Sudden termination of atmosphere model

[518er2005]

Dear all, I am trying to simulate a tropical cyclone case using ERA-5 as initial condition, but each time the atmosphere model gets terminated with the error message, "
Beginning MPAS-atmosphere Error Log File for task 19 of 24
Opened at 2023/04/12 02:59:42
----------------------------------------------------------------------

CRITICAL ERROR: NaN detected in 'w' field.
Logging complete. Closing file at 2023/04/12 02:59:42. However, if I run the simulations using FNL data as initial conditions, there isn't any error and the model runs properly. If anyone can help me resolve the issue while using the ERA-5 data as initial conditions, it will be greatly appreciated. I am attaching the "Namelist" along with the log file for a better understanding.

 

[Briggs1]

Dear all, I am trying to simulate a tropical cyclone case using ERA-5 as initial condition, but each time the atmosphere model gets terminated with the error message, "
Beginning MPAS-atmosphere Error Log File for task 19 of 24
Opened at 2023/04/12 02:59:42
----------------------------------------------------------------------

CRITICAL ERROR: NaN detected in 'w' field.
Logging complete. Closing file at 2023/04/12 02:59:42. However, if I run the simulations using FNL data as initial conditions, there isn't any error and the model runs properly. If anyone can help me resolve the issue while using the ERA-5 data as initial conditions, it will be greatly appreciated. I am attaching the "Namelist" along with the log file for a better understanding.

Hello, everyone

I'm sorry to hear about the error you're encountering in your simulation. Without the specific details of your namelist and log file, it is challenging to provide a precise solution. However, NaN (Not a Number) errors in the 'w' field often indicate numerical instability or issues with the initial conditions.

Here are a few suggestions to consider:

1. Check the compatibility: Ensure that the ERA-5 data you are using as initial conditions is compatible with the atmospheric model you are using. Verify the format, resolution, and variable names to ensure they match the requirements of the model.
2. Data preprocessing: Perform thorough data preprocessing to eliminate any missing or corrupted values in the ERA-5 data. Additionally, check for inconsistencies or anomalies in the data that could cause instability during the simulation.
3. Review namelist settings: Double-check the namelist settings to ensure they are correctly configured for using ERA-5 data as initial conditions. Pay attention to any specific settings related to the 'w' field or the initialization process.
4. Grid and resolution: Verify that the grid and resolution used in the atmospheric model are suitable for handling the ERA-5 data. Mismatches in grid settings can lead to interpolation errors and inconsistencies.
5. Consult documentation or support: Review the documentation of the atmospheric model or contact the support team for assistance. They may be able to provide specific guidance based on the model and error message you are encountering. Pay by Plate Ma


By carefully examining these aspects and making the necessary adjustments, you may be able to address the NaN error when using ERA-5 data as initial conditions.


Best regard,
Beiggs

 

[518er2005]

Thanks for the reply,
I have already attached the namelist and log file.

 

[mgduda]

Apologies for not replying long ago! In case you haven't already resolved this issue, please don't hesitate to follow up and I'll be glad to help.

 

[518er2005]

Thanks for the reply. The issue still remains

 

[mgduda]

Would you be willing to try the simulation again using MPAS v8.0.0? Since the simulation was successful out beyond a day and a half, it may not necessarily be the ICs that are at fault, but perhaps a growing mismatch between the model solution and the LBCs derived from ERA5. There was one change (see MPAS-Dev/MPAS-Model PR #1036) that was introduced in the 8.0.0 release to improve the robustness of regional simulations, and perhaps that could be helpful here?

 

[518er2005]

Ya, I will try and let you know. Also, when I checked "namemelist.atmosphere", the option "config_len_disp " was not available. Is this option no longer required or is it a mistake in the namelist.atmosphere.

 

[mgduda]

Regarding the 'config_len_disp' option, with MPAS v8.0 it's no longer necessary to specify this length scale if you're using new meshes from the mesh download page, since these new mesh NetCDF files have a variable 'nominalMinDc' that provides this length scale. If you're curious, you can see the corresponding code changes to read and use the 'nominalMinDc' value in MPAS-Dev/MPAS-Model PR #1079. However, if you're using old meshes (downloaded before the v8.0.0 release), you will still need to add config_len_disp to your namelist.atmosphere file.

 

[518er2005]

Yes, I came to know about that and also after using MPAS v8. 0 the simulations are running properly. I just need one more help, is their any code or way i can directly convert the "diag" or "history" files directly to netcdf.I know by using "convert_mpas" one can interpolate and get the outputs in desired lat lon grid, but i intend to directly use those outputs so want to directly convert them to regular lat_lon grid instead of interpolation.

 

[mgduda]

sdfsd

Yes, I came to know about that and also after using MPAS v8. 0 the simulations are running properly. I just need one more help, is their any code or way i can directly convert the "diag" or "history" files directly to netcdf.I know by using "convert_mpas" one can interpolate and get the outputs in desired lat lon grid, but i intend to directly use those outputs so want to directly convert them to regular lat_lon grid instead of interpolation.

Is this the same question that you've posted about here: convert to regular lat lon grid ? Could you please follow up in that thread if the convert_mpas tool doesn't do what you're trying to accomplish?