Prepare for PR to dev/gfdl #6
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
nikizadehgfdl
pushed a commit
that referenced
this pull request
Aug 12, 2022
Add initialization tests using CS%initialized
marshallward
force-pushed
the
dev/gfdl
branch
3 times, most recently
from
8a66adc to
9de6ce7
Compare
Cleaned up the error messages when there are fatal problems in interpolate_for_nondim_position() in the MOM_neutral_diffusion module to explicitly give an indications of all the problems. These had previously triggered fatal errors (or should have), but had less explicit or incorrect error messages in some cases, so all solutions and behavior are identical in any cases that worked previously.
Corrected the case of the do-loop index variable in 10 places in 6 files, initialized two whole diagnostic arrays to 0 instead of having shared loops initializing u- and v-point arrays, and shortened two loops initializing pressures to the size that is actually used in the corresponding equation of state calls. With these changes, there are no i- or j- loops with extents that are inconsistent with the stagger-based index case convention used throughout the MOM6 code. All answers are bitwise identical in all cases.
Replaced 382 scale arguments with unscale arguments in calls to chksum routines, and did the same in another 5 calls to global_mean or global_integral routines. Also added 2 missing unscale arguments to checksums for the initial velocities, which probably had gone undetected because we usually initialize velocities to 0. All answers and diagnostic debugging output (apart from the afore noted checksums of the initial velocities in some rare case) are bitwise identical.
Added the new function query_debugging_checks to the MOM_debugging module to return flags indicating what types of debugging are enabled, including checksums or redundant value checks. This new routine is used to allow for the redundant value checks in various step_MOM routines to be enabled or disabled at run-time when debugging is on overall (by setting DEBUG = True). For cases with inconsistent redundant points, this change allows for a reduction in the volume of debugging information sent to stdout from multiple PEs (in an arbitrary order), while for cases without inconsistent redundant points this change will allow for faster debugging with checksums alone. All answers are bitwise identical, but there is a new publicly visible routine.
This commit restores the effectiveness of the runtime parameter USE_WRIGHT_2ND_DERIV_BUG in determining whether a bug is corrected in the calculation of two of the second derivative terms returned by calculate_density_second_derivs_elem() with the "WRIGHT" equation of state, recreating the behavior (and answers) that are currently on the main branch of MOM6. To do this, it adds and calls the new routine set_params_buggy_Wright() when appropriate, and adds the new element "three" to the buggy_Wright_EOS type. When the bug is fixed, buggy_Wright_EOS%three = 3, but ...%three = 2 to recreate the bug. This commit does change answers for cases using the "WRIGHT" equation of state and one of the "USE_STANLEY_..." parameterizations from those on the dev/gfdl branch of MOM6, but in so doing it restores the answers on the main branch of MOM6. There is also a new publicly visible subroutine.
Moved a check for whether to initialize an uninitialized visc%h_ML field from visc%MLD outside of the test for the value of CS%mixedlayer_restart. This change will prevent answer changes between code versions for certain cases that initialize from a restart file, use the melt_potential field and have MLE_USE_PBL_MLD = True but also have MIXEDLAYER_RESTRAT = False. This commit corrects a very specific oversight that was introduced when the roles of visc%h_ML and visc%MLD were separated, and it can change answers (reverting them to answers from older versions of the main branch of MOM6) in very specific cases where MOM6 is initialized from a restart file from an older versions of the code, including the 5x5-degree test case in the dev/emc regression suite.
Added checksum calls for the melt_potential, ocean_mass, ocean_heat and ocean_salt elements of the surface state in MOM_surface_chksum if these fields are allocated for more comprehensive debugging. Also added the symmetric optional argument to the call to MOM_surface_chksum form extract_surface_state so that all of the surface velocity values that could contribute to the ocean surface velocities that are seen by the coupler are checksummed. All solutions are bitwise identical, but there are enhancements to the MOM6 debugging capabilities.
The module MOM_harmonic_analysis computes the constant coefficients of sine/cosine functions for the SSH and barotropic velocity fields of each tidal constituent.
Performance optimization for subroutine HA_solver.
Optionally corrected a bug (essentially a sign error) in the selection of where MASS_WEIGHT_IN_PRESSURE_GRADIENT is applied in non-Boussinesq test cases. The (incorrect) non-Boussinesq version of the test for where interfaces are outside of the range of hydrostatic consistency due to interactions with bathymetry was converted from the (correct) Boussinesq version without properly taking into account that pressure increases downward while height increases upward. This bug is repeated 12 times in 6 different specific volume integral routines, including in the analytical specific volume integrals with 4 equations of state. To accommodate these corrections as well as a future expansion of the mass weighting to apply near the surface under ice-shelves or icebergs, the previous optional logical argument (useMassWghtInterp) was replaced with a new optional integer argument (MassWghtInterp) that can be used to encode information about where this weighting is applied as well as whether to fix this bug. This combination of settings (MASS_WEIGHT_IN_PRESSURE_GRADIENT = True and non-Boussinesq) does not seem to be widely used yet (it is not used in the GFDL regression suite), so rather than preserving the old (incorrect) solutions by default, this bug is corrected by default. However, the previous answers can be recovered by setting the new runtime parameter MASS_WEIGHT_IN_PGF_NONBOUS_BUG to true. This parameter that is only used (and logged) for non-Boussinesq cases with MASS_WEIGHT_IN_PRESSURE_GRADIENT set to true. It is intended that this new runtime bug-fix parameter will be obsoleted with an aggressive schedule if it is not needed to recreate any production runs. In addition, there are two new diagnostics, MassWt_u and MassWt_v, that show the fractional (0 to 1) effects of the mass weighting in the pressure gradient forces at the velocity points, along with the new diagnostic subroutines diagnose_mass_weight_Z and diagnose_mass_weight_p that calculate these diagnostics by layer in code that replicates the 13 copies for various equations of state and vertical structures within layers. By default, this commit does change answers in non-Boussinesq cases that use MASS_WEIGHT_IN_PRESSURE_GRADIENT = True, and there is a new runtime parameter in such cases. There are also two new diagnostics. Answers are bitwise identical in all Boussinesq cases.
* Correct bug in kappa shear viscosity with vertex shear option. - Viscosities at vertices along the coast were incorrectly zero'd out. This commit removes that mask so the non-zero shear driven viscosities can be interpolated from in the model. This bug caused diffusivities to be very large in channels and potentially near coastlines. - A bugfix flag is added with an option to use the current behavior for legacy purposes. * Fix missing paranthesis in previous commit (VS_viscosity_bug) * Update logging of vertex shear viscosity bug parameter
Add the new runtime parameter PHILLIPS_ANSWER_DATE to enable the option to use mathematically equivalent expressions in Phillips_initialize_velocity() that exactly specify the arithmetic to be used, avoid excess divisions and permit full rescaling of the internal variables and the elimination of rescaling variables. This new slightly answer-changing option is enabled by setting PHILLIPS_ANSWER_DATE >= 20250101. For now, the default for PHILLIPS_ANSWER_DATE is set to 20241231 to avoid changing answers without explicitly setting it. This commit also introduces code to use G%grid_unit_to_L to detect and handle various choices for the units of the G%geolat and G%geolon variables. By default, all answers are bitwise identical, but there is a new runtime parameter in some MOM_parameter_doc files. This commit changes answers at roundoff when there is an explicit setting of PHILLIPS_ANSWER_DATE >= 20250101,
Added the new zero_zeros optional argument to the 10 MOM_write_field routines in MOM_io and the 6 rescale_comp_data routines in MOM_domain_infra to cause negative zeros to replaced with ordinary signless zeros before they are written out to files. This has no impact at all on answers, but it does help with comparisons between rotated restart files, in which meaningless differences between positive and negative zeros were leading to false differences between files. All answers are bitwise identical, and all output is equivalent, but there are new optional arguments to 16 routines covered by 2 publicly visible interfaces.
Added a new turns optional argument to 6 versions of the MOM_read_data routines to allow for the reading to override the number of turns in the MOM_domain that is passed into these routines. Several internal turns variables in the same routines were renamed qturns to allow for the new optional arguments. Also check for whether the MOM_domain%domain_in pointer is associated before it is used, avoiding a segmentation fault that was occurring when a restart file is read and ROTATE_INDEX is true. Also added rotate_array calls to ensure that the halo values are retained while reading data into a rotated array. These changes are necessary to allow for the model to be initialized from a restart files with rotated grids. Several instances of continuation line indentation that do not follow the typical 4-space pattern used elsewhere in the MOM6 code and documented in the MOM6 style guide were also altered to follow the standard. All answers that previously worked are bitwise identical, but there are new optional arguments to publicly visible interfaces.
Modified MOM_restart so that restart files generated by rotated runs match unrotated restart files, and the model can be properly initialized from a restart file when the grid is rotated. Also added runtime options to convert negative zeros into ordinary zeros before they are written to restart files (selected with RESTART_UNSIGNED_ZEROS) and to properly do the checksums on the velocity points on all of the faces (selected with RESTART_SYMMETRIC_CHECKSUMS). Also added the new interfaces copy_restart_var and copy_restart vector to use the names of restart variables and the pointers stored in the restart control structure to obtain a copy of the variables as the restart variables with the option to undo the rotation. These routines are necessary because the reading of restart files occurs during a phase of the model initialization that works on an unrotated grid, and they are called from inside of MOM_initialize_state. The ranges for the checksums are now set correctly for each variable, depending on where it is discretized, but when RESTART_SYMMETRIC_CHECKSUMS is false, the previous ranges are still used so answers do not change in unrotated test case. The conversion factors used for the pair of register_restart_field calls in register_restart_pair now include the necessary sign changes for the rotation, as set in the new internal routine set_conversion_pair. There is also now a scalar_pair optional argument to the register_restart_pair routines to accommodate the rotation of pairs of scalars that do not change sign when rotated (e.g., grid-lengths). Instead of working with the hor_grid character string, the restart code has been modified to instead use the encoded integer position argument returned from query_vardesc. This avoids several redundant blocks of code that translate the hor_grid strings into positions. All answers are bitwise identical when there is no grid rotation, but with grid rotation the restart files that are created are modified to have the correct signs and replicate the restart fields with no rotation. Also, cases with grid rotation can now be reinitialized from restart files, while previously this simply did not work, either giving an incorrect reinitialized state or a segmentation fault. There are also two new runtime parameters in some MOM_parameter_doc files.
Added a description of the units of the return values to the comments describing 33 real functions in 8 modules. Only comments are changed and all answers are bitwise identical.
…810) * Add MASS_WEIGHT_IN_PGF_VANISHED_ONLY to modify mass weighting This commit introduces the runtime variable `MASS_WEIGHT_IN_PGF_VANISHED_ONLY` which has default False. If true, then the `MASS_WEIGHT_IN_PRESSURE_GRADIENT` and `MASS_WEIGHT_IN_PRESSURE_GRADIENT_TOP` effect of weighting T/S integrals in slanted grid cell FV PGF calculation is turned off if both sides of the grid cell are nonvanished, where nonvanished means thickness greater than `RESET_INTXPA_H_NONVANISHED` which defaults to 1e-6 m. Since the benefit of `MASS_WEIGHT_IN_PRESSURE_GRADIENT` happens in vanished layers (creating a fake PGF away from vanished layer, which is arrested by upwinded viscosity) the benefit is still there, but now we can use `MASS_WEIGHT_IN_PRESSURE_GRADIENT` for coordinates that also have slanted layers in the open ocean that are not vanished, e.g. sigma coordinates or SIGMA_SHELF_ZSTAR coordinates in the ice shelf where we DO trust T and S values. Additionally, this is required near a grounding line in a 3D z-coord ice shelf as some strange looking slanted non-vanished cells can emerge, and MWIPG being on would create fake PGFs in non-vanished cells (and therefore generating spurious currents). Reccommend `MASS_WEIGHT_IN_PGF_VANISHED_ONLY` to be set to True, as well as `MASS_WEIGHT_IN_PRESSURE_GRADIENT` and `MASS_WEIGHT_IN_PRESSURE_GRADIENT_TOP` if you have vanishing layers with min thickness < 0.1m. Also modifies MassWt_u and MassWt_v diagnostics to reflect usage of MASS_WEIGHT_IN_PRESSURE_GRADIENT. This commit should not change answers since it defaults to False. However, my implementation is not very efficient and should probably be optimised. * Minor style updates to previous commit of Add MASS_WEIGHT_IN_PGF_VANISHED_ONLY to modify mass weighting * Address comments from Bob by adding scaling to dimensional numbers, replacing Boussinesq rho in nonBoussinesq code, and removing white space to follow 2-space indenting.
This commit cleans up several aspects of the MOM_internal_tides code, including the elimination of one runtime option, the correction of some units in comments and a start toward enabling the propagation directions to alternate. 1. Eliminated the internal routine propagate_corner_spread. This routine was never completely implemented, and it has never been used in any realistic configurations. In particular, it appears that it will only work on a Cartesian grid in which the position variables (G%geoLonBu and G%geoLatBu) have the same units as G%dx. The comment describing propagate_corner_spread() indicates that "it is too numerically diffusive to be of much use as of yet" and that "it is not yet compatible with reflection schemes", so we have no plans to complete its implementation. Although it is being deleted, it can always be recovered from older branches of MOM6. 2. The runtime parameter INTERNAL_TIDE_CORNER_ADVECT, which triggered the use of propagate_corner_spread was obsoleted. 3. Started adding the option to alternate the direction of the internal tide propagation. The code that actually enacts the change is still commented out to avoid changing answers, but it could readily be enabled later with changes to a single line, and all the required code is present. 4. The units of 9 area-integrated energy budget debugging diagnostics were corrected, as is rescaling factor when their sum is written to stdout. With these changes, all answers are expected to be bitwise identical in any cases that are in active use, but an under-used runtime option has been eliminated from some MOM_parameter_doc files.
* Reorder optional input arguments in set_dtbt Change the order of optional input arguments of subroutine set_dtbt so that they are grouped logically. The decription of the subroutine is also updated. * Add runtime parameter SET_DTBT_USE_BT_CONT The runtime parameter allows to use BT_CONT optional input argument in subroutine set_dtbt if BT_CONT type is available. Originally, when BT_CONT is used, set_dtbt is estimated with zero surface height. The eta optinoal input argument has no effect since it only works with NONLINEAR_BT_CONTINUITY is true, which is by default false when BT_CONT is used. The added option allows accurate estimate of dtbt when the mean surface height of the domain is not at the same level, while maintains old answers. Also, an unused field calc_dtbt in type MOM_dyn_split_RK2_CS is removed. * Minor fixes on setting dtbt in barotropic_init * Fix a bug that local variable calc_dtbt is not initialized. * Rename dtbt_tmp to dtbt_restart for clarity * Add a comment about the usage of subroutine call set_dtbt * Remove an unused input argument eta in barotropic_init
This patch adds the MOM_murmur_hash module, a non-cryptographic hash function used to generate unique hashes for arrays. The particular property of interest is that murmur hashes are sensitive to order. This will allow us to detect a wider range of answer changes which would otherwise be undetected by the current min/max/mean/bitcount method. The checksum functions have also been modified to produce murmur hashes, although the compile-time flag to generate them is currently disabled.
Revised the calculation of the netFWGlobalAdj, vPrecGlobalAdj and saltFluxGlobalAdj elements of the forcing type to work in dimensionally rescaled units, making use of the unscale argument to reproducing_sum. This change includes the addition of conversion factors to the register_scalar_field calls, and changes to the units of 15 variables to be rescaled. A total of 27 instances of multiplication by unit conversion factors were eliminated as a part of these changes, with 15 other unit conversion factors added via unscale arguments. All answers and output are bitwise identical, but there are changes to the units of three variables in a transparent type.
MOM_state_stats() has been refactored to work primarily in dimensionally rescaled units using the unscale argument to reproducing_sum(). As a result of these changes, the units of 8 variables were changed to be rescaled, and six instances of multiplication by rescaling factors were changed into unscale arguments. MOM_state_stats() is exercised by setting DEBUG_CONSERVATION = True, and it has been verified that MOM_state_stats() gives results that are invariant to the use of dimensional rescaling. All answers and output are bitwise identical.
* Frequency-dependent internal wave drag Utilizing streaming band-pass filters, the frequency-dependent wave drag can be applied to the narrowband tidal velocities, instead of the broadband barotropic velocity. This allows internal wave drag to be optimized for different frequency bands independently, without affecting the low-frequency, non-tidal motions. Also fixed the bug that Filt_accum() does not return a valid output if it is called at the corrector step. * Frequency-dependent internal wave drag Minor performance optimization. Also, this commit allows the filters to be activated without the frequency-dependent drag being activated.
* Commit for coupling of land to ocean adot * Remove unused ice sheet enabled flag * add adot to Shelf restart * Register sfc mass flux outside of initialize fluxes * Gfdl cryo merge (#3) * Removed second registration of sfc_mass_flux with register_diag_field. * Registered sfc_mass_flux to runs with static and dynamic ice sheets. * This commit removed the second registration of 'sfc_mass_flux' in 'register_diag_field()' * Testing shelf_sfc_mass_flux restart. * The earlier removed 'register_diag_field()' for 'mass_flux' has been added back and 'register_restart_field()' of 'fluxes_in%shelf_sfc_mass_flux' is moved before 'restore_state()' * Corrected the line length error. * Added back 'CS%id_mass_flux = register_diag_field()' --------- Co-authored-by: NJSchlegel <nschlegel@gmail.com>
Expand range of halos for checksums for Kh_h, Kh_q, Ah_h and Ah_q to cover the full range of values that are actually used. This includes expanding the range of values where the viscosities are calculated to those used in symmetric memory mode when DEBUG is true, even though these are not used for efficiency. One set of vorticity-point loop ranges was also corrected when the recently added parameter EY24_EBT_BS is true to avoid setting an extra unused value, perhaps correcting non-symmetric memory mode cases that would not reproduce across PE count and layout. All answers are bitwise identical in any case that reproduces across PE count.
Corrected the indenting in most of the lines of ePBL_column to respect the MOM6 2-space indenting convention, and eliminated the comment explaining why the incorrect indenting had previously been retained to facilitate the review of a pull request with substantial actual changes in this routine. Apart from a single comment that was removed, only white space is changed, and all answers are bitwise identical.
* Add MLD_out to diagnose MLD call Add an optional argument to pass the MLD out of the diagnose MLD routine. * Change MLD_out to be only computaional domain Includes three changes: - Make MLD_out optional - initalize MLD_out to 0. everywhere - copy MLD to MLD_out only in the computational domain. * Remove unnedded MLD=0 initalization Remove an extra initalization of MLD in diagnoseMLDbyEnergy --------- Co-authored-by: Theresa Morrison <Theresa.Morrison@gaea68.ncrc.gov> Co-authored-by: Theresa Morrison <Theresa.Morrison@gaea63.ncrc.gov>
Revised the rescaled units of forces%tau_mag, fluxes%tau_mag and fluxes%tau_mag_gustless from [R Z L T-2 ~> Pa] to [R Z2 T-2 ~> Pa] to avoid the need for further rescaling when this field is used to calculate turbulent friction velocities and TKE fluxes in 29 places. However, this requires the addition of other rescaling factors when forces%tau_mag is set from the wind stresses. A total of 40 rescaling factors were eliminated, while another 36 were added, all where the components of the wind stress are used to calculate the magnitude of the wind stress. Several other internal variables were also rescaled analogously for simplicity. All answers are bitwise identical, but there are changes to the rescaling factors for three elements in a transparent type.
The field is never used by MOM_set_visc and incorrectly logged under MOM_set_visc section in MOM_parameter_doc, rather than MOM_kappa_shear.
* In subroutine set_viscous_BBL, move and merge `u2_bg` calculation to avoid redundancy and if-branch within do-loops. * In subroutine set_viscous_BBL, add a comment on explaining the reason to reset Ray_[uv] * In subroutine find_coupling_coef, replace an if-branch with min()
Added the new publicly visible interface field_checksum to return the checksum that is used to verify the arrays that are written to or read from files, while handling both the unscaling and rotation of the arrays. The 5 internal rotated_field_chksum_real_... functions were renamed to field_checksum_real_... to reflect their slightly broadened purpose and now have new optional unscale arguments. The previous rotated_field_chksum interface has been retained because it is still being used in SIS2, but this may change in the future. All answers are bitwise identical, but there is a new public function interface and a new optional argument to a preexisting one.
Use field_checksum in get_depth_list_checksum to relocate two rescaling factors into unscale arguments and work in scaled internal variables in that routine. All answers are bitwise identical.
Use field_checksum and its unscale argument in save_restart and restore_state to avoid 10 instances of array-syntax math which are unnecessary in most cases because the conversion factor is 1. All answers are bitwise identical.
* Remove unused local variable I_Rho in int_density I_Rho = 1.0/rho_0 is never used and therefore removed. * Add runtime flag RHO_PGF_REF_BUG This new parameter addresses a bug in Boussinesq finite volume pressure gradient forces (MOM_PressureForce_FV) that the mean seawater density Rho_0 and reference density Rho_ref are used incorrectly in several instances. It should be noted that by default Rho_ref is Rho_0 and Rho_ref is rarely set to other than Rho_0. * Bugfix: reference height offset for SAL using bpa Recover reference height offset (Z_ref) in calculating bottom pressure for self-attraction and loading in Boussinesq mode.
* Alternative interface to EQN_OF_STATE="LINEAR" The existing interface to EQN_OF_STATE="LINEAR" is based on RHO_T0_S0, the density at T=0, S=0. This is the most computationally efficient way to specify a linear EOS, but we are usually linearizing about a reference T and S that are not 0. The new interface is based on TREF, SREF, and RHO_TREF_SREF, where: RHO(T,S) = RHO_TREF_SREF + DRHO_DT*(T-TREF) + DRHO_DS*(S-SREF) RHO_T0_S0 = RHO_TREF_SREF - DRHO_DT*TREF - DRHO_DS*SREF RHO(T,S) = RHO_T0_S0 + DRHO_DT*T + DRHO_DS*S The defaults for TREF and SREF are zero and for RHO_TREF_SREF is 1000.0. So if the new interface is not used answers are not changed but there are always new model parameters. * corrected spelling error
* Adding a flag to turn off the bug fix (mostly) * Fix OBC check if not associated.
* Split grid generation and remapping from step thermo Move grid generation, remapping, and subsequent processes into a new routine ALE_gridgen_and_remapping which is called immediately after step_MOM_thermo. Clean up inputs and local variables in step_MOM_thermo by removing those only used for gid generation and remapping. Routines that need to be done after step_thermo, grid generation and remapping, regardless of use_ALE or not into another routine since they do not make sense in ALE_gridgen_and_remapping. Answers and diagnostics have not changed in Baltic case in ALE mode. Testing still needs to be done layer mode and with particles. * Update Names of routines * Fix doxygen and calltree errors --------- Co-authored-by: Theresa Morrison <Theresa.Morrison@gaea68.ncrc.gov> Co-authored-by: Theresa Morrison <Theresa.Morrison@gaea57.ncrc.gov>
* + Add `RESET_INTXPA_INTEGRAL_FLATTEST` option to pressure gradient This new parameter RESET_INTXPA_INTEGRAL_FLATTEST modifies the reference interface for the horizontal pressure integrals in the finite volume pressure gradient force calculation. If true, in ocean columns where no non-tilted, non-vanished layer can be found (e.g. near specific shaped ice shelf grounding lines), rather than using the top surface to integrate downwards as the reference, the code finds and chooses the flattest (i.e. minimum difference between depth or pressure) interface as the reference. The parameter defaults to False therefore should not change default answers. However, if true, combined with other ice shelf code changes e.g. MASS_WEIGHT_IN_PGF_VANISHED_ONLY = True, RESET_INTXPA_INTEGRAL = True, MASS_WEIGHT_IN_PRESSURE_GRADIENT = True and MASS_WEIGHT_IN_PRESSURE_GRADIENT_TOP = True + initialisation code fixes I get very low spurious currents in the Boussinesq ZSTAR and SIGMA_SHELF_ZSTAR ISOMIP+ test cases. * Amend RESET_INTXPA_INTEGRAL_FLATTEST description to refer to flattest interface rather than flattest layer.
* Update btstep negative eta warning In Boussinesq mode the eta below bathyT warning includes the location, but in non-Boussinesq mode the corresponding negative eta warning does not. Added the location to the negative eta warning. No answers are changed. * 4 space indenting of continuation lines
This commit takes the first steps toward correcting problems with the Kelvin initialization code, including making the OBC nudging timescale an explicit parameter and adding a number of comments identifying concerns and suggesting possible improvements with the code that is used when KELVIN_WAVE_MODE > 0. This commit includes the introduction of the new runtime parameter KELVIN_WAVE_VEL_NUDGING_TIMESCALE, with a default value set to recover the current hard-coded answers. A new warning message cautions about the suspected problems with the internal wave version of the code. By default, all answers are bitwise identical in the 4 Kelvin wave test cases that can be found at github.com/ESMG/ESMG-configs, but there is a new parameter in the MOM_parameter_doc.all files these cases.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.