Initialising areaC

[anton-seaice]

Hi

I was going through the code to intialise the MOM grid, and have a question about the areas used on for the C grid U and v points.

The normal grid areas (for the tracer points and the the b-grid velocity point are initialised from the mom_mosaic grid file:

  call MOM_read_data(filename, 'area', tmpT, SGdom, scale=US%m_to_L**2)
  call pass_var(tmpT, SGdom)
  call extrapolate_metric(tmpT, 2*(G%jsc-G%jsd)+2, missing=0.)


  do j=G%jsd,G%jed ; do i=G%isd,G%ied ; i2 = 2*i ; j2 = 2*j
    G%areaT(i,j) = (tmpT(i2-1,j2-1) + tmpT(i2,j2)) + &
                   (tmpT(i2-1,j2) + tmpT(i2,j2-1))
  enddo ; enddo
  do J=G%JsdB,G%JedB ; do I=G%IsdB,G%IedB ; i2 = 2*i ; j2 = 2*j
    G%areaBu(I,J) = (tmpT(i2,j2) + tmpT(i2+1,j2+1)) + &
                    (tmpT(i2,j2+1) + tmpT(i2+1,j2))
  enddo ; enddo

This seems as expected - it sums the area of the relevant four cells in the mosaic file.

However the area for the C-grid points are initialised assuming each grid cell is square and multiplying the side-lengths:

  do j=G%jsd,G%jed ; do I=G%IsdB,G%IedB
    ! This open face length may be revised later.
    G%dy_Cu(I,j) = G%mask2dCu(I,j) * G%dyCu(I,j)
    G%areaCu(I,j) = G%dxCu(I,j) * G%dy_Cu(I,j)
    G%IareaCu(I,j) = G%mask2dCu(I,j) * Adcroft_reciprocal(G%areaCu(I,j))
  enddo ; enddo


  do J=G%JsdB,G%JedB ; do i=G%isd,G%ied
    ! This open face length may be revised later.
    G%dx_Cv(i,J) = G%mask2dCv(i,J) * G%dxCv(i,J)
    G%areaCv(i,J) = G%dyCv(i,J) * G%dx_Cv(i,J)
    G%IareaCv(i,J) = G%mask2dCv(i,J) * Adcroft_reciprocal(G%areaCv(i,J))
  enddo ; enddo

Why is areaCu and areaCv not initialised following the same method as areaT and areaBu? The differences will be significant near the poles of the grid ...

Thanks

Anton

[herrwang0]

Hi @anton-seaice,

The reason is that grid widths at velocity point G%dy_Cu and G%dx_Cv can be later modified by CHANNEL_CONFIG. This allows narrow channels to have their actual widths, rather than limited by the model's grid size.

See how G%dy_Cu, G%dx_Cv and velocity point areas are modified by the following subroutines, depending on the choices of CHANNEL_CONFIG. This is 14 lines after subroutine initialize_masks is called.

MOM6/src/initialization/MOM_fixed_initialization.F90

Lines 113 to 132 in 80d8b5f

	call get_param(PF, mdl, "CHANNEL_CONFIG", config, &
	"A parameter that determines which set of channels are \n"//&
	"restricted to specific widths. Options are:\n"//&
	" \t none - All channels have the grid width.\n"//&
	" \t global_1deg - Sets 16 specific channels appropriate \n"//&
	" \t\t for a 1-degree model, as used in CM2G.\n"//&
	" \t list - Read the channel locations and widths from a \n"//&
	" \t\t text file, like MOM_channel_list in the MOM_SIS \n"//&
	" \t\t test case.\n"//&
	" \t file - Read open face widths everywhere from a \n"//&
	" \t\t NetCDF file on the model grid.", &
	default="none")
	select case ( trim(config) )
	case ("none")
	case ("list") ; call reset_face_lengths_list(G, PF, US)
	case ("file") ; call reset_face_lengths_file(G, PF, US)
	case ("global_1deg") ; call reset_face_lengths_named(G, PF, trim(config), US)
	case default ; call MOM_error(FATAL, "MOM_initialize_fixed: "// &
	"Unrecognized channel configuration "//trim(config))
	end select