Why do met_em files include interior points when only boundaries are needed after Time=0?

[hzanganeh]

Hello all,


I'm running a 36-year WRF simulation with 6-hourly met_em output, which generates ~5-6 TB of intermediate files. This raises a storage question:


Why do met_em files contain interior grid points for all timesteps when they're only needed for the initial condition (T=0)?


After the first time step, WRF only needs lateral boundary conditions, yet every met_em file contains the full 3D domain. For long climate simulations, this creates massive storage overhead.


Questions:

1. Is there a way to generate met_em files with only boundary points after T=0?
2. Or can ungrib/metgrid be configured to skip interior points for t>0?
3. Has anyone implemented a workflow to delete interior data from met_em files before running real.exe?

I understand real.exe needs complete fields for vertical interpolation and consistency checks, but it seems inefficient to store full domains for what becomes pure boundary forcing.


Any insights or workarounds would be appreciated!


Best,

 

[Ming Chen]

Please see my answers below:

Hello all,


I'm running a 36-year WRF simulation with 6-hourly met_em output, which generates ~5-6 TB of intermediate files. This raises a storage question:


Why do met_em files contain interior grid points for all timesteps when they're only needed for the initial condition (T=0)?

This is not only for convenience, but also for a roughly comparison of WRF simulations with its large scale foricng.

After the first time step, WRF only needs lateral boundary conditions, yet every met_em file contains the full 3D domain. For long climate simulations, this creates massive storage overhead.


Questions:

1. Is there a way to generate met_em files with only boundary points after T=0?

This is feasible. You can modify relevant code to save data only for the buffer zone.

1. Or can ungrib/metgrid be configured to skip interior points for t>0?

ungrib processes large-scale forcing data and it has no idea where the WRF domain is located. Hence it is more practical to modify mergrid.

1. Has anyone implemented a workflow to delete interior data from met_em files before running real.exe?

I am not aware of such information.

I understand real.exe needs complete fields for vertical interpolation and consistency checks, but it seems inefficient to store full domains for what becomes pure boundary forcing.

WRF has been applied to various projects and these projects may have different requirements. It is hard for WRF developers to consider all these various situations. That is why WRF users usually modify codes based on their specific needs.

Any insights or workarounds would be appreciated!


Best,