Skip to content

Latest commit

 

History

History
230 lines (189 loc) · 14.1 KB

RELEASE.md

File metadata and controls

230 lines (189 loc) · 14.1 KB

RELEASE NOTES for FV3 202411: Summary

FV3-202411-public --- November 2024
Primary Point of Contact: Lucas Harris, GFDL [email protected]

This version has been tested with:
SHiELD physics release FV3-202411-public from https://github.com/NOAA-GFDL/SHiELD_physics
FMS release 2024.03 from https://github.com/NOAA-GFDL/FMS
FMS Coupler release 2024.03.01 from https://github.com/NOAA-GFDL/FMScoupler
Atmos Drivers release FV3-202411-public from https://github.com/NOAA-GFDL/atmos_drivers

This release includes the following:

  • Numerics updates (Lucas, Joseph, Linjiong):
    • Removed USE_COND and MOIST_CAPPA compiler directives and replaced with runtime options use_cond and moist_kappa in the new namelist fv_thermo_nml. Both default to .T. in nonhydrostatic simulation and .F. in hydrostatic.
    • Added a simple limiter to prevent dissipative heating from creating spurious cooling. Set prevent_diss_cooling = .F. to turn off.
    • Fixed bugs in hydrostatic nesting for west and east BCs in setup_pt_BC_k() and calculated true pressure in BCs if no BC remapping is done (compute_peBC, compute_peBC_k).
    • Revision of Lagrangian_to_Eulerian to fix variable dimension mismatch.
    • Revision of FV3's dissipation heating diss_est option to improve numerical consistency with other dissipation options.
    • Fixed edge noise for hord 6 and 7 (suggested by Bill Putman, GMAO).
    • Add mixed precision compilation mode to support 32bit FV3 with other 64bit components (with Uriel Ramirez).
    • New tracers:
      • w_diff to allow subgrid mixing of vertical velocity by physics. This requires compiling with the option -DW_DIFF to enable.
      • pbl_age and tro_pbl_age tracers representing the age of air since leaving the PBL and tropical PBL, respectively.
      • Removed obsolete clock tracers
      • Refer to docs/HOWTO_tracer-2024.11.md for more information
  • GFDL Microphysics updates (Linjiong)
    • Included fast microphysics tendencies diagnostics
    • Added two namelist options (fast_fr_mlt and fast_dep_sub) to control freezing/melting and deposition/sublimation in the fast microphysics.
    • Included a missing term in the energy conservation formula (credit: Tristan Abbott). May affect prediction of processes depending strongly on microphysics. Compile the model with -DENG_CNV_OLD to revert this change.
    • Added a namelist option, prog_cin, to define the source of CIN (cloud ice nuclei) concentration. This is similar to prog_ccn but for ice nuclei.
    • Added diagnostics for cloud content and cloud effective radii of all cloud hydrometeors (qc*, re*).
    • Added diagnostics for microphysical process rates (mpp*).
    • Removed unused Keihl et al. (1994) cloud water effective radius diagnosis
  • Driver update (Joseph):
    • Implemented a new atmosphere driver to run SHiELD and SHiEMOM with the full FMScoupler.
  • Updates to $2\delta z$ filter (fv_sg) (Lucas, Linjiong):
    • Included a missing term in the energy conservation formula (credit: Tristan Abbott). May affect prediction of processes depending strongly on microphysics. Compile the model with -DENG_CNV_OLD to revert this change.
    • Added option, fv_sg_adj_weak, to apply a weaker 2dz filter below sg_cutoff. This may be useful in controlling tropospheric instabilities without interfering with the behavior of the PBL scheme.
    • Renamed routines and eliminated ifdefs for SHiELD vs. AM4 versions.
  • Physics interface updates (Linjiong, Kai, Spencer):
    • Fixed negative tracers in the dynamics-physics interface.
    • Enhanced the fill_gfs function to remove negative tracers.
    • Enabled data_override for nest domain
    • Fixed a precipitation diagnostic issue when ntimes > 1 in the GFDL MP.
    • MPI fix for sedimentation mass transport in GFDL MP.
  • Updates to nudging (Lucas):
    • Added an option to turn TC breeding off.
    • Bugfixes for nudging on a nest/regional domain (in which tendencies in the halo are undefined).
  • Coarse-graining updates (Spencer, Kai):
    • Added options strategy = 'pressure_level_extrapolate' ’blended_area_weighted’ (developed with support from Chris Bretherton, AI2), and simplest model_level_area_weighted (like FREgrid first-order conservative scheme).
    • Renamed model_level strategy to model_level_mass_weighted.
    • Coarse-grained plev diagnostics for u, v, w, omega, vorticity, height, temperature, tracers, and RH.
    • Coarse-grained plev diagnostics use plevs defined in coarse-grained plev diagnostics for fv_diag_plevs_nml
    • OpenMP multi-threaded calculations
  • Code refactors (Lucas):
    • Cleaned up external_ic_nml and fv_surf_map_nml.
    • Cleaned up fv_mapz.F90 to move vertical remapping operators and thermodynamics/energetics routines into their own modules
  • Diagnostics (Lucas, Linjiong, Kai, Spencer):
    • Fixes for nudging and fv_sg diagnostics
    • Cleaned up fv_diagnostics stdout messages
    • True instantaneous and timestep-mean divergence and dissipative heating.
    • 40 dBz reflectivity height diagnostic.
    • Dissipative heating and dissipation estimate as, even if stochastic physics isn't enabled.
    • Introduced a flag PRT_LEVEL (now hard-coded) to control which min/max fields are written to stdout.
    • Fixed a bug for CAPE/CIN/BRN when nonhydrostatic pressure perturbation is also being output.
    • Refactor of plev and standard pressure level diagnostics, added new variables (vort, theta, theta_e, w, RH, dew point) to plevs, and removed unnecessary arguments to cs3_interpolator
  • Deprecated/removed options (Lucas):
    • Removed outdated options: scale_z, w_max, w_limiter, z_min, d2_divg_max_k[12], damp_k_k[12], old_divg_damp, do_am4_remap, use_new_ncep, use_ncep_phy, a2b_ord, c2l_ord.
    • Interpolation from cell-means to corner values (a2b) and from local staggered winds to A-grid lat-lon winds, have been hard-coded to be fourth-order, except where it had previously been hard-coded to be second-order. Supporting codes have been cleaned up.
    • Deprecation notice for conserve_ke
    • Added warning messages for poorly-chosen advection scheme options (hord_xx), and a FATAL is thrown for invalid scheme choices.

RELEASE NOTES for FV3 202305: Summary

FV3-202305-public --- May 2023
Lucas Harris, GFDL [email protected]

This version has been tested with SHiELD physics release 202305 and with FMS release 2023.01 from https://github.com/NOAA-GFDL/FMS

  • Revised Vertical Remapping Operators (Lucas)
    • kord=10 reverted back to AM4 version.
    • Post-AM4 version of kord=10 is now kord=12.
    • do_am4_remap no longer does anything and is deprecated.
    • New strictly-monotone operators kord=14, 15 for improving tracer correlations, and kord=13 without subgrid limiting.
    • kord <= 7 now deprecated; may be removed in a future release.
  • New Test Cases: (Joseph, Kun, Lucas)
    • Idealized TC test case with SHiELD physics
    • Zurita-Gotor et al. 2022 Held-Suarez variant
    • New Stable Boundary Layer (Beale at al.) doubly-periodic test case
  • New nesting updates: (Joseph)
    • Enable nesting in solo core and add a new idealized test case (58)
    • Enable adding multiple nests in doubly-periodic test cases using absolute coordinates
  • Additional idealized capability (Linjiong, Kun, Lucas)
    • Added namelist variable is_ideal_case, which must be used for runs starting (or re-starting) from idealized states.
    • Begin saving the initial wind fields (u0 and v0) to the restart files
  • GFDL MP and Integrated Physics (Linjiong):
    • Added options to sub-cycling condensation evaporation (nconds), control timescale or evaporation (do_evap_timescale), and delay condensation and evaporation (delay_cond_evap)
    • Removed unused 3d microphysics diagnostics to save time and memory
    • Optimized the mpp domain updates for fast physics
    • Update gfdl_mp_nml reading code to avoid model crash for absent gfdl_mp_nml
    • Added an option (do_intermediate_phys) to disable intermediate phys
    • Removed grid size in GFDL MP energy and mass calculation
    • Updates to use dry_cp instead of moist_cp in a hydrostatic case
  • Added a function to use O3 data from IFS ICs (Jan-Huey)
    • Namelist parameter: “use_gfsO3” with the default value = “false”
    • This function only works when ecmwf_ic = T
    • If the IFS IC does not include O3 data, or the run would like to use GFS O3 with other IFS ICs, set use_gfsO3 = T
  • Solver Updates (Lucas)
    • Revised semi-implicit solver to partially linearize vertical sound wave propagation about the hydrostatic state. This removes a specific instability causing deep “columnar” modes in the vertical velocity field due to the equation for the pressure perturbation being updated partially forward-in-time. This removes the spurious modes, reduces vertical velocities, and makes the solver slightly more stable.
    • MPI bug fix for tracer diffusion
    • Fast Rayleigh Damping on w controlled by fast_tau_w_sec.

RELEASE NOTES for FV3 202210: Summary

FV3-202210-public --- October 2022
Lucas Harris, GFDL [email protected]

This version has been tested with SHiELD physics release 202210 and with FMS release 2022.03 from https://github.com/NOAA-GFDL/FMS

This release includes the following:

  • Release of the GFDL Microphysics Version 3
  • Fix pressure-coarse-graining weighting from AI2's fork of FV3GFS
  • Add A-grid restart functionality from AI2's fork of FV3GFS
  • Fix for telescoping nest and GFS FIX file read
  • Total precipitation diag field has changed from prec to pret
  • Clean-up of the diagnostic messages to stdout

RELEASE NOTES for FV3 202204: Summary

FV3-202204-public --- April 2022
Lucas Harris, GFDL [email protected]

This version has been tested against the current SHiELD physics and with FMS release 2022.01 from https://github.com/NOAA-GFDL/FMS

This release includes the following:

  • Release of stand-alone solo_core functionality with simple physics.
  • Updated GFDL Microphysics, used for real-time 2021 C-SHiELD and T-SHiELD. (L Zhou)
  • Merges numerous updates from dev/emc.
  • Leverage DA functionality from UFS with additional changes (M Tong).
  • Updates to use the latest FMS release, including fms2_io.
  • Adds license header to missing files and fixes typo in header.
  • Fixes a bug where long_name and units attributes were not being captured in restart files.
  • Adds the ability to specify prefix and directory when reading and writing restarts.
  • The planetary radius and rotation rate are now re-scalable by a namelist parameter (small_earth_scale) instead of using exclusively the hard-coded FMS constant.
  • Removes obsolete driver/SHiELD files.
  • Removes unused function fv_diagnostics::max_vorticity_hy1.
  • Removes avec timer remnants.
  • Removes old style namelist read in favor of read from internal character variable.
  • Adds option for a mean wind.
  • Addresses GNU warnings.

RELEASE NOTES for FV3 202107: Summary

FV3-202107-public --- 08 July 2021
Lucas Harris, GFDL [email protected]

This version has been tested against the current SHiELD physics and with FMS release 2021.02 from https://github.com/NOAA-GFDL/FMS

This release includes the following:

  • Comprehensive documentation in LaTEX format (FV3 team)
  • Default changes to some namelist options and updated inline documentation
  • Multiple same-level and telescoping nests for the Regional domain (J Mouallem)
  • Updated fms2_io functionality (L Chilutti)
  • Revised Regional domain code (K-Y Cheng)
  • Reproducibility fixes for global+nests and regional+nests (tested for absolute reproducibility across PE counts, restarts)
  • Other updates and general cleanup

RELEASE NOTES for FV3 202101: Summary

FV3-202101-public --- 22 January 2021
Lucas Harris, GFDL [email protected]

This version has been tested against the current SHiELD (formerly fvGFS) physics and with FMS release candidate 2020.04 from https://github.com/NOAA-GFDL/FMS

This release includes the following:

  • Positive-definite advection scheme
  • In-line GFDL Microphysics
  • Fast-timescale Rayleigh damping
  • Updated namelist documentation
  • Implemented multiple same-level and telescoping nests for the global system (J Mouallem)
  • Updated coarse-graining capabilities (S Clark)
  • Re-organized fv_diagnostics, moving the revised fv_diag_column functionality and the declaration of diagnostic IDs to separate files
  • and other updates and general cleanup

This version of FV3 is described as component of SHiELD in Harris et al. (2020, JAMES).

Interface changes in 202101

atmosphere.F90: if using the in-line GFDL microphysics the precipitation rates (available in the structure Atm%inline_mp for rain, ice, snow, and graupel separately) must be passed into the physics and/or land model as appropriate. Here we demonstrate how to do this in SHiELD by copying them into IPD_Data(nb)%Statein%prep (and so on), which are newly defined in the IPD_Data structure within the SHiELD physics.

RELEASE NOTES for FV3 201912: Summary

FV3-201912-public --- 10 January 2020 Lucas Harris, GFDL [email protected]

This version has been tested against the current SHiELD (formerly fvGFS) physics and with FMS release candidate 2020.02 from https://github.com/NOAA-GFDL/FMS

Includes all of the features of the GFDL Release to EMC, as well as:

  • Updated 2017 GFDL Microphysics (S-J Lin and L Zhou included in GFSv15)
  • Updates for GFSv15 ICs (T Black/J Abeles, EMC)
  • Updates to support new nesting capabilities in FMS (Z Liang)
  • Re-written grid nesting code for efficiency and parallelization
  • Re-organized fv_eta for improved vertical level selection
  • 2018 Stand-alone regional capabilities (T Black/J Abeles, EMC)
  • Refactored model front-end (fv_control, fv_restart)
  • Support for point soundings
  • And other updates

Interface changes

drivers: renamed 'fvGFS' directory to SHiELD

atmosphere.F90: 'mytile' is renamed 'mygrid'

The non-functional gfdl_cloud_microphys.F90 has been removed and replaced with the 2017 public release given to EMC. Also added a proper initialization routine, that includes the use of INTERNAL_FILE_NML and thereby requires the input_nml_file argument. If you do not define the compiler flag INTERNAL_FILE_NML then you can delete this argument.

The namelist nggps_diag_nml has been eliminated. 'fdiag' is no longer handled by the dynamical core, and should be handled by the physics driver.

For a complete technical description see the NOAA Technical Memorandum OAR GFDL: https://repository.library.noaa.gov/view/noaa/23432