Idealized LES with Non-uniform Bottom BCs

[jgilbert]

Dear WRF Community,

I would like to set up an idealized WRF-LES case which includes spatially varying SST, ocean surface currents, and ocean surface roughness. My current understanding is that this may not be possible in an idealized WRF simulation today. So, my questions are:

1. Is such an idealized LES case possible without significant code modifications? Is it simply a matter of initializing the SST field? I'm guessing no, based on other posts on this forum.

2. Is it possible to initialize a real WRF-LES case with the same initial conditions and lateral boundary conditions as the idealized case? This may be easier than option 1. from a workflow/code modification standpoint, but I'm not yet familiar enough with WRF to know.

If anyone has suggestions for how I might go about setting up such an analysis, I would appreciate the help.

Thanks,
John

 

[Ming Chen]

John,
I am sorry to tell that both options you proposed are not practical.
For options I, it will involve huge amounts of code changes. The default LES case is designed to produce LES of free convective boundary layer. Surface heat flux is specified.
For option II, I expect the model will crash soon due to the inconsistency between initial condition and boundary forcing. Such a design is physically unreasonable.

 

[jgilbert]

Ming,

Thanks for the response. To be honest, I had expected option 1 to be a non-starter. However, I don't believe option II is physically unreasonable, as there are other coupled atmosphere-ocean models that have used a similar approach when validating their models. That said, I fully appreciate what you are saying about the sensitivity of the model to consistent initial and boundary conditions.

Let's assume, for the sake of argument, that such a simulation would run. Is it possible to initialize an idealized case (i.e. an analytically specified profiles of velocity, temperature, etc.) using the real case workflow (real.exe -> wrf.exe) so that I can incorporate spatially varying SST in the model?

I'll also say, that my reason for wanting to do this is to simplify the problem so that I can more easily characterize the effects of these bottom forcings.

Thanks for the help,
John

 

[Ming Chen]

John,
I am not sure whether it works. LES requires high resolution run, which makes it an issue how to produce boundary forcing from coarse-resolution data. Nesting could be a solution, but the lateral forcing that doesn't contain any eddy structures may still damage the fine-grid simulation.
Anyway this is just my preliminary idea, which may not be correct. Please try and keep us updated for progress. I guess many people will b interested in this topic.

 

[jgilbert]

Ming,

Thanks again for taking the time to respond. I feel like we might be talking past one another here, so let me ask a two different questions.

1. What are the minimum number of meteorological fields WRF requires to run a real case? This information might be available in existing documentation and I just missed it.

2. What do you think the best approach would be to modify initial conditions in WRF? Is it better to edit Metgrid files, wrfinput_*, wrfbdy_*, or possibly edit one or more of the *.F files directly and use input soundings? Really, I'm looking for a concrete place to start my digging.

[Update]: Put simply, I'm looking to set up a canonical flow over a flat plat profile and include spatially varying bottom boundary conditions.

Thanks,
John

 

[Ming Chen]

John,

(1) The following variables are required for running real-data case:
Mandatory:
3d upper air: horizontal winds, temperature, relative humidity, geopotential height
3d soil: soil temperature and soil moisture
2d fields: surface pressure, sea-level pressure, land mask, Skin temperature, 2-m T & RH,, 10-m U and V
Optional (but desirable):
2d fields: topography elevation of input data, SST, sea-ice

(2) I am not sure how to only modify initial condition, while keep the boundary forcing to be consistent with the modified wrfinput. We usually don't recommend such applications.

 

[zhangzhida]

John,

Have you found a good solution? I have the same problem as you.

Looking forward to your reply, I hope you can provide some suggestions on this study.

zhida

 

[Ming Chen]

LES simulation requires high-resolution with grid intervals smaller than 100m. SST and ocean currents usually have pretty coarse resolution and we won't expect distinct changes within a short distance. In this case, unless LES domain covers large enough area, SST variation should be small within the LES domain. Lots of coding work will be involved to couple ideal LES case with varying SST and ocean currents.
Due to limited human power, this is not our top priority at NCAR. However, we do welcome code contributions to make this option work.