Main candidate from NCAR (12/15/2020) #1275

gustavo-marques · 2020-12-16T15:02:44Z

Major changes

Make long character variables deferred length (Make long character variables deferred length NCAR/MOM6#165)
Major updates to the lateral boundary diffusion module (Major updates to the LBD module NCAR/MOM6#164)
Introduce a controllable stdout unit in MOM_io (Introduce a controllable stdout unit in MOM_io NCAR/MOM6#163)

New diagnostics

Adds option to output u^2, v^2 and u*v (Adding option to output u^2, v^2 and u*v NCAR/MOM6#159)

This candidate does not changes answers for the MOM6 regression tests (Intel 19.0.5).

* turned off pressure force * hard-coded BLD * turned off advect_tracer * set NTR = 2

This commit adds a linear transition from full LBD at k=k_min to zero LBD at k=k_max. This is applied to both methods currently available in the LBD module. Another modification is the fact that both methods no longer compute average values at k_min (done previously via average_value_ppoly). Instead, the full layer thicknesses are now used.

This patch adds the option to apply a linear decay of the fluxes at the base of hbl. This had been already implemented but since it breaks the unit tests, which were designed to work without this option, adding this option will avoid breaking the tests.

When using the option to apply neutral diffusion only below the surface boundary layer we were using (1.-zeta). This is wrong. It should be just (zeta).

Valid options are: * PARAM - use the vector-parameter LBD_DZ_TOP * UNIFORM[:N] - uniformly distributed * FILE:string - read from a file

For each tracer point: * tracers at (I,j), (I+1,j), (i,J) and (i,J+1) are remapped to a defined zgrid; * Apply LBD, uflux and vflux are calculated using the zgrid * Remap fluxes to native grid * Apply tracer convergence in the native grid TODO: * cleanup * create a zgrid for each pair of column taking into consideration h_L, h_R, BLD_L and BLD_R.

* Add functions to merge thicknesses and BLDs * z_top is now defined every time-step using this information * added unit tests

* adding new functions to sort, swap, and remove duplications in 1D arrays * updating unit tests * clean the module

…idate-ncar-2020-08-11

A few fixes to get it to compile with gfortran.

Introduce a controllable stdout unit in MOM_io

…ed_10nov2020 Major updates to the LBD module

marshallward · 2020-12-16T22:01:11Z

@gustavo-marques I am seeing a unit_tests fail in our GNU debug build:

 MOM_lateral_boundary_diffusion, UNIT TEST FAILED: Equal hbl and same layer thicknesses (gradient from right to left)
Layer=  1 F_calc=  0.0000000000000000 F_ans -2.0000000000000000                 
Layer=  2 F_calc=  0.0000000000000000 F_ans  0.0000000000000000                 FATAL: MOM_unit_tests: near_boundary_unit_tests FAILED

Do you want to look into this? Can you try to reproduce?

adcroft · 2020-12-16T22:03:18Z

@jiandewang, @kshedstrom, @abozec (or @awallcraft) can you review/test this PR per usual process? We're fairly happy with it other than the above unit test issue but continue to test.

gustavo-marques · 2020-12-16T22:08:08Z

@marshallward, it passed all unit tests with Intel 19.0.5. Let me try with GNU...

marshallward · 2020-12-16T22:13:08Z

I get a fail on Intel 18.0.6 (with slightly more info since it's a REPRO build):

 MOM_lateral_boundary_diffusion, UNIT TEST FAILED: 
 Equal hbl and same layer thicknesses (gradient from right to left)             
  
Layer=  1 F_calc=  0.0000000000000000 F_ans -2.0000000000000000
Layer=  2 F_calc=  0.0000000000000000 F_ans  0.0000000000000000
 MOM_lateral_boundary_diffusion, UNIT TEST FAILED: 
 Equal hbl and same layer thicknesses (gradient from left to right)             
  
Layer=  1 F_calc=  0.0000000000000000 F_ans  1.0000000000000000
Layer=  2 F_calc=  0.0000000000000000 F_ans  0.0000000000000000
Layer=  1 F_calc= -1.0000000000000000 F_ans -1.0000000000000000
 MOM_lateral_boundary_diffusion, UNIT TEST FAILED: 
 hbl < column thickness, hbl same, linear profile right, khtr=2                 
  
Layer=  2 F_calc=  0.0000000000000000 F_ans -4.0000000000000000
Layer=  1 F_calc=  0.0000000000000000 F_ans  0.0000000000000000
Layer=  2 F_calc=  0.0000000000000000 F_ans  0.0000000000000000
 MOM_lateral_boundary_diffusion, UNIT TEST FAILED: 
 Different hbl and different column thicknesses (gradient from left to right)   
  
Layer=  1 F_calc=  0.0000000000000000 F_ans 10.0000000000000000
Layer=  2 F_calc=  0.0000000000000000 F_ans  0.0000000000000000

FATAL: MOM_unit_tests: near_boundary_unit_tests FAILED

I'll try to repeat this with Intel 19.

gustavo-marques · 2020-12-16T22:14:33Z

Okay, I found the problem (I thought I had merged this NCAR#169, but apparently I did not). I will send a fix in a few.

…0-12-15

kshedstrom · 2020-12-16T23:48:38Z

I approve this PR.

marshallward · 2020-12-17T02:56:20Z

I still seem to be getting a fail, although I am now getting different answers (-0.4 rather than 0.0):

 MOM_lateral_boundary_diffusion, UNIT TEST FAILED: Equal hbl and same layer thicknesses (gradient from right to left)              
Layer=  1 F_calc= -0.4000000000000000 F_ans -2.0000000000000000
Layer=  2 F_calc=  0.0000000000000000 F_ans  0.0000000000000000

FATAL: MOM_unit_tests: near_boundary_unit_tests FAILED

Could be something to do with the merge; I will pick this up tomorrow.

marshallward · 2020-12-17T02:58:17Z

I just tried an Intel "REPRO" build and it did pass, so there could still be an underlying problem here (e.g. assumed zero initialization?).

Edit: PGI also passes, so this problem seems restricted to GNU.

I don't know anything about this test, so I won't be able to sort this out quickly, but I'll start looking into it. In the meantime, can you re-try your test using GNU?

gustavo-marques · 2020-12-17T15:47:23Z

Thanks, @marshallward. I will try to reproduce the fail with GNU.

awallcraft · 2020-12-17T15:53:40Z

I approve this PR.

gustavo-marques · 2020-12-18T00:58:37Z

@marshallward, I was able to reproduce the fail with GNU in debug mode. The test is failing because of an undefined logical (CS%limiter_remap) in the unit test function (near_boundary_unit_tests). For some reason, GNU/debug sets it to True by default. The fix coming in a few.

jiandewang · 2020-12-18T04:21:49Z

EMC wcoss machine is reserved for production test only at this moment, will test this MOM code once access to machine is back

marshallward · 2020-12-18T10:43:56Z

Thanks, @gustavo-marques , good to see it was nothing serious. This is now passing on our branch with no regressions (after resolving a minor git conflict).

Hallberg-NOAA

I have visually examined all of these code changes, and I approve of them. (I especially like the use of the deferred allocation character strings, which is a very elegant solution!) Also, I have been informed that this PR is now passing all of the testing from the GFDL side.

Hallberg-NOAA · 2020-12-18T14:32:48Z

The merge of this PR onto the main branch is now just waiting for the go-ahead from the EMC team via @jiandewang, having obtained approval from the other main forks.

jiandewang · 2020-12-18T15:08:36Z

I shall have access to machine this afternoon, will try to finish testing no later than today

jiandewang · 2020-12-18T17:28:13Z

works fine on EMC (PS: I started my testing 1 minute after machine is back, so my job got finished quickly as nodes are empty)

marshallward · 2020-12-19T00:08:05Z

Approvals:

NCAR (submitted)
FSU
GFDL
Rutgers
EMC

I am going to merge this now.

gustavo-marques and others added 30 commits March 2, 2020 12:34

Remove dependence on the boundary layer scheme

a446d0f

Modifications to run idealized two-column LBD

b5132d9

* turned off pressure force * hard-coded BLD * turned off advect_tracer * set NTR = 2

Merge branch 'fix_LBD_tests' into LBD_idealized_experiments

3b121cb

Minor improvements in the bulk method

92581e2

Fix bug when applying ND only in the interior

7299d86

When using the option to apply neutral diffusion only below the surface boundary layer we were using (1.-zeta). This is wrong. It should be just (zeta).

Add option to apply linear decay at the base of hbl

09ad695

Merge branch 'dev/ncar' into LBD_idealized_experiments

7e39059

comment calls to BLD schemes

a49954a

Fix bug in loop-indices when applying linear decay

1fd5393

Fix bug in loop-indices when using discontinuous ND and interior_only

9957e55

Rename lateral_boundary_diffusion_CS to lbd_CS

d8ae09f

First attempt to use remapping in LBD

2a80964

Merge branch 'LBD_idealized_experiments2' into regrid_LBD

fa376e4

Adding options to specify dz to apply LBD @ top

5c56268

Valid options are: * PARAM - use the vector-parameter LBD_DZ_TOP * UNIFORM[:N] - uniformly distributed * FILE:string - read from a file

Adding option to output u^2, v^2 and u*v

1849bcc

Fix typo when posting vsq

e0db9ab

Delete fluxes_bulk_method

593ebfc

Define vertical grid on-the-fly using h's and BLDs

441c34e

* Add functions to merge thicknesses and BLDs * z_top is now defined every time-step using this information * added unit tests

Passes GV to lateral_boundary_diffusion_init

40136b6

Add option to initialize passive tracer from z file

e537ef2

Cleanup + add unit tests

365b298

Remove hard-coded BLD and uncomment code for adv and PGF

4ab4a7a

Comment out write statements

f444aff

Improve the merging of interfaces

2538c4d

* adding new functions to sort, swap, and remove duplications in 1D arrays * updating unit tests * clean the module

Adds missing args for KPP_get_BLD and energetic_PBL_get_MLD

87ba533

Adds missing args for KPP_get_BLD and energetic_PBL_get_MLD

f0face4

Merge pull request mom-ocean#158 from gustavo-marques/dev-master-cand…

ddbd9c7

…idate-ncar-2020-08-11

alperaltuntas and others added 6 commits December 10, 2020 19:30

Merge pull request mom-ocean#168 from kshedstrom/gnu_fix

0c16805

A few fixes to get it to compile with gfortran.

Merge pull request mom-ocean#163 from NCAR/stdout

8215b70

Introduce a controllable stdout unit in MOM_io

Fix bug in linear decay and set F_layer = 0 below htot_max

cf97095

Merge pull request mom-ocean#164 from gustavo-marques/LBD_regrid_merg…

eddbd6c

…ed_10nov2020 Major updates to the LBD module

Merge branch 'main' into main-candidate-ncar-2020-12-15

42d79c2

Delete forgotten code

51ed387

gustavo-marques requested a review from Hallberg-NOAA December 16, 2020 15:03

Merge branch 'delete_fotgotten_code_LBD' into main-candidate-ncar-202…

5739a34

…0-12-15

Sets CS%limiter_remap=.false. in near_boundary_unit_tests

9d04f38

Hallberg-NOAA approved these changes Dec 18, 2020

View reviewed changes

marshallward merged commit 0f664aa into mom-ocean:main Dec 19, 2020

Hallberg-NOAA mentioned this pull request Apr 19, 2021

Buffer overflow logging ALE_RESOLUTION #214

Closed

gustavo-marques deleted the main-candidate-ncar-2020-12-15 branch April 21, 2021 23:26

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Main candidate from NCAR (12/15/2020) #1275

Main candidate from NCAR (12/15/2020) #1275

gustavo-marques commented Dec 16, 2020

marshallward commented Dec 16, 2020

adcroft commented Dec 16, 2020

gustavo-marques commented Dec 16, 2020

marshallward commented Dec 16, 2020

gustavo-marques commented Dec 16, 2020

kshedstrom commented Dec 16, 2020

marshallward commented Dec 17, 2020

marshallward commented Dec 17, 2020 •

edited

Loading

gustavo-marques commented Dec 17, 2020

awallcraft commented Dec 17, 2020

gustavo-marques commented Dec 18, 2020 •

edited

Loading

jiandewang commented Dec 18, 2020

marshallward commented Dec 18, 2020

Hallberg-NOAA left a comment

Hallberg-NOAA commented Dec 18, 2020

jiandewang commented Dec 18, 2020

jiandewang commented Dec 18, 2020

marshallward commented Dec 19, 2020

Main candidate from NCAR (12/15/2020) #1275

Main candidate from NCAR (12/15/2020) #1275

Conversation

gustavo-marques commented Dec 16, 2020

Major changes

New diagnostics

marshallward commented Dec 16, 2020

adcroft commented Dec 16, 2020

gustavo-marques commented Dec 16, 2020

marshallward commented Dec 16, 2020

gustavo-marques commented Dec 16, 2020

kshedstrom commented Dec 16, 2020

marshallward commented Dec 17, 2020

marshallward commented Dec 17, 2020 • edited Loading

gustavo-marques commented Dec 17, 2020

awallcraft commented Dec 17, 2020

gustavo-marques commented Dec 18, 2020 • edited Loading

jiandewang commented Dec 18, 2020

marshallward commented Dec 18, 2020

Hallberg-NOAA left a comment

Choose a reason for hiding this comment

Hallberg-NOAA commented Dec 18, 2020

jiandewang commented Dec 18, 2020

jiandewang commented Dec 18, 2020

marshallward commented Dec 19, 2020

marshallward commented Dec 17, 2020 •

edited

Loading

gustavo-marques commented Dec 18, 2020 •

edited

Loading