Eliminate artificial supercooled water (rain and cloud) on pressure levels #771

jaymes-kenyon · 2023-09-06T20:23:02Z

Updates to prevent artificial supercooled rain and cloud water (an aviation hazard) from appearing on pressure levels. This code seeks to fix a problem that was identified/described by Greg Thompson (NCAR) in HRRR pressure-level GRIB2 output. The accompanying figure (linked below) illustrates a scenario in which supercooled rain water can appear in pressure levels when no supercooled rain water is actually present on the adjacent model levels.
isobaric_interpolation.pdf

…pressure levels

jaymes-kenyon · 2023-09-06T20:33:40Z

@WenMeng-NOAA @HuiyaChuang-NOAA @EricJames-NOAA Here is a draft PR intended to prevent spurious supercooled rain water from appearing on pressure levels. Testing is still ongoing, but feedback / ideas are welcome. Should this new code only be used for RRFS (i.e., if MODELNAME = FV3R ...)? Thanks!

WenMeng-NOAA · 2023-09-06T20:36:25Z

@jaymes-kenyon Do you have an UPP issue for this PR? If not, please open one.

(2) Slight change to freezing-temp threshold (now 273.15 K) (3) Revised description in comments (4) Added print statements (temporary)

jaymes-kenyon · 2023-09-07T16:05:04Z

@gthompsnJCSDA : Here's a draft PR to fix the issue of spurious supercooled rain water; comments are welcome.

…arlier in MDL2P.f)

…instead of local, hardcoded value

jaymes-kenyon · 2023-09-11T19:16:55Z

Code has been tested in RRFS 3-km CONUS cases for summer and winter; no GFS tests have been performed.

In the summer case, the code results in Qrain adjustments at ~140k grid points on pressure levels (in 3D). Of these adjustments, ~18k (or ~13%) are set to zero (i.e., supercooled rain is eliminated).
In the winter case, Qrain adjustments occur at ~193k grid points. Of these, ~11k (or ~6%) are set to zero.

These results suggest instances of supercooled rain on pressure levels arising purely as an interpolation artifact are relatively rare. In other words, instances of supercooled rain on pressure levels (near the melting level) are usually accompanied by explicit supercooled rain on the overlying model level. It's not known whether this is the expected result; ideas or concerns are welcome.

sorc/ncep_post.fd/MDL2P.f

(2) Applying adjustments to all hydrometeors

gthompsnJCSDA · 2023-09-12T15:08:48Z

I can think of a really serious potential pitfall. How about the most complex case of more than one crossing (in vertical) of the melting level. I.e., a classic freezing rain profile with the "warm nose?" This really should be checked before proceeding. I believe we would have a really obvious ICICLE case - but I have to check the date, perhaps 2019Feb05.

And I agree if the idea works right for rain, then adapt to cloud water as well. For cloud ice going lower in altitude than melting level, I don't think there would be a problem.

jaymes-kenyon · 2023-09-12T15:34:51Z

I can think of a really serious potential pitfall. How about the most complex case of more than one crossing (in vertical) of the melting level. I.e., a classic freezing rain profile with the "warm nose?" This really should be checked before proceeding. I believe we would have a really obvious ICICLE case - but I have to check the date, perhaps 2019Feb05.

And I agree if the idea works right for rain, then adapt to cloud water as well. For cloud ice going lower in altitude than melting level, I don't think there would be a problem.

Thanks, @gthompsnJCSDA — I mulled the possibility of a "warm nose" and I think this code will handle that situation appropriately, but I am prepared to test further to confirm. Lines 593–595 require a true melting level to proceed; i.e., "cold above, warm below":

                   IF ( TSL(I,J) <= TFRZ .AND.    & ! This pressure level is subfreezing and located just above a melting level;
                        T(I,J,LL-1) <= TFRZ .AND. & ! i.e., the overlying model level is subfreezing,
                        T(I,J,LL) > TFRZ ) THEN     ! but the underlying model level is above freezing.

(note that the LL index increases toward the ground, so LL-1 is actually above LL).

Given this temperature check, a re-freezing level (i.e., "warm above, cold below")—such as would exist at the base of the "warm nose" in a freezing rain situation—will be untouched by this code.

… the interpolated value, whichever is greater. This modified approach is reccomended by G. Thompson in order to retain larger values of snow/graupel whenever these mixing ratios increase downward across the pressure level (e.g., snow growth via cloud-drop collection prior to melting).

jaymes-kenyon · 2023-09-15T20:28:02Z

At the recommendation of @gthompsnJCSDA, I tested this code in a RRFS winter case (8 Feb 2022), then searched for a model column containing a "warm nose" thermal profile with freezing rain at the surface. I found such a column over New England (p = 1002 mb on the lowest-layer midpoint). Using "print" statements, the profiles of temperature (middle column) and Qrain (right column) for 600 mb and below were extracted:

  600 mb   -15.80606     degC  1.0000000E-12
  625 mb   -13.89532     degC  1.0000000E-12
  650 mb   -11.59238     degC  1.0000000E-12
  675 mb   -9.432922     degC  1.0000000E-12
  700 mb   -8.086395     degC  1.0000000E-12
  725 mb   -6.734436     degC  1.0000000E-12
  750 mb   -5.336792     degC  1.0000000E-12
  775 mb   -3.949310     degC  1.6138863E-12
  800 mb   -2.553467     degC  5.9035853E-11
  825 mb   -1.372528     degC  3.3919179E-10
  850 mb  -0.4472656     degC  1.8346782E-08  <- updated
  875 mb   0.4566956     degC  2.6089851E-05
  900 mb    1.366028     degC  8.8202418E-05
  925 mb    1.859070     degC  1.4896749E-04
  950 mb    2.156738     degC  1.8997306E-04
  975 mb    1.247162     degC  2.1207312E-04
 1000 mb  -0.2186890     degC  2.2372934E-04

The warm nose (with T > 0C) spans the ~875–975-mb layer. At 850 mb, the "updated" tag indicates where Qrain was modified from its original interpolated value. Otherwise, all other values of Qrain were unaffected by the code in this PR, including the freezing rain at 1000 mb. This suggests that the code retains supercooled rain beneath re-freezing levels, as intended. Note that the values of Qrain aloft (1.0000000E-12) correspond to the 'zero' value in MDL2P.f; this can be modified to be actual zeros, if desired.

gthompsnJCSDA · 2023-09-15T20:35:33Z

what about those super tiny amounts of rain at 825 and 800 mb? Those appear possibly interpolated values as well.

jaymes-kenyon · 2023-09-15T21:44:34Z

@gthompsnJCSDA — That's right. In fact, all of the Qrain values in the table above are the interpolated values that the original UPP code is providing, except at 850 mb. The "<-updated" tag at 850 mb indicates that my code supplied this value by extracting it from the overlying model level, overwriting the original interpolation.

Here are some more print statements placed before my code block. First, at 850 mb:

 At requested pressure:          850 mb 
 ...overlying model     P:    844.1834    
 ...overlying model     T:  -0.7026978    
 ...overlying model Qrain:   1.8346782E-08
  ---   
 ...underlying model     P:    862.0761    
 ...underlying model     T:   7.7484131E-02
 ...underlying model Qrain:   5.9126928E-06
   
 *** Interpolated Qrain is :  1.9481072E-06

Here at 850 mb, the interpolated value was later replaced by the value from the overlying model level (two orders of magnitude smaller) as a result of my code (see previous table).

At 825 mb, small amounts of supercooled rain are present on the adjacent model levels; the interpolated value goes unmodified by my code:

 At requested pressure:          825 mb 
 ...overlying model     P:    824.7617    
 ...overlying model     T:   -1.380951    
 ...overlying model Qrain:   1.1281667E-10
  ---  
 ...underlying model     P:    844.1834    
 ...underlying model     T:  -0.7026978    
 ...underlying model Qrain:   1.8346782E-08
   
 *** Interpolated Qrain is :  3.3919179E-10

At 800 mb, very small amounts of supercooled rain are present on the adjacent model levels (again, the interpolated value is retained):

 At requested pressure:          800 mb 
 ...overlying model     P:    781.2208    
 ...overlying model     T:   -3.609955    
 ...overlying model Qrain:   2.1661874E-12
  ---  
 ...underlying model     P:    803.7781    
 ...underlying model     T:   -2.343872    
 ...underlying model Qrain:   7.0317543E-11
   
 *** Interpolated Qrain is :  5.9035853E-11

Hopefully this is insightful. It looks like our proposed code is behaving as intended. Other concerns/ideas?

jaymes-kenyon · 2023-09-20T12:46:15Z

@WenMeng-NOAA — If no further questions/concerns, I think we can continue moving forward with this PR. Testing in RRFS confirms that this code will fix the problem of spurious supercooled water arising purely from interpolation; it leaves most (real) supercooled water intact. No tests have been performed using the GFS. Thanks to @gthompsnJCSDA for helping to describe the original problem and advising on this PR.

WenMeng-NOAA · 2023-09-20T16:14:17Z

@WenMeng-NOAA — If no further questions/concerns, I think we can continue moving forward with this PR. Testing in RRFS confirms that this code will fix the problem of spurious supercooled water arising purely from interpolation; it leaves most (real) supercooled water intact. No tests have been performed using the GFS. Thanks to @gthompsnJCSDA for helping to describe the original problem and advising on this PR.

@jaymes-kenyon I will start my testing. Thanks!

WenMeng-NOAA · 2023-09-22T13:19:11Z

@jaymes-kenyon Can you sync your branch with the latest UPP develop branch? We will process your PR next. Thanks!

…re_interpolation_update

jaymes-kenyon · 2023-09-22T13:40:10Z

@WenMeng-NOAA — I just synched with "develop"; hopefully this helps. Thanks!

sorc/ncep_post.fd/MDL2P.f

WenMeng-NOAA · 2023-09-22T17:39:56Z

@jaymes-kenyon During reviewing your PR, I find a bug of cwm initialization in INITPOST_NETCDF.f. Would you mind to pick up my fix and combine in your PR? My fix is at /scratch1/NCEPDEV/stmp2/Wen.Meng/jaymes/UPP/sorc/ncep_post.fd/INITPOST_NETCDF.f

[Wen.Meng@hfe03 ncep_post.fd]$ git status
On branch pressure_interpolation_update
Your branch is up to date with 'origin/pressure_interpolation_update'.

Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git checkout -- <file>..." to discard changes in working directory)

        modified:   INITPOST_NETCDF.f

Thanks!

jaymes-kenyon · 2023-09-22T18:44:22Z

Good idea, @WenMeng-NOAA — I added checks for each hydrometeor species. Based on line 512, I think that "C1D" can be formulated the same: if C1D exists, then CWM exists, which also means that all five species exist. But, let me know if I am mistaken. You bug fix has also been added to the PR.

WenMeng-NOAA · 2023-09-22T19:04:35Z

Good idea, @WenMeng-NOAA — I added checks for each hydrometeor species. Based on line 512, I think that "C1D" can be formulated the same: if C1D exists, then CWM exists, which also means that all five species exist. But, let me know if I am mistaken. You bug fix has also been added to the PR.

@jaymes-kenyon Looks good for me. Thanks!

WenMeng-NOAA · 2023-09-24T22:00:22Z

The UPP RT indicate changed results of these MP variables on isobaric levels for GFS, RRFS, 3DRTMA and HAFS such as:

gfs:  gfs.20190830/gfs.t00z.master.grb2f006
493:1277344354:CLMR:500 mb:rpn_corr=0.999999:rpn_rms=5.8857e-08
495:1278718847:RWMR:500 mb:rpn_corr=0.999853:rpn_rms=2.00914e-10
509:1321693641:CLMR:525 mb:rpn_corr=0.999724:rpn_rms=1.35698e-06
510:1322077105:ICMR:525 mb:rpn_corr=0.999997:rpn_rms=5.70781e-09
511:1323177936:RWMR:525 mb:rpn_corr=0.918638:rpn_rms=3.71626e-08
512:1323192010:SNMR:525 mb:rpn_corr=0.999953:rpn_rms=7.95879e-07
541:1412183285:CLMR:575 mb:rpn_corr=0.997774:rpn_rms=4.23497e-06
542:1412654320:ICMR:575 mb:rpn_corr=0.997294:rpn_rms=1.21677e-07
543:1413607815:RWMR:575 mb:rpn_corr=0.99955:rpn_rms=9.61382e-08
544:1413741913:SNMR:575 mb:rpn_corr=0.999704:rpn_rms=1.74645e-06
545:1414900843:GRLE:575 mb:rpn_corr=0.999999:rpn_rms=3.9135e-09
...

RRFS: fv3r_2023062800/PRSLEV10.tm00.diff
425:379040095:RWMR:500 mb:rpn_corr=0.999915:rpn_rms=2.71782e-07
438:394290403:CLMR:525 mb:rpn_corr=0.999966:rpn_rms=2.93396e-07
439:394397871:ICMR:525 mb:rpn_corr=0.999999:rpn_rms=7.27848e-10
440:394570609:RWMR:525 mb:rpn_corr=0.99921:rpn_rms=8.47557e-07
442:394692533:GRLE:525 mb:rpn_corr=0.999963:rpn_rms=7.90778e-07
453:409902027:CLMR:550 mb:rpn_corr=0.999229:rpn_rms=1.66224e-06
455:410166644:RWMR:550 mb:rpn_corr=0.998717:rpn_rms=1.21641e-06
456:410178311:SNMR:550 mb:rpn_corr=0.999979:rpn_rms=7.39704e-07
457:410291276:GRLE:550 mb:rpn_corr=0.999938:rpn_rms=1.00225e-06
468:425758237:CLMR:575 mb:rpn_corr=0.998616:rpn_rms=2.59907e-06
469:425898238:ICMR:575 mb:rpn_corr=0.999649:rpn_rms=3.98602e-09
470:426032130:RWMR:575 mb:rpn_corr=0.990812:rpn_rms=3.87284e-06
471:426047312:SNMR:575 mb:rpn_corr=0.999704:rpn_rms=2.47564e-06
472:426152535:GRLE:575 mb:rpn_corr=0.999072:rpn_rms=3.79369e-06
483:440837063:CLMR:600 mb:rpn_corr=0.996896:rpn_rms=4.37183e-06
484:440992321:ICMR:600 mb:rpn_corr=0.988521:rpn_rms=8.98694e-09
...

@jaymes-kenyon I would think the above changes are expected. You may look at RRFS tests at


- With your changes at /scratch1/NCEPDEV/stmp2/Wen.Meng/fv3r_2023062800 on Hera.
- Without your changes (baseline) at /scratch2/NCEPDEV/ovp/Wen.Meng/ncep_post/post_regression_test_new/data_out/fv3r

Please let me know if you see issues from my testing.

jaymes-kenyon · 2023-09-25T13:33:56Z

Hey @WenMeng-NOAA — These tests look good to me. All of the changes that were found would be expected. Thanks for the update!

WenMeng-NOAA · 2023-09-25T13:45:23Z

Hey @WenMeng-NOAA — These tests look good to me. All of the changes that were found would be expected. Thanks for the update!

@jaymes-kenyon Thanks for confirming.

WenMeng-NOAA · 2023-09-25T13:55:50Z

The UPP RTs were completed on WCOSS2. There will be new baselines needed for gfs, fv3r, 3drtma and hafs with this PR.

hafs:
-rw-r--r-- 1 wen.meng emc  5293 Sep 25 13:02 fv3hafs_2022092800/HURPRS09.tm00.diff
3drtma:
-rw-r--r-- 1 wen.meng emc  6503 Sep 25 13:02 rtma_2023040400/PRSLEV00.tm00.diff
fv3r:
-rw-r--r-- 1 wen.meng emc  4005 Sep 25 13:02 fv3r_2023062800/PRSLEV10.tm00.diff
gfs:
-rw-r--r-- 1 wen.meng emc  8435 Sep 25 13:13 gfs.20190830/gfs.t00z.master.grb2f006.diff

WenMeng-NOAA · 2023-09-25T14:42:18Z

@FernandoAndrade-NOAA You may start the UPP RT on Orion. See the baseline changes above.
The new baselines on Hera can be copied from my testing at /scratch1/NCEPDEV/stmp2/Wen.Meng:
gfs.20190830/
rtma_2023040400/
fv3r_2023062800/
fv3hafs_2022092800/

FernandoAndrade-NOAA · 2023-09-25T16:18:41Z

@FernandoAndrade-NOAA You may start the UPP RT on Orion. See the baseline changes above. The new baselines on Hera can be copied from my testing at /scratch1/NCEPDEV/stmp2/Wen.Meng: gfs.20190830/ rtma_2023040400/ fv3r_2023062800/ fv3hafs_2022092800/

UPP RTs have completed successfully on Orion, changes in results are consistent with the ones mentioned above for 3drtma, fv3r, fv3hafs, and gfs

WenMeng-NOAA · 2023-09-25T16:52:33Z

This PR is ready for merging.

jaymes-kenyon added 3 commits September 6, 2023 19:43

Updates to prevent spurious supercooled rain water from appearing on …

0fa1e2a

…pressure levels

Updates to comments/documentation; fix to continuation line

777fd09

Fix typo in comments

4253ec9

jaymes-kenyon mentioned this pull request Sep 6, 2023

Pressure-level interpolation may produce artificial supercooled rain water #778

Closed

WenMeng-NOAA linked an issue Sep 6, 2023 that may be closed by this pull request

Pressure-level interpolation may produce artificial supercooled rain water #778

Closed

(1) Adding "IF MODELNAME" checks

1153656

(2) Slight change to freezing-temp threshold (now 273.15 K) (3) Revised description in comments (4) Added print statements (temporary)

jaymes-kenyon added 3 commits September 8, 2023 18:31

Adding "MAX" functions, as are also used in the interpolation code (e…

ffff4e2

…arlier in MDL2P.f)

Restating freezing temperature as "TFRZ" (= 273.15 K, from params.F) …

263da42

…instead of local, hardcoded value

Removing print statements; finalizing PR

e1df9fd

jaymes-kenyon marked this pull request as ready for review September 11, 2023 19:16

jaymes-kenyon requested a review from WenMeng-NOAA as a code owner September 11, 2023 19:17

AndersJensen-NOAA reviewed Sep 11, 2023

View reviewed changes

sorc/ncep_post.fd/MDL2P.f Outdated Show resolved Hide resolved

jaymes-kenyon marked this pull request as draft September 11, 2023 21:45

jaymes-kenyon added 2 commits September 12, 2023 14:08

(1) Refactoring logic,

e702d73

(2) Applying adjustments to all hydrometeors

Updating entry in history log

a2b0b1b

jaymes-kenyon changed the title ~~Eliminate artificial supercooled rain water on pressure levels~~ Eliminate artificial supercooled water (rain and cloud) on pressure levels Sep 12, 2023

jaymes-kenyon marked this pull request as ready for review September 12, 2023 14:21

Re-wording comment

d695311

WenMeng-NOAA added RRFS GFSV17 labels Sep 22, 2023

Merge branch 'develop' of https://github.com/NOAA-EMC/UPP into pressu…

bbf4f47

…re_interpolation_update

WenMeng-NOAA reviewed Sep 22, 2023

View reviewed changes

sorc/ncep_post.fd/MDL2P.f Outdated Show resolved Hide resolved

jaymes-kenyon added 3 commits September 22, 2023 17:58

Adding bug fix from Wen

f879aa5

Adding checks for undefined arrays.

2425400

Fix typo

44d205c

WenMeng-NOAA added Ready for Review This PR is ready for code review. Baseline Change The baselines of the UPP regression tests are changed. labels Sep 25, 2023

WenMeng-NOAA requested a review from FernandoAndrade-NOAA September 25, 2023 13:56

FernandoAndrade-NOAA approved these changes Sep 25, 2023

View reviewed changes

WenMeng-NOAA approved these changes Sep 25, 2023

View reviewed changes

WenMeng-NOAA merged commit 998c2fe into NOAA-EMC:develop Sep 25, 2023

WenMeng-NOAA mentioned this pull request Sep 25, 2023

Updates in develop branch since 06/2023 #715

Closed

jaymes-kenyon deleted the pressure_interpolation_update branch September 25, 2023 16:55

WenMeng-NOAA added the HAFS label Mar 3, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Eliminate artificial supercooled water (rain and cloud) on pressure levels #771

Eliminate artificial supercooled water (rain and cloud) on pressure levels #771

jaymes-kenyon commented Sep 6, 2023 •

edited

Loading

jaymes-kenyon commented Sep 6, 2023

WenMeng-NOAA commented Sep 6, 2023

jaymes-kenyon commented Sep 7, 2023

jaymes-kenyon commented Sep 11, 2023

gthompsnJCSDA commented Sep 12, 2023

jaymes-kenyon commented Sep 12, 2023 •

edited

Loading

jaymes-kenyon commented Sep 15, 2023 •

edited

Loading

gthompsnJCSDA commented Sep 15, 2023

jaymes-kenyon commented Sep 15, 2023 •

edited

Loading

jaymes-kenyon commented Sep 20, 2023

WenMeng-NOAA commented Sep 20, 2023

WenMeng-NOAA commented Sep 22, 2023

jaymes-kenyon commented Sep 22, 2023

WenMeng-NOAA commented Sep 22, 2023

jaymes-kenyon commented Sep 22, 2023

WenMeng-NOAA commented Sep 22, 2023

WenMeng-NOAA commented Sep 24, 2023

jaymes-kenyon commented Sep 25, 2023

WenMeng-NOAA commented Sep 25, 2023

WenMeng-NOAA commented Sep 25, 2023

WenMeng-NOAA commented Sep 25, 2023

FernandoAndrade-NOAA commented Sep 25, 2023

WenMeng-NOAA commented Sep 25, 2023

Eliminate artificial supercooled water (rain and cloud) on pressure levels #771

Eliminate artificial supercooled water (rain and cloud) on pressure levels #771

Conversation

jaymes-kenyon commented Sep 6, 2023 • edited Loading

jaymes-kenyon commented Sep 6, 2023

WenMeng-NOAA commented Sep 6, 2023

jaymes-kenyon commented Sep 7, 2023

jaymes-kenyon commented Sep 11, 2023

gthompsnJCSDA commented Sep 12, 2023

jaymes-kenyon commented Sep 12, 2023 • edited Loading

jaymes-kenyon commented Sep 15, 2023 • edited Loading

gthompsnJCSDA commented Sep 15, 2023

jaymes-kenyon commented Sep 15, 2023 • edited Loading

jaymes-kenyon commented Sep 20, 2023

WenMeng-NOAA commented Sep 20, 2023

WenMeng-NOAA commented Sep 22, 2023

jaymes-kenyon commented Sep 22, 2023

WenMeng-NOAA commented Sep 22, 2023

jaymes-kenyon commented Sep 22, 2023

WenMeng-NOAA commented Sep 22, 2023

WenMeng-NOAA commented Sep 24, 2023

jaymes-kenyon commented Sep 25, 2023

WenMeng-NOAA commented Sep 25, 2023

WenMeng-NOAA commented Sep 25, 2023

WenMeng-NOAA commented Sep 25, 2023

FernandoAndrade-NOAA commented Sep 25, 2023

WenMeng-NOAA commented Sep 25, 2023

jaymes-kenyon commented Sep 6, 2023 •

edited

Loading

jaymes-kenyon commented Sep 12, 2023 •

edited

Loading

jaymes-kenyon commented Sep 15, 2023 •

edited

Loading

jaymes-kenyon commented Sep 15, 2023 •

edited

Loading