search

ufs-community / ufs-weather-model

UFS Weather Model
Other
140 stars 247 forks source link

p7b not reproduceable itself #684

Closed jiandewang closed 3 years ago

jiandewang commented 3 years ago

Description

It is found that p7b is not reproduceable by itself. This is confirmed from Xu Li, Weizhong and Jiande's independent tests. Attached is the tmpsfc difference between two runs from sfcf00.nc Tsfc

To Reproduce:

checkout https://github.com/ufs-community/ufs-weather-model/tree/release/P7b compile and run any one of the p7b case twice compare the results

jiandewang commented 3 years ago

a good start to track this issue is turn back to p7a to see whether same issue exists or not. However p7a is no longer available to run as "input_data_20210608" has been deleted. A question for EIB group: is it fine to use the existing input data before 20210608 to run p7a ? what's new in 20210608? will the new stuff have any impact to p7a setting ?

DeniseWorthen commented 3 years ago

There is a directory prototype7-input-data-20210608. In that directory, the MERRA2 input has been corrected.

SMoorthi-emc commented 3 years ago

My PR should fix this Moorthi

On Fri, Jul 9, 2021 at 11:36 AM jiandewang @.***> wrote:

Description

It is found that p7b is not reproduceable by itself. This is confirmed from Xu Li, Weizhong and Jiande's independent tests. Attached is the tmpsfc difference between two runs from sfcf00.nc [image: Tsfc] https://user-images.githubusercontent.com/14180427/125102825-b70b5480-e0a9-11eb-99b8-071d6d952dc2.png To Reproduce:

checkout https://github.com/ufs-community/ufs-weather-model/tree/release/P7b compile and run any one of the p7b case twice compare the results

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ufs-community/ufs-weather-model/issues/684, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALLVRYQPXEPY6TENZZZBOADTW4JQLANCNFSM5AC7WHMQ .

-- Dr. Shrinivas Moorthi Research Meteorologist Modeling and Data Assimilation Branch Environmental Modeling Center / National Centers for Environmental Prediction 5830 University Research Court - (W/NP23), College Park MD 20740 USA Tel: (301)683-3718

e-mail: @.*** Phone: (301) 683-3718 Fax: (301) 683-3718

jiandewang commented 3 years ago

@SMoorthi-emc : thanks for the good news

jiandewang commented 3 years ago

@DeniseWorthen thanks for the information. I am going to try p7a with that fixed file dir.

jiandewang commented 3 years ago

my two p7a runs with "prototype7-input-data-20210608" got identical results. So the issue must be related to the new features in p7b.

DeniseWorthen commented 3 years ago

To eliminate Mushy thermo as a cause, you can comment out these two settings:

export KTHERM=2 export TFREEZE_OPTION='mushy'

That will return to the default (old) settings.

jiandewang commented 3 years ago

@DeniseWorthen : thanks for the info., will try

SMoorthi-emc commented 3 years ago

As I wrote before it is fixed in my PR and the bug in ATM. Moorthi

Sent from my iPhone

On Jul 9, 2021, at 4:07 PM, Denise Worthen @.***> wrote:

To eliminate Mushy thermo as a cause, you can comment out these two settings:

export KTHERM=2 export TFREEZE_OPTION='mushy'

That will return to the default (old) settings.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

jiandewang commented 3 years ago

@SMoorthi-emc OK will wait for your PR's merge. Thanks

DeniseWorthen commented 3 years ago

@SMoorthi-emc Isn't this gcycle fix the only change that is in p7b but not in FV3 develop right now? But the current develop branch does reproduce itself; is there a bug in the gcycle fix that P7b has which will get fixed your PR?

SMoorthi-emc commented 3 years ago

This has nothing to do with cycle or sfcsub. Does the standalone fv3 reproduce with threading and fractional grid? If yes, maybe I am wrong. Nevertheless, code in my PR reproduces coupled or uncoupled and I think I am using mushy. I will double check later. Moorthi

Sent from my iPhone

On Jul 9, 2021, at 4:48 PM, Denise Worthen @.***> wrote:

@SMoorthi-emc Isn't this gcycle fix the only change that is in p7b but not in FV3 develop right now? But the current develop branch does reproduce itself; is there a bug in the gcycle fix from your PR?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

DeniseWorthen commented 3 years ago

@SMoorthi-emc Jiande reminded me we are running the P7 w/ threads so that explains my confusion.

SMoorthi-emc commented 3 years ago

Denise, It turned out that I was not using mushy thermodynamics. So, I made three new runs with mushy thermodynamics and the results are reproducing, including restart run (all three runs are with two threads, including the ocean).

jiandewang commented 3 years ago

I turned off mushy, results not reproduceable, so mushy is not the cause

bingfu-NOAA commented 3 years ago

I had one issue related to NSST restart, might be not related to this one. But I suggest turn off NSST spinup and give a try? change 2,1,0,0,0 to 2,0,0,0,0.