MARBL reset module

[matt-long]

In order to represent "preformed" tracers, we need a tracer reset capability, which will likely entail a new MARBL interface.

[mnlevy1981]

There are efficiency concerns with copying the full tracer arrays in and out of MARBL un-necessarily.

[mnlevy1981]

Idea would be that preformed tracers have no source / sink term, and every timestep the surface value is reset to match the non-preformed version of the same tracer.

[mnlevy1981]

@matt-long and @kristenkrumhardt -- I think this will be a fairly trivial feature to implement. It will require a marbl_preformed_init_mod.F90 module to set the tracer metadata, but since we set surface_fluxes = 0 and interior_tendencies = 0 at the beginning of surface_flux_compute() and interior_tendency_compute(), respectively, I don't think we would need marbl_preformed_surface_flux_mod.F90 or marbl_preformed_interior_tendency_mod.F90.

Also, it doesn't seem like we would need any preformed-specific diagnostics because we would just rely on the GCM's tracer diags. So we need a place in the code to reset the preformed surface values, but that's it. (There will also be some infrastructure work, but I think that's mostly introducing a preformed_on setting with a default of .false. and updating marbl_tracer_index_type)

Anyway, I think a day or two would be plenty of time to get it up and running, with maybe another half-day to run sanity checks and look at output. What's the timeline for this experiment? I definitely want to wait until we bring in #330 (later this week if all goes well) but I can spend a few days working on this before finishing #156.

[mnlevy1981]

Oh, I forgot to ask the big question: which tracers will have preformed counterparts? All the tracers are listed in the indexing type:

  type, public :: marbl_tracer_index_type
    ! Book-keeping (tracer count and index ranges)
    integer (int_kind) :: total_cnt = 0
    type (marbl_tracer_count_type) :: ecosys_base
    type (marbl_tracer_count_type) :: ciso

    ! General tracers
    integer (int_kind) :: po4_ind         = 0 ! dissolved inorganic phosphate
    integer (int_kind) :: no3_ind         = 0 ! dissolved inorganic nitrate
    integer (int_kind) :: sio3_ind        = 0 ! dissolved inorganic silicate
    integer (int_kind) :: nh4_ind         = 0 ! dissolved ammonia
    integer (int_kind) :: fe_ind          = 0 ! dissolved inorganic iron
    integer (int_kind) :: lig_ind         = 0 ! Fe-binding Ligand
    integer (int_kind) :: o2_ind          = 0 ! dissolved oxygen
    integer (int_kind) :: dic_ind         = 0 ! dissolved inorganic carbon
    integer (int_kind) :: dic_alt_co2_ind = 0 ! dissolved inorganic carbon with alternative CO2
    integer (int_kind) :: alk_ind         = 0 ! alkalinity
    integer (int_kind) :: alk_alt_co2_ind = 0 ! alkalinity with alternative CO2
    integer (int_kind) :: doc_ind         = 0 ! dissolved organic carbon
    integer (int_kind) :: don_ind         = 0 ! dissolved organic nitrogen
    integer (int_kind) :: dop_ind         = 0 ! dissolved organic phosphorus
    integer (int_kind) :: dopr_ind        = 0 ! refractory DOP
    integer (int_kind) :: donr_ind        = 0 ! refractory DON
    integer (int_kind) :: docr_ind        = 0 ! refractory DOC

    ! CISO tracers
    integer (int_kind) :: di13c_ind       = 0 ! dissolved inorganic carbon 13
    integer (int_kind) :: do13ctot_ind    = 0 ! dissolved organic carbon 13 (semi-labile+refractory)
    integer (int_kind) :: di14c_ind       = 0 ! dissolved inorganic carbon 14
    integer (int_kind) :: do14ctot_ind    = 0 ! dissolved organic carbon 14 (semi-labile+refractory)

    ! Living tracers
    type(marbl_living_tracer_index_type), allocatable :: auto_inds(:)
    type(marbl_living_tracer_index_type), allocatable :: zoo_inds(:)
    ! For CISO, don't want individual C13 and C14 tracers for each zooplankton
    ! Instead we collect them into one tracer for each isotope, regardless of
    ! zooplankton_cnt
    integer (int_kind) :: zootot13C_ind   = 0 ! total zooplankton carbon 13
    integer (int_kind) :: zootot14C_ind   = 0 ! total zooplankton carbon 14

  contains
    procedure, public :: add_tracer_index
    procedure, public :: construct => tracer_index_constructor
    procedure, public :: destruct => tracer_index_destructor
  end type marbl_tracer_index_type

Would the general tracers all have preformed copies as well? And would we do the same for the isotopic tracers if a user runs with both ciso_on = .true. and preformed_on = .true?

[mnlevy1981]

@klindsay28 pointed out that POP already has a reset function for passive tracers that we want to take advantage of, so instead of shoehorning the reset into interior_tendency_compute(), we really need a new marbl_instance%reset() stage. (That will require updating the documentation and the stand-alone driver.)

Other design details that came up:

• a new derived setting named reset_required (that at the moment will be equal to preformed_on, but if we add new_tracer_module_on that also makes use of the reset stage then reset_required would reflect that)
• marbl_instance%reset() will return immediately if reset_required is .false.
• POP should use get_settings('reset_required') to determine if it should even call marbl_instance%reset() (and avoid the overhead of copying tracer data if it's not needed)