restrict post computation from undefined points

[junwang-noaa]

Currently in POST, there is no restriction to keep the computation only happening on valid points with reasonable value. The undefined value are participating in the computation in many places which causes wrong value in post products. One example is when regional output domain is larger than the forecast computational domain. FillValue is put on the extra domain for all the fields, however those fields are still involved in post computation and causes issues.

[HuiyaChuang-NOAA]

After Wen tested and merged Jun's UPP updates, Jesse will take over and scan through all UPP subroutines to make sure derivations are only done when none of input has missing values.
Jesse will also follow Jun's advice to test use of negative value as missing value instead, e.g., -1.e10
Jesse's task is tracked and documented in
https://vlab.ncep.noaa.gov/redmine/issues/87352

[JesseMeng-NOAA]

Wen, please let me know when you are ready. Thanks, Jesse

…

On Wed, Jan 27, 2021 at 3:21 PM HuiyaChuang-NOAA ***@***.***> wrote: After Wen tested and merged Jun's UPP updates, Jesse will take over and scan through all UPP subroutines to make sure derivations are only done when none of input has missing values. Jesse will also follow Jun's advice to test use of negative value as missing value instead, e.g., -1.e10 Jesse's task is tracked and documented in https://vlab.ncep.noaa.gov/redmine/issues/87352 — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#240 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALXNRCNJGUBDZHFLFZHJHLTS4BYV5ANCNFSM4UWURXAA> .

[WenMeng-NOAA]

@JesseMeng-NOAA The UPP branch was updated with the PR #255 for regional in-line post.

[WenMeng-NOAA]

@JesseMeng-NOAA Any updates on this issue?

[JesseMeng-NOAA]

@JesseMeng-NOAA Any updates on this issue?

working on it

[JesseMeng-NOAA]

pull request submitted
#306

[WenMeng-NOAA]

On Thu, Apr 29, 2021 at 4:21 PM Wen Meng - NOAA Affiliate wen.meng@noaa.gov wrote:
Jesse,

Would you run the regression tests before and after your changes, and compare all fields in the datasets from all applications to
ensure no fields having all 0 values changes or from all 0 to valid values changes? It seems the utility cmp_grib2_grib2_new might fail to
compare two fields either one with all 0 values.

Thanks,

Wen

[WenMeng-NOAA]

On Fri, Apr 30, 2021 at 2:47 PM Jesse Meng - NOAA Affiliate jesse.meng@noaa.gov wrote:
All output checked.
The fields show up in *.diff are due to one of the following reasons,

1. the entire field is undef in both rt base file AND the test file
2. boundaries of regional models
3. hrrr and rap pe test, Wen has mentioned it's associated with pending GSD unification
4. surface ALBEDO, fixed bug in current FIXED.f for uninitialized GRID1

Resolved.

[WenMeng-NOAA]

On Fri, Apr 30, 2021 at 3:05 PM Wen Meng - NOAA Affiliate wen.meng@noaa.gov wrote:
Jesse,

My understanding is that you have checked all datasets from gfs, gefs, rap, hrrr, hafs, fv3lam. There were no the following
situations:

1. values in fields are changed from all 0 (undefined or invalid ) into valid.
2. values in fields are changed from in valid into all 0 (undefined or invalid).

Thanks,

Wen

[WenMeng-NOAA]

Correct.

Dr. Jesse Meng
IMSG at NOAA/NWS/NCEP/EMC
5830 University Research Ct. Room 2037
College Park, MD 20740
email: jesse.meng@noaa.gov