Output sea flux species dry deposition velocity with DryDep collection

[yantosca]

Name and Institution (Required)

Name: Bob Yantosca
Institution: Harvard + GCST

Describe the update

This is the companion PR to issue #2564 by @eamarais. This moves the dry deposition velocity for HEMCO extension seaflux species and satellite diagnostic species to the DryDep diagnostic collection.

Expected changes

See: #2564 (comment)

Reference(s)

N/A

Related Github Issue

• Closes Output sea flux species dry deposition velocity with DryDep collection #2564

[msulprizio]

Looks good to merge @yantosca!

[lizziel]

I think this PR will break the drydep diagnostics for full mixing since HEMCO only applies drydep fluxes if using non-local mixing.

[lizziel]

This is a bug fix we should note in the changelog. These diagnostics must have been many times larger than they should be before this fix.

[yantosca]

@lizziel: We can test this to make sure we get zero-diff output for this diagnostic

[yantosca]

@lizziel: I did a test run for this diagnostic and it seems to be zero-diff. I think the only thing that would be different is that the old way would take much longer since you're adding an array on each iteration instead of just a grid box.

[eamarais]

@lizziel, would a conditional "if non-local-pbl" statement resolve this?

[yantosca]

@lizziel, would a conditional "if non-local-pbl" statement resolve this?

That is precisely what I did. In comment d9ee2e2 I restored the lines that were taken out of drydep_mod.F90 and wrapped them in an IF statement that will only execute when the full PBL mixing is selected. I will post plots to verify that this works.

[yantosca]

I've restored the archival of DryDepVel and SatDiagnDryDepVel diagnostics to drydep_mod.F90 when the full PBL mixing is selected in commits d9ee2e2 and c7cfcfa. In the difference plots below, we show ACET and NO2 when using TURBDAY (full PBL) as Ref, and VDIFF (non-local PBL) as Dev.

The differences in ACET are attributed to the HEMCO seaflux extension, but there are also some small differences in the ratio plots. This may be numerical noise.

@eamarais: There are several differences in the NO2 drydep velocity. This may be due to the fact that when TURBDAY is used, we take dry dep velocity from State_Chm%DryDepVel and when we use VDIFF we compute dry dep velocity from the dry dep flux in hco_interface_gc_mod.F90:

        !------------------------------------------------------------------
        ! Also add drydep frequencies calculated by HEMCO (e.g. from the
        ! air-sea exchange module) to DFLX.  These values are stored
        ! in 1/s.  They are added in the same manner as the DEPSAV values
        ! from drydep_mod.F90.  DFLX will be converted to kg/m2/s later.
        ! (ckeller, 04/01/2014)
        !------------------------------------------------------------------
        IF ( DepSpec ) THEN
          CALL GetHcoValDep( Input_Opt, State_Grid, NA, I, J, L, found, dep )
          IF ( found ) THEN
             dflx(I,J,NA) = dflx(I,J,NA) + ( dep                   &
                            * State_Chm%Species(NA)%Conc(I,J,1) &
                            / (AIRMW / ThisSpc%MW_g)  )
             ! Get dry deposition velocity in cm/s:
             dvel(I,J,NA) = dvel(I,J,NA) + ( dep * State_Met%BXHEIGHT(I,J,1) * &
                                             1.0e+2_fp )
          ENDIF

The differences are pretty small but are these what we would expect?

Also for NO2, the differences are mostly over the continents. That might rule out e.g. PARANOX.

[yantosca]

@eamarais @lizziel @msulprizio: Another thought, the way in which we compute dry deposition velocity in hco_interface_gc_mod.F90 might be dependent on the concentration of the species (which is included in the drydep flux). This might vary slightly from version to version as opposed to the dry dep velocity in State_Chm%DryDepVel, which is computed from purely from resistances and surface parameters. air mass which determines the grid box height.

In drydep_mod.F90 we have:

       ! THIK = thickness of surface layer [m]
       THIK   = State_Met%BXHEIGHT(I,J,1)
       ... 
          !-----------------------------------------------------------
          ! Compute drydep velocity and frequency
          !-----------------------------------------------------------

          ! Dry deposition velocities [m/s]
          State_Chm%DryDepVel(I,J,NDVZ) = DVZ / 100.e+0_f8

          ! Dry deposition frequency [1/s]
          State_Chm%DryDepFreq(I,J,D) = State_Chm%DryDepVel(I,J,NDVZ) / THIK

But if AIRQNT is called to reset the airmass quantities in between the time when drydep is called and when the Compute_Sflx_for_Vdiff is called this could cause the differences observed when comparing full mixing vs. non-local mixing.

Ideally for a species that has no air-sea exchange in HEMCO we'd want to see differences that are more on the order of numerical noise than we do now.

[eamarais]

Hi @yantosca, this does look odd. Is this issue as a result of switching between TUBDAY and VDIFF? Or has my updated handling of the DryDepVel diagnostic introduced this issue? Not looking to avoid blame. Just on how to resolve it.

[yantosca]

@eamarais: I suspect it is a difference in air mass (thus causing a difference in box height) between where drydep is called an where hco_interface_gc_mod is called. I will try to investigate further and generate some debug output which should clear things up.