Implementing CoupledDA

[ateruzzi]

The coupled DA aims at use the physical DA (PHY DA) results to updated nutrients in biogeochemistry.
3DVar with coupled DA will:

• read the density increments provided by PHY DA
• read the pre-computed correlations between density and nutrients
• calculate the nutrient increments (nitrate and phosphate)
• update the restarts accordingly

[ateruzzi]

The name of the main subroutine has been updated from oceanvar to biovar (not necessary for coupled DA, just a matter of formality).

Starting to intorduce the reading of PHY DA increments.

[ateruzzi]

The code compiles correctly BUT it is only for testing density-nutrient DA in a simplified framework:

• The update is done on nitrate outside to the DA iteration

  3dVarBio/cnv_inn.f90

  Lines 65 to 68 in fc4b5dc

  	if (drv%densnut .eq. 1) then
  	grd%n3n(:,:,:) = 0.0
  	call ver_hor_densnut(grd%n3n)
  	endif

• Single profiles of density increments are read

  3dVarBio/get_densincr_arg.f90

  Line 1 in fc4b5dc

  subroutine get_densincr_arg

• Nitrate increments are calculated on the model grid levels

  3dVarBio/get_densincr_arg.f90

  Line 1 in fc4b5dc

  subroutine get_densincr_arg

• The horizontal operator is applied

  3dVarBio/ver_hor_densnut.f90

  Line 71 in fc4b5dc

  !********** APPLY RECURSIVE FILTERS ********** !

• Phosphate is updated accordingly and RSTafter files are written

  3dVarBio/biovar.f90

  Lines 140 to 142 in fc4b5dc

  	if (drv%dnc .eq. 1) then
  	call wrt_nut_stat
  	endif

NOTE1 that this version is for testing density-nutrent DA alone (nitrate profiles assimilation is overwritten).
NOTE2 To test the code an assimilation has to be performed (NOT nitrate)

[ateruzzi]

Compiled and tested for:

• assimilation of some O2 profiles
• apply increments on n3n and n1p based on a single density increment profile

NB: it does work only if drv%nut = 1

• density increments should be relatively dense on the vertical, as it is not interpolated by EOFs
• correlation radius quite small at the moment

[ateruzzi]

From now starting a new strategy (will overwrite the previous approach)

• The assimilation will be based on multivariate nutrient-density EOFs
• Density increments from Med-PHY will serve as pseudo-observations

The cost function will minimize the distance from nitrate observations (if availables) and from density increments (weighted by their uncertainties) and from the nitrate model background (expressed as the dot products of the EOF control space solution, i.e. the weigths v assigned to both nitrate and dens-nit EOFs.

[ateruzzi]

Makefile to be still updated and then to be compiled and tested.

[ateruzzi]

• Code has been changed to adopt the strategy of multivariate density-nutrient EOFs
• Check of consistency carried out with the help of AI https://github.com/inogs/3dVarBio/blob/CoupledDA/CODE_REVIEW_SESSION.md
• Code compiles

To be tested!

And to decide about the copyright policy