optional)

No response

[ClaraDraper-NOAA]

@aerorahul @WalterKolczynski-NOAA Do either of you know why we're calculating the ensemble mean increment using a purpose-designed executable, rather than using the option in the EnKF to output the ensemble mean increment? (asking, as I need to add calculation of the ensemble mean increment for the soil analysis, and am wondering if I can just use the EnKF output option).

[ClaraDraper-NOAA]

@aerorahul @WalterKolczynski-NOAA Do either of you know why we're calculating the ensemble mean increment using a purpose-designed executable, rather than using the option in the EnKF to output the ensemble mean increment? (asking, as I need to add calculation of the ensemble mean increment for the soil analysis, and am wondering if I can just use the EnKF output option).

I figured it out. The code to write out increments from within the enkf requires the background (for conversion to Tv, and things like that). The background mean file is not available. Also it's non-linear, so using the mean background to write out the mean increment wouldn't give the same answer as calculating the mean of the ensemble output increments.