add ability to taper analysis perts near top of model #729

jswhit · 2024-03-24T23:53:25Z

Fixes #728

Addresses issue #728 by adding a logical taperanalperts to the namelist. The vertical profile of the taper profile

log(ak(k) - ak_top)/log(ak_bot - ak_top)

is controlled by parameters taperanalperts_aktop and taperanalperts_akbot. Default values are 500 and -1, which results in a taper that starts at level 19 and has a value of 0.11 at level 1 for the 127 level GFS.

@CatherineThomas-NOAA - I decided it would be simpler and less intrusive to the workflow to include this functionality in the EnKF code instead of the recentering utility.

TODO: implement for regional models

Checklist

My code follows the style guidelines of this project
I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
New and existing tests pass with my changes
Any dependent changes have been merged and published

RussTreadon-NOAA · 2024-03-25T10:50:26Z

@jswhit , who might we ask to review these changes? We need a minimum of two peer reviewers and approvals.

jswhit2 · 2024-03-25T17:35:58Z

@jswhit , who might we ask to review these changes? We need a minimum of two peer reviewers and approvals.

I suggest @CatherineThomas-NOAA and @TingLei-NOAA

RussTreadon-NOAA · 2024-03-25T17:37:33Z

Thanks @jswhit . Cathy and Ting have been added as peer reviewers.

TingLei-NOAA · 2024-03-26T11:26:59Z

@jswhit I am a little confused on the profile generated by

log(ak(k) - ak_top)/log(ak_bot - ak_top)

For example, as ak(k) approaches ak_top from the positive side (always such case), with a difference from ak_top less than 1, the magnitude of log(ak(k)-ak_top) would increase, while it is a negative number. Such values are not what we expect for the tamper coefficients, right? While "(log(ak(k))-log(ak_top))/(log(ak_bot/ak_top) seems to be no such issues with 0 on ak_top. Wondering if you would have better profiles. Please let me know if I missed any implicit properties for ak. Your further clarification will be appreciated.

jswhit2 · 2024-03-26T19:04:56Z

@Ting.Lee the default values of akbot and aktop are 500 and -1. log(akbot-aktop) = 6.21. For 500<ak<-1, log(ak-aktop) is decreases from 6.21 to about 7 for ak=1 (the top level of the model). Thus the taper profile decreases from 1 when ak=500 to about 0.1 when ak=1.

TingLei-NOAA · 2024-03-26T19:42:30Z

@jswhit Thanks for this explanation. With the aktop=-1 , this indeed works as your described.

src/enkf/gridinfo_nmmb.f90

RussTreadon-NOAA · 2024-03-27T19:38:31Z

Orion ctests

Clone jswhit:taperanalperts on Orion and merge develop at b53740a into working copy. Run ctests with following results

Test project /work2/noaa/da/rtreadon/git/gsi/pr729/build
    Start 1: global_4denvar
    Start 2: rtma
    Start 3: rrfs_3denvar_glbens
    Start 4: netcdf_fv3_regional
    Start 5: hafs_4denvar_glbens
    Start 6: hafs_3denvar_hybens
    Start 7: global_enkf
1/7 Test #3: rrfs_3denvar_glbens ..............   Passed  786.10 sec
2/7 Test #4: netcdf_fv3_regional ..............   Passed  844.88 sec
3/7 Test #7: global_enkf ......................   Passed  909.35 sec
4/7 Test #2: rtma .............................   Passed  1269.14 sec
5/7 Test #6: hafs_3denvar_hybens ..............   Passed  1404.80 sec
6/7 Test #5: hafs_4denvar_glbens ..............   Passed  1641.61 sec
7/7 Test #1: global_4denvar ...................   Passed  2043.23 sec

100% tests passed, 0 tests failed out of 7

Total Test time (real) = 2043.25 sec

All tests pass. This is expected.. This PR only changes files in src/enkf. All updat and contrl gsi.x based tests should be identical since the source code is identical. We also do not expect any difference in the enkf.x test. The taper option added by the PR is off by default. src/enkf/params.f90 contains

logical, public :: taperanalperts = .false.

Thus, the global_enkf analysis results should be identical. The Orion tests confirm this.

RussTreadon-NOAA · 2024-03-27T19:51:08Z

Hercules ctests

Repeat Orion test on Hercules with the following results

Test project /work/noaa/da/rtreadon/git/gsi/pr729/build
    Start 1: global_4denvar
    Start 2: rtma
    Start 3: rrfs_3denvar_glbens
    Start 4: netcdf_fv3_regional
    Start 5: hafs_4denvar_glbens
    Start 6: hafs_3denvar_hybens
    Start 7: global_enkf
1/7 Test #4: netcdf_fv3_regional ..............   Passed  482.88 sec
2/7 Test #3: rrfs_3denvar_glbens ..............   Passed  484.93 sec
3/7 Test #7: global_enkf ......................   Passed  725.90 sec
4/7 Test #2: rtma .............................   Passed  964.86 sec
5/7 Test #6: hafs_3denvar_hybens ..............   Passed  1090.78 sec
6/7 Test #5: hafs_4denvar_glbens ..............***Failed  1095.93 sec
7/7 Test #1: global_4denvar ...................   Passed  1741.54 sec

86% tests passed, 1 tests failed out of 7

Total Test time (real) = 1741.64 sec

The following tests FAILED:
          5 - hafs_4denvar_glbens (Failed)

hafs_4denvar_glbens failed due to

The fv3_dynvars are reproducible
The fv3_sfcdata are reproducible
The results between the two runs (hafs_4denvar_glbens_loproc_updat and hafs_4denvar_glbens_loproc_contrl) are not reproducible. Thus, the case has Failed siganl of the regression tests.

Comparison of the loproc_contrl and loproc_updat fv3_tracer output files shows very small differences in the o3mr field

o3mr min/max 1=1.4650413e-08,1.6894655e-05 min/max 2=1.4650413e-08,1.6894655e-05 max abs diff=0.0000000450

Interestingly, the hiproc_updat fv3_tracer file is identical to the fv3_tracer file generated by the loproc_contrl and hiproc_contrl runs. The loproc_updat is the anomalous run. Non-reproducibility in regional analysis files is a known problem on Hercules.

All other ctests Passed on Hercules.

RussTreadon-NOAA · 2024-03-27T20:29:35Z

Hera ctests

Repeat Hercules test on Hera with the following results

Test project /scratch1/NCEPDEV/da/Russ.Treadon/git/gsi/pr729/build
    Start 1: global_4denvar
    Start 2: rtma
    Start 3: rrfs_3denvar_glbens
    Start 4: netcdf_fv3_regional
    Start 5: hafs_4denvar_glbens
    Start 6: hafs_3denvar_hybens
    Start 7: global_enkf
1/7 Test #3: rrfs_3denvar_glbens ..............   Passed  550.06 sec
2/7 Test #4: netcdf_fv3_regional ..............   Passed  608.17 sec
3/7 Test #7: global_enkf ......................***Failed  998.36 sec
4/7 Test #2: rtma .............................   Passed  1149.47 sec
5/7 Test #6: hafs_3denvar_hybens ..............   Passed  1223.71 sec
6/7 Test #5: hafs_4denvar_glbens ..............   Passed  1348.94 sec
7/7 Test #1: global_4denvar ...................   Passed  1986.77 sec

86% tests passed, 1 tests failed out of 7

Total Test time (real) = 1986.81 sec

The following tests FAILED:
          7 - global_enkf (Failed)

global_enkf failed due to the time threshold test.

The runtime for global_enkf_loproc_updat is 168.053597 seconds.  This has exceeded maximum allowable threshold time of 163.541045 seconds, resulting in Failure timethresh of the regression test.

A check of enkf.x wall times returns

global_enkf_hiproc_contrl/stdout:The total amount of wall time                        = 125.916447
global_enkf_hiproc_updat/stdout:The total amount of wall time                        = 128.483509
global_enkf_loproc_contrl/stdout:The total amount of wall time                        = 149.715914
global_enkf_loproc_updat/stdout:The total amount of wall time                        = 170.181588

While the loproc_updat enkf.x wall time is higher than the loproc_contrl is it not anomalously larger. This is not a fatal fail.

RussTreadon-NOAA · 2024-03-27T20:30:49Z

WCOSS2 (Dogwood) ctests

Repeat Hera test on Dogwood with the following results

Test project /lfs/h2/emc/da/noscrub/russ.treadon/git/gsi/pr729/build
    Start 1: global_4denvar
    Start 2: rtma
    Start 3: rrfs_3denvar_glbens
    Start 4: netcdf_fv3_regional
    Start 5: hafs_4denvar_glbens
    Start 6: hafs_3denvar_hybens
    Start 7: global_enkf
1/7 Test #4: netcdf_fv3_regional ..............   Passed  604.61 sec
2/7 Test #3: rrfs_3denvar_glbens ..............   Passed  665.51 sec
3/7 Test #7: global_enkf ......................   Passed  970.00 sec
4/7 Test #6: hafs_3denvar_hybens ..............   Passed  1212.18 sec
5/7 Test #2: rtma .............................   Passed  1327.94 sec
6/7 Test #5: hafs_4denvar_glbens ..............   Passed  1392.89 sec
7/7 Test #1: global_4denvar ...................   Passed  1862.29 sec

100% tests passed, 0 tests failed out of 7

Total Test time (real) = 1862.30 sec

As expected, all tests pass.

RussTreadon-NOAA · 2024-03-27T20:36:26Z

@jswhit , when you have time, please merge recent commits to the authoritative develop into jswhit:taperanalperts. Your branch is currently 3 commits behind develop.

I am on leave 4/1 through 4/11. Hence my running ctests with the hope of merging this PR into develop before COB 3/29 pending approval from peer reviewers.

jswhit2 · 2024-03-27T21:45:33Z

@jswhit , when you have time, please merge recent commits to the authoritative develop into jswhit:taperanalperts. Your branch is currently 3 commits behind develop.

I am on leave 4/1 through 4/11. Hence my running ctests with the hope of merging this PR into develop before COB 3/29 pending approval from peer reviewers.

done!

RussTreadon-NOAA

Changes look good. Review comments are minor and fall into the nitpick category - nothing serious or major.

src/enkf/gridinfo_fv3reg.f90

src/enkf/gridinfo_gfs.f90

src/enkf/gridinfo_nmmb.f90

src/enkf/gridinfo_wrf.f90

src/enkf/inflation.f90

src/enkf/params.f90

CatherineThomas-NOAA · 2024-03-28T13:22:44Z

@jswhit Just want to confirm that the description:

is controlled by parameters taperanalperts_aktop and taperanalperts_akbot. Default values are 500 and -1,

is swapped, correct? aktop=-1 and ak_bot=500? This is what's in the code and what mathematically makes sense.

RussTreadon-NOAA · 2024-03-28T13:46:18Z

Rerun ctests on Hera Rocky 8 nodes using taperanalperts at 87af1e2. All tests pass on rerun.

Test project /scratch1/NCEPDEV/da/Russ.Treadon/git/gsi/pr729b/build
    Start 1: global_4denvar
    Start 2: rtma
    Start 3: rrfs_3denvar_glbens
    Start 4: netcdf_fv3_regional
    Start 5: hafs_4denvar_glbens
    Start 6: hafs_3denvar_hybens
    Start 7: global_enkf
1/7 Test #3: rrfs_3denvar_glbens ..............   Passed  1330.32 sec
2/7 Test #4: netcdf_fv3_regional ..............   Passed  1387.52 sec
3/7 Test #7: global_enkf ......................   Passed  1522.12 sec
4/7 Test #6: hafs_3denvar_hybens ..............   Passed  1821.52 sec
5/7 Test #2: rtma .............................   Passed  2052.07 sec
6/7 Test #5: hafs_4denvar_glbens ..............   Passed  2068.81 sec
7/7 Test #1: global_4denvar ...................   Passed  3967.34 sec

100% tests passed, 0 tests failed out of 7

Total Test time (real) = 3967.40 sec

jswhit2 · 2024-03-28T14:50:40Z

@jswhit Just want to confirm that the description:

is controlled by parameters taperanalperts_aktop and taperanalperts_akbot. Default values are 500 and -1,

is swapped, correct? aktop=-1 and ak_bot=500? This is what's in the code and what mathematically makes sense.

yes, it is swapped - sorry about that

RussTreadon-NOAA

Thank you @jswhit for patiently responding to my nitpicks. Looks good. Just two more lines which need precision specified in params.f90. I missed this on my first sweep. Sorry.

src/enkf/params.f90

RussTreadon-NOAA

Great! Thank you @jswhit for the quick commit.

Approve.

CatherineThomas-NOAA

Changes look good to me. Thanks @jswhit for adding this!

Jeffrey S Whitaker added 6 commits March 24, 2024 18:45

add ability to taper analysis perts near top of model

0a6c84a

update

1bfb756

add taper_vert to regional models (but set to 1.0)

5c6d350

fix typo

c2f8bbf

fix typo

9f7c049

fix name clash

432220d

RussTreadon-NOAA added this to the GFS v17 GSI/EnKF development milestone Mar 25, 2024

RussTreadon-NOAA assigned jswhit and RussTreadon-NOAA Mar 25, 2024

RussTreadon-NOAA requested review from CatherineThomas-NOAA and TingLei-NOAA March 25, 2024 17:36

define tapervert using ak,bk if taperanalperts=T

a0b482b

TingLei-NOAA reviewed Mar 27, 2024

View reviewed changes

src/enkf/gridinfo_nmmb.f90 Show resolved Hide resolved

TingLei-NOAA approved these changes Mar 27, 2024

View reviewed changes

Merge branch 'develop' into taperanalperts

87af1e2

RussTreadon-NOAA self-requested a review March 28, 2024 12:38

RussTreadon-NOAA reviewed Mar 28, 2024

View reviewed changes

address @RussTreadon-NOAA's comments

3116c80

RussTreadon-NOAA self-requested a review March 28, 2024 17:13

RussTreadon-NOAA reviewed Mar 28, 2024

View reviewed changes

src/enkf/params.f90 Outdated Show resolved Hide resolved

specify kind of default namelist parameters

9df1215

RussTreadon-NOAA self-requested a review March 28, 2024 17:34

RussTreadon-NOAA approved these changes Mar 28, 2024

View reviewed changes

CatherineThomas-NOAA approved these changes Mar 29, 2024

View reviewed changes

RussTreadon-NOAA merged commit b2fc4fd into NOAA-EMC:develop Mar 29, 2024
4 checks passed

CatherineThomas-NOAA mentioned this pull request Nov 13, 2024

Enable tapering of atm ensemble spread at the model top NOAA-EMC/global-workflow#3096

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add ability to taper analysis perts near top of model #729

add ability to taper analysis perts near top of model #729

jswhit commented Mar 24, 2024 •

edited by RussTreadon-NOAA

Loading

RussTreadon-NOAA commented Mar 25, 2024

jswhit2 commented Mar 25, 2024

RussTreadon-NOAA commented Mar 25, 2024

TingLei-NOAA commented Mar 26, 2024

jswhit2 commented Mar 26, 2024

TingLei-NOAA commented Mar 26, 2024

RussTreadon-NOAA commented Mar 27, 2024

RussTreadon-NOAA commented Mar 27, 2024

RussTreadon-NOAA commented Mar 27, 2024

RussTreadon-NOAA commented Mar 27, 2024

RussTreadon-NOAA commented Mar 27, 2024

jswhit2 commented Mar 27, 2024

RussTreadon-NOAA left a comment

CatherineThomas-NOAA commented Mar 28, 2024

RussTreadon-NOAA commented Mar 28, 2024

jswhit2 commented Mar 28, 2024

RussTreadon-NOAA left a comment

RussTreadon-NOAA left a comment

CatherineThomas-NOAA left a comment

add ability to taper analysis perts near top of model #729

add ability to taper analysis perts near top of model #729

Conversation

jswhit commented Mar 24, 2024 • edited by RussTreadon-NOAA Loading

RussTreadon-NOAA commented Mar 25, 2024

jswhit2 commented Mar 25, 2024

RussTreadon-NOAA commented Mar 25, 2024

TingLei-NOAA commented Mar 26, 2024

jswhit2 commented Mar 26, 2024

TingLei-NOAA commented Mar 26, 2024

RussTreadon-NOAA commented Mar 27, 2024

RussTreadon-NOAA commented Mar 27, 2024

RussTreadon-NOAA commented Mar 27, 2024

RussTreadon-NOAA commented Mar 27, 2024

RussTreadon-NOAA commented Mar 27, 2024

jswhit2 commented Mar 27, 2024

RussTreadon-NOAA left a comment

Choose a reason for hiding this comment

CatherineThomas-NOAA commented Mar 28, 2024

RussTreadon-NOAA commented Mar 28, 2024

jswhit2 commented Mar 28, 2024

RussTreadon-NOAA left a comment

Choose a reason for hiding this comment

RussTreadon-NOAA left a comment

Choose a reason for hiding this comment

CatherineThomas-NOAA left a comment

Choose a reason for hiding this comment

jswhit commented Mar 24, 2024 •

edited by RussTreadon-NOAA

Loading