rdb_ocean_meke Module

Mesoscale eddy kinetic energy state. All fields default to the inert (enable=.false.) configuration so an ocean run that never sets &ocean_meke_nml is bit-identical.

Components

Type-Bound Procedures

Harmonic inverse mixing length 1/Lmix = Sum aX/LX over the five length scales (deformation, frictional, Rhines, Eady, grid). Each scale is gated aX*LX > 0 so a zero weight or a degenerate scale contributes nothing. Returns 1/Lmix (0 ⇒ Lmix degenerate).

Arguments

Return Value real(kind=wp)

Counted allocatable footprint of the MEKE slot (0 when unallocated). One arr_bytes term per array — add a term here when a new allocatable joins the type.

Arguments

Return Value integer(kind=int64)

Inject the MEKE harmonic backscatter into the per-face resolved harmonic viscosity (capability Gap 2, v1). Subtracts a face-average of the cell-centred ku field from ah_face_x/ah_face_y so the NET coefficient A_net = A_resolved − Ku can go NEGATIVE — that negative viscosity is the energy return into the momentum tendency (the hvisc Laplacian kernel reads these same face fields).

Arguments

Advance the MEKE field one thermo step (Strang split), update the derived diffusivity kh_diff, and feed the geometric-mean kh into VarMix’s per-face KhTh/KhTr (the GM↔MEKE feedback). Run once per outer step at thermo cadence, after varmix_compute and before gm_compute_transports (MEKE reads gm%gm_src from the previous thermo step — a one-step lag). No-op when enable=.false., uninitialised, or the GM slot is absent.

Arguments

Upwind flux-form advection of E by the barotropic mass transport. advFac = adv_fac/sdt uflux(I) = baroHu(I)advFacE_upwind (E_{I-1} if baroHu>0 else E_I) E += sdtIareaTI_mass((uflux_{i-1}-uflux_i)+(vflux_{j-1}-vflux_j)) Conservative on a closed domain (interior faces only; the divergence telescopes ⇒ Sum Earea*mass conserved). adv_fac=0 never reaches here (gated by the caller) ⇒ default bit-identity.

Arguments

harmonic viscosity and floor the net at the CFL-stable minimum. Explicit-shape dummies for NVHPC stdpar (no descriptor walk). The backscatter coefficient is z-independent in v1 (BS_struct=1), so the cell-centred ku(i,j) is broadcast to every layer.

Arguments

Depth-integrated, mass-weighted barotropic transport through each C-grid face: baroHu(I,j) = Sum_k rho_face * mass_flux_x_layer, where mass_flux_*_layer is the per-layer VOLUME transport (m^3/s, = uh_facedy) and rho_face the two-cell average density. The result is a MASS transport (kg/s) so the advective divergence pairs exactly with IareaT*I_mass (1/(areamass)) ⇒ Sum Earea*mass is conserved. Array-edge faces carry zero transport (closed domain).

Arguments

Resolved bed-layer speed² at cell centres: u_bbl2 = u_c² + v_c² with u_c = ½(u_face(i)+u_face(i+1)), v_c = ½(v_face(j)+v_face(j+1)), each face read on ITS OWN bed layer k_bot_u/v (the first layer live on both sides counting up; 1 off z_fixed ⇒ the historical k=1 read). Under z_fixed the layers below are inert fillers whose velocity the closed-face mask has zeroed, so a k = 1 read reported a motionless bed. The bottom eddy velocity the MEKE drag law needs (MOM6 drag_rate_visc).

Arguments

Implicit (backward-Euler) bottom-drag half-step. drag_rate = rho0i_masssqrt(cdrag^2(max(0,2bf2E)+u_bbl2+uscale^2)) [1/s] damp_rate = damping + drag_ratebf2 ; =0 where E<0 E <- E/(1 + sdt_dampdamp_rate) rho0*i_mass = rho0/(Sum_k rho_kh_k) ~= 1/depth_tot [1/m], so drag_rate ~= cdrag|U_d|/H – the MOM6 GV%H_to_RZ * I_mass factor. Without it drag_rate is m^3/(kgs), not a rate. i_mass=0 on dry columns still gives drag_rate=0. u_bbl2 is the resolved bed-layer speed² (MOM6 drag_rate_visc); it is 0 unless use_bbl_drag is set, so the default is bit-identical.

Arguments

Add the geometric mean of neighbour kh_diff into VarMix’s per-face KhTh (and KhTr) base BEFORE GM’s CFL clamp: khth_u(i,j) += khth_facsqrt(kh(i-1,j)kh(i,j)) khth_fac=0 ⇒ nothing added ⇒ bit-identical seam.

Arguments

Derived diffusivity kh = khcoeff*sqrt(2*max(0,gamma_t2*E))*Lmix. khcoeff<=0 ⇒ kh left at 0.

Arguments

Derived harmonic backscatter viscosity ku = visc_coeff_ku*sqrt(2*max(0,E))*Lmix (m²/s), matching MOM6 MEKE%Ku = MEKE_VISCOSITY_COEFF_KU*sqrt(2*MEKE)*Lmix. Unlike the kh closure (which carries the barotropic-mode factor gamma_t2 inside the eddy velocity), MOM6’s Ku uses the PLAIN sqrt(2*MEKE) — no vertical-structure factor — so gamma_t2 is deliberately absent here (vertical structure BS_struct = 1; EBT/SQG deferred). Off ⇒ ku left at 0 (bit-identical seam). Always ≥ 0; the SIGN of the momentum effect is set by the subtraction downstream.

Arguments

Harmonic-mass Laplacian diffusion of MEKE (+ optional biharmonic). Flux-form, conservative on a closed domain (interior faces only; array-edge faces carry zero flux). Kh_u = max(0,kh_bg) + khmeke_fac0.5(kh_i+kh_{i+1}), CFL-capped 0.25 uflux = Kh_u(dy_cuidxCu)[2 m_i m_{i+1}/(m_i+m_{i+1}+eps)](E_i-E_{i+1}) E += sdtiareaTi_mass((uflux_{i-1}-uflux_i)+(vflux_{j-1}-vflux_j)) Biharmonic: del2 = iareaT(d uflux’ + d vflux’) with the bare-gradient flux uflux’ = (dy_cuidxCu)(E_{i+1}-E_i); then a harmonic-mass flux of del2 with CFL cap 0.3 and E += that divergence (additive).

Arguments

Fill the structure factors gamma_b^2 (bottom_fac2) and gamma_t^2 (barotr_fac2) plus the mixing length le (Lmix) at each cell centre. Ldeform/Lfrict drives both gammas; Lmix is the harmonic sum of the alpha-weighted scales. beta = |grad f| from centred f_centre differences scaled by idxT/idyT (zero when f_centre is unfilled ⇒ Rhines inert). SN = 0.25*(sn_u(i)+sn_u(i-1)+ sn_v(j)+sn_v(j-1)) only when aEady>0.

Arguments

Column mass mass = Sum_k rho*max(h,H_VANISHED) (kg/m^2), its inverse i_mass (0 where mass<=0), depth_tot = Sum_k h (m), and mass_ws = mass (the harmonic-mass input for the lateral flux).

Arguments

Aggregate source src = bgsrc + gmcoeff*I_mass*gm_src - frcoeff*I_mass*ke_diss and the explicit bump E += sdt*src. gmcoeff<0 ⇒ GM source off; frcoeff<0 ⇒ frictional source off. ke_diss is the lateral-viscosity KE dissipation rate (≤0), so -frcoeff*I_mass*ke_diss ≥ 0 is a mean→eddy source (0 ⇒ inert).

Arguments

Copy rd_over_dx onto the device workspace. Thermo-cadence; explicit-shape; on-device (source slot is device-resident).

Arguments

Copy the SN faces onto the device workspaces. Thermo-cadence; explicit-shape; on-device (VarMix slot is device-resident).

Arguments

Zero a device-resident 2D workspace (absent-source fallback).

Arguments

Arguments

Arguments

Arguments

Arguments

Arguments

Allocate the prognostic field, the derived diffusivity, and the Strang-stage workspaces. Always allocates (configure runs after init); setup uses plain host allocation (no do concurrent before enter_data).

Arguments

Copy a pre-filled cell-centre Coriolis magnitude |f| (1/s) onto the MEKE slot, so beta = |grad f| for the Rhines length is live. The caller (setup) builds f_centre with the same metrics_fill_coriolis path the Coriolis / VarMix / EPBL slots use (handles beta-plane AND spherical). Host loop — call after init, before enter_data.

Arguments