NoahMP restart runs likely not b4b identical

[climbfuji]

I have reason to assume that restart runs using NoahMP are not b4b identical, because (at least)

• the precipitation rates from the previous time step are not saved to the restart files

This requires a change in fv3atm, file gfsphysics/GFS_layer/GFS_restart.F90, similar to what was done for RUC LSM (but for twice as many variables?):

Sfcprop(nb)%raincprv   
Sfcprop(nb)%rainncprv  
Sfcprop(nb)%iceprv     
Sfcprop(nb)%snowprv    
Sfcprop(nb)%graupelprv 
Sfcprop(nb)%draincprv  
Sfcprop(nb)%drainncprv 
Sfcprop(nb)%diceprv    
Sfcprop(nb)%dsnowprv   
Sfcprop(nb)%dgraupelprv

Whether this is sufficient to achieve b4b identical results through restarts for NoahMP or not I cannot say.

[climbfuji]

@grantfirl @HelinWei-NOAA for your information / for you to look at

[grantfirl]

@climbfuji Thanks. I was aware of this, although I didn't see any of the variables needed for NoahMP restarts in GFS_restart, so I assumed that this functionality didn't exist and/or is a work in progress at EMC. Since this isn't CCPP-related, I don't view this as DTC responsibility.

[climbfuji]

@climbfuji Thanks. I was aware of this, although I didn't see any of the variables needed for NoahMP restarts in GFS_restart, so I assumed that this functionality didn't exist and/or is a work in progress at EMC. Since this isn't CCPP-related, I don't view this as DTC responsibility.

I agree, but I wanted to have it documented somewhere.

[HelinWei-NOAA]

If you turn on Noah MP option, those variables needed for restarting Noah MP will be saved in GFS_restart title files.

…

On Fri, Dec 6, 2019 at 10:53 AM Dom Heinzeller ***@***.***> wrote: @climbfuji <https://github.com/climbfuji> Thanks. I was aware of this, although I didn't see any of the variables needed for NoahMP restarts in GFS_restart, so I assumed that this functionality didn't exist and/or is a work in progress at EMC. Since this isn't CCPP-related, I don't view this as DTC responsibility. I agree, but I wanted to have it documented somewhere. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#367?email_source=notifications&email_token=ALPHKYAGQUFQZYKSQUZJCVLQXJYRJA5CNFSM4JW2L7I2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEGEQN7Q#issuecomment-562628350>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALPHKYABMDJ4D56KMHE5N7DQXJYRJANCNFSM4JW2L7IQ> .

[ligiabernardet]

@HelinWei-NOAA (separate email conversation) believes the Noah-MP restarts are working.
@climbfuji @grantfirl Is it your understanding that runs with Noah-MP are not b4b when restarted? Do we need to run a test to find out (att @linlin-pan)?

[panll]

I can run a test on that @climbfuji @ligiabernardet @HelinWei-NOAA

[grantfirl]

My understanding was that it was not working, but maybe it has been fixed by someone in the last few months?

[climbfuji]

It is possible that this was fixed since we tested last time (beginning of December 2019).

[HelinWei-NOAA]

It is possible. We fixed the precip forcing issue which definitely has an impact on reproducibility.

…

On Fri, May 15, 2020 at 12:47 PM Dom Heinzeller ***@***.***> wrote: It is possible that this was fixed since we tested last time (beginning of December 2019). — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#367 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALPHKYCFOF3FGE5ZALLTGIDRRVWZNANCNFSM4JW2L7IQ> .

[panll]

I tested cold start run and restart run, and the results are not B4B match. Any comments/suggestions? Thanks!

[climbfuji]

This was fixed for the SRW App release/public-v2 branch of the ufs-weather-model / fv3atm. It turns out that making the changes suggested above was sufficient, provided that the code gives b4b identical restarts for runs with Noah LSM already. See NOAA-EMC/fv3atm#246 and ufs-community/ufs-weather-model#417.

[panll]

Thanks, Dom!

[climbfuji]

Fixed for fv3atm develop in NOAA-EMC/fv3atm#249