How should we handle hybrid coordinates in iHAMOCC?

[TomasTorsvik]

@JorgSchwinger @jmaerz @matsbn

We should make some changes in iHAMOCC to accommodate the change to hybrid coordinates. At the moment it is assumed that the top two layers represent the surface mixed layer (layer 1: top 10m water column, layer 2: rest of the mixed layer). Special treatment of the surface mixed layer mostly reference the INTEGER, PARAMETER :: kmle = 2 index (i.e. do k=1,kmle etc.) that is defined in the mo_vgrid module. The relevant modules/subroutines where this comes into play (that I have found):

Preformed tracers in the mixed layer

These tracers are initialized in beleg_vars and updated each time step in preftrc.

ocetra(i,j,k,iprefo2) =ocetra(i,j,k,ioxygen)
ocetra(i,j,k,iprefpo4)=ocetra(i,j,k,iphosph)
ocetra(i,j,k,iprefalk)=ocetra(i,j,k,ialkali)
ocetra(i,j,k,iprefdic)=ocetra(i,j,k,isco212)

Saturation DIC concentration in mixed layer

carchm does some calculations specifically for the surface layer (i.e. if (k.eq.1) then etc.), and at one point calculates the saturation DIC concentration in the mixed layer, without reference to kmle

ocetra(i,j,k,  idicsat)=tc_sat * rrho ! convert mol/kg to kmol/m^3
ocetra(i,j,k+1,idicsat)=tc_sat * rrho ! k+1 = the rest of the mixed layer

Riverine tracer inputs

Riverine tracer inputs are distributed over the surface mixed layer, for instance

ocetra(i,j,1:kmle,iano3) = ocetra(i,j,1:kmle,iano3) + rivin(i,j,irdin)*fdt/volij

Nitrogen fixation

The cyano subroutine is limited by kmle to update tracers only in the surface mixed layer.

Particle aggregation

This only comes into play if the AGG pre-processor flag is defined. In this case the following parameters from beleg_parm are used

shear = 43200.   !shear in the mixed layer,   0.5  d-1
fsh = 0.163 * shear *dtb
nmldmin = 0.1    ! minimum particle number in mixed layer

This comes into play in ocprod to set the minimum particle number in the mixed layer

if ( k <= kmle ) then
   ocetra(i,j,k,inos) = MAX(nmldmin,ocetra(i,j,k,inos))
endif

and in the calculation of the aggregate and dustagg fields.

[TomasTorsvik]

One possible solution, if we want minimal changes to the behavior in iHAMOCC, would be to change kmle from an integer parameter to a 2D mask field, with values corresponding to the depth of the surface mixed layer. We could then make most of the changes very minimal by doing do k=1,kmle(i,j), or something similar. Then we can also ensure that the new formulation can be backwards compatible with using isopycnic coordinates, i.e. set kmle(i,j) = 2 everywhere.

[JorgSchwinger]

Hi Tomas
thanks for identifying all critical points related to kmld. I think your suggestion to use a spatially explicit kmle-array is good. We just have to discuss whether this can be passed by BLOM (mld is calculated somewhere in BLOM) or whether iHAMOCC calculates kmld (density is available, so calculating kmld according to some density criterion in mo_vgrid should be doable).

@matsbn, what do you think about passing an index field defining the depth of the mixed layer from BLOM to iHAMOCC?

[jmaerz]

Hi @JorgSchwinger and @TomasTorsvik , I was a bit reading on the preformed tracers (particularly oxygen-related), looked into the code and started wondering, why we update them through the full mixed layer depth and not only at surface? Particularly in deep mixed layers (polar & subpolar regions) this update thus includes already remineralization processes (when kmle > kwrbioz). With the new hybrid coordinates, this approach seems even more questionable to me. Do you have any thoughts on it? Any reasons for this approach (and not only updating surface values)?

I just spoke briefly with Nadine about it and she mentioned some discussion with you, @JorgSchwinger earlier these days about the same issue. I am currently debating, whether it would make sense to introduce another preformed oxygen tracer (or all of the preformed tracers) in my simulations (only updating at the surface) which would allow to estimate the difference to what was done before.

[JorgSchwinger]

I think this is a question of how you define "preformed". My understanding was (is) that mixing inside the mixed layer is quick, so that it doesn't really make a difference. The bigger problem with the current implementation is how we initialize the preformed tracers: Currently it is initialized to zero (except in the mixed layer), so that it takes extremely long to get meaningful values in some places (deep North Pacific). It would probably be better to initialize to an estimate of preformed tracers using WOA/GLODAP - then it would still be off (=not an estimate according to our model) in less ventilated regions, but at least have a meaningful value there.