[FEATURE REQUEST] Supporting curvilinear grids in MESSy/A2A regridding

[jimmielin]

Background

During IGC10 (Emissions and deposition working group?) there was discussion of support for regridding of lambert conformal grids for HEMCO. However, at present, HEMCO only supports regridding to/from rectilinear grids.

Feature request

It will be useful to support curvilinear grids such as Lambert in the HEMCO regridding for GEOS-Chem Classic, without introducing ESMF dependencies. This will also improve capabilities in WRF-GC and HEMCO within CESM (until we move to the CESM input/output system)

This implementation is technically possible because MAP_A2A and MESSY_NCREGRID should be able to do this, just that the feature was not implemented. For example, in HCO_REGRID_MOD's REGRID_MAPA2A, the following construct forces a rectilinear grid:

    ! get longitude / latitude sizes
    nLonEdge = SIZE(LonE,1)
    nLatEdge = SIZE(LatE,1)

    ! Write input grid edges to shadow variables so that map_a2a accepts them
    ! as argument.
    ! Also, for map_a2a, latitudes have to be sines...
    ALLOCATE(LonEdgeI(nlonEdge), LatEdgeI(nlatEdge), STAT=AS )
    IF ( AS /= 0 ) THEN
       CALL HCO_ERROR( 'alloc error LonEdgeI/LatEdgeI', RC, THISLOC=LOC )
       RETURN
    ENDIF
    LonEdgeI(:) = LonE
    LatEdgeI(:) = SIN( LatE * HcoState%Phys%PI_180 )

    ! Get output grid edges from HEMCO state
    LonEdgeO(:) = HcoState%Grid%XEDGE%Val(:,1)
    LatEdgeO(:) = HcoState%Grid%YSIN%Val(1,:)

Similarly in HCOIO_MESSY_MOD's HCO_MESSY_REGRID:

    !-----------------------------------------------------------------
    ! Destination grid description.
    ! This creates a MESSy axis object for the target (=HEMCO) grid.
    !-----------------------------------------------------------------

    ! Get horizontal grid directly from HEMCO state
    lon   => HcoState%Grid%XEDGE%Val(:,1)
    lat   => HcoState%Grid%YEDGE%Val(1,:)

...

    ! ----------------------------------------------------------------
    ! Assign longitude: this is always the first dimension
    ! ----------------------------------------------------------------
    IF ( ASSOCIATED(lon) ) THEN
       N        = N + 1
       ax(N)%lm = .true.     ! LONGITUDE IS MODULO AXIS

       ! Axis dimension
       XLON = SIZE(lon,1)

       ! FOR NOW, ASSUME NO DEPENDENCIES. NEED TO EDIT HERE
       ! IF WE WANT TO USE CURVILINEAR GRIDS
       ax(N)%ndp    = 1          ! LONGITUDE IS ...
       ALLOCATE(ax(N)%dep(1), STAT=status)
       IF ( status/= 0 ) THEN
          CALL HCO_ERROR ( HcoState%Config%Err, 'Cannot allocate lon dependencies', RC )
          RETURN
       ENDIF
       ax(N)%dep(1) = N          ! ... INDEPENDENT
       ndep_lon     = N

It should be possible to instead use the full 2-D information for input latitudes/longitudes. However I lack expertise in the MESSy regridding code to implement this immediately, so I am posting this on the HEMCO GitHub as a feature request so we could potentially figure out a solution.

[ktravis213]

Thanks for this request. I just want to add that as the EPA NEI emission (and other regional inventories) come in this format, this will be a very useful feature for streamlining future updates.