MPAS-A stops and raises error with Noah MP land surface model option on global variable resolution 15km-3km mesh
- Thread starter jkukulies
- Start date
Hi Julia,
We found the issue that caused MPAS crash.This is because, SWE from ERA5 is up to 10m, which leads to more than 30m of snow depth. As a result, NoahMP snow package becomes physically unreasonable and yields TSK that is totally wrong. TSK subsequently affects radiation, eventually leading to model crash.
There are 2 temporary solutions to this issue:
(1) exclude snow from ERA5,
or
(2) cap SWE in NoahMP
Michael suggests that we add limitations to snow in init_atmopshere. This is the approach I like, because it will address similar issues caused by wrong information in MPAS input data, --- while it is hard for us to explore each individual forcing data to detect wrong information, this approach potentially can correct wrong information at the right beginning. I suppose this new feature will be included in MPASV8.3.0, which will be released before the MPAS/WRF workshop.
If you din't feel like to wait until the release of MPASV8.3.0, then you can modify the code:
MPAS-Model/src/core_atmosphere/physics/physics_noahmp/drivers/mpas/NoahmpInitMainMod.F90, in which please find the following piece of code (`~lines 53-65):
And change the first line from do i = its, its to do i = its, ite. Then recompile and run your case again.
Please let me know if you still have any issues. Thanks.
We found the issue that caused MPAS crash.This is because, SWE from ERA5 is up to 10m, which leads to more than 30m of snow depth. As a result, NoahMP snow package becomes physically unreasonable and yields TSK that is totally wrong. TSK subsequently affects radiation, eventually leading to model crash.
There are 2 temporary solutions to this issue:
(1) exclude snow from ERA5,
or
(2) cap SWE in NoahMP
Michael suggests that we add limitations to snow in init_atmopshere. This is the approach I like, because it will address similar issues caused by wrong information in MPAS input data, --- while it is hard for us to explore each individual forcing data to detect wrong information, this approach potentially can correct wrong information at the right beginning. I suppose this new feature will be included in MPASV8.3.0, which will be released before the MPAS/WRF workshop.
If you din't feel like to wait until the release of MPASV8.3.0, then you can modify the code:
MPAS-Model/src/core_atmosphere/physics/physics_noahmp/drivers/mpas/NoahmpInitMainMod.F90, in which please find the following piece of code (`~lines 53-65):
Code:
do i = its, its
if ( NoahmpIO%snow(i) < 0.0 ) NoahmpIO%snow(i) = 0.0
if ( NoahmpIO%snowh(i) < 0.0 ) NoahmpIO%snowh(i) = 0.0
if ( (NoahmpIO%snow(i) > 0.0) .and. (NoahmpIO%snowh(i) == 0.0) ) &
NoahmpIO%snowh(i) = NoahmpIO%snow(i) * 0.005
if ( (NoahmpIO%snowh(i) > 0.0) .and. (NoahmpIO%snow(i) == 0.0) ) &
NoahmpIO%snow(i) = NoahmpIO%snowh(i) / 0.005
if ( NoahmpIO%snow(i) > 2000.0 ) then
NoahmpIO%snowh(i) = NoahmpIO%snowh(i) * 2000.0 / NoahmpIO%snow(i) !snow in mm and snowh in m.
NoahmpIO%snow (i) = 2000.0 !cap snow at 2000 to maintain
!density.
endif
enddoAnd change the first line from do i = its, its to do i = its, ite. Then recompile and run your case again.
Please let me know if you still have any issues. Thanks.
Thank you so much for figuring this out, Ming! This is so helpful and I am glad that this change will also benefit other users and prevent potential issues with the input data.
Before I will re-run my simulations, I just have a couple of questions.
1. Will the snow cap that you suggest mean that instead of going over a certain threshold that makes the model crash due to unreasonable TSK, the snow just stays at a certain maximum? Could that also lead to unrealistic values although it prevents the model from crashing? I just want to make sure that I understand it right that this is the preferred option over replacing the ERA5 snow with other snow input data?
2. Will the code snippet that you suggest in practice do the same thing as in the new release with the only difference that the limitations to snow will be handled in init_atmosphere in the new release? Or would you suggest to wait until the new release for more appropriate handling of the snow input?
Thanks once again!!
Before I will re-run my simulations, I just have a couple of questions.
1. Will the snow cap that you suggest mean that instead of going over a certain threshold that makes the model crash due to unreasonable TSK, the snow just stays at a certain maximum? Could that also lead to unrealistic values although it prevents the model from crashing? I just want to make sure that I understand it right that this is the preferred option over replacing the ERA5 snow with other snow input data?
2. Will the code snippet that you suggest in practice do the same thing as in the new release with the only difference that the limitations to snow will be handled in init_atmosphere in the new release? Or would you suggest to wait until the new release for more appropriate handling of the snow input?
Thanks once again!!
Hi Julia,
Please see my answers below:
What i suggest is a temporary fix. If possible, I would suggest you wait for the release of MPASv8.3.0. Michael, Laura and Cenlin will get involved in this issue and hopefully we can find a more reasonable solution.
You are very welcome !
Please see my answers below:
This change only affects snow at the initial time, and it caps the maximum snow to 2000mm. Once the model starts, snow will be updated based on its physics package. However, I agree that this modification gives initial snow an artificial value, which doesn't really reflect the real-world condition. In this sense, you are right that we make such changes to prevent model from crashing, while we need to be cautious when interpreting the results.
What i suggest is a temporary fix. If possible, I would suggest you wait for the release of MPASv8.3.0. Michael, Laura and Cenlin will get involved in this issue and hopefully we can find a more reasonable solution.
You are very welcome !