Running MYNN on coarse and LES on fine grid - impossible or user error?

[Pechudin]

Hello,

I am running WRF to study downslope windstorms, and would like to use the MYNN (or other higher-level closure schemes) for the coarse grid (1 km or larger), as they've proved the best for this kind of application, and LES for a finer grid (say 100 m). However, when attempting to choose bl_pbl_opt = 5 and 0 (for LES domain), REAL part of the workflow fails with the error saying that this pairing is impossible. The WRF version I was using is 4.3.3 or 4.5.2.

Now, the GitHub page says the following about 4.3.2:

A test was introduced to stop the incorrect combination of particular PBL schemes on coarser grids with the LES PBL option selected on a finer grid. Previously, this problem caused an inconsistent number of variables on the CG and FG, which caused segmentation faults when trying to do feedback or advection of unavailable fields.

I understood that as "in this version, we temporarily disabled this particular coupling". In the next version however (4.3.3), the log says:

The v4.3.2 release contained an error in the commit that addressed the situation of segmentation faults when running a PBL on a coarse grid and no PBL on the fine grid. If a user is running the option to get the time series output, then the model will always report that there are PBL + LES issues if the user is choosing the MYNN or EEPS PBL schemes. This bug is fixed in the v4.3.3 release.

To me, this means that it should now be possible to use this pairing. But it does not seem to work, so is this my problem or is this a problem that will persist in perpetuity? Thank you in advance for your response,

Petar Golem

EDIT: I forgot to mention that I am using tslist to extract time-step output as well - is that the culprit here? I suppose I'll test it forthwith.

 

[Pechudin]

Update - no, it's not the tslist. Unsure what's the solution besides reverting to previous WRF version (pre-4.3)?

 

[kwerner]

Hi,
Can you attach your namelist.input file, as well as your error file(s) - if you have several rsl* files, you can package those together in a .zip or .tar file and just attach that with the namelist. Thanks!

 

[Pechudin]

Thank you for the response!

Since the problem is in the REAL part of the process, there is only one rsl file. I have attached it here with the namelist.input.

 

[kwerner]

The error you're seeing is simply because using an LES PBL option (i.e., setting it to 0 in the fine-resolution domains) does not work properly with the MYNN pbl options. The reason this error doesn't show up prior to V4.3 is simply because the bug wasn't discovered until after that code was released. That check was added to the code to protect against doing something that would likely lead to bad results. Unfortunately if you want to use LES PBL, your only option is to choose an non-MYNN PBL for the coarser domains.

 

[Pechudin]

That's a bummer. I've been using YSU (1st-order K-profile closure) instead. I assume using ndown so the simulations are not ran concurrently would also not work (and, as you say, would produce bad results so what would be the point)? Is there a paper on why this is the case?

In either case, glad to know this is not an user error.

 

[kwerner]

I'm not aware of any literature on the topic, but that doesn't mean there isn't any. You could try searching to see if you can find anything. If you're interested, you can take a look at the commit with the change to the code in our GitHub repository.

 

[Ming Chen]

I would suggest using the SMS-3DTKE scale-adaptive LES/PBL scheme, which can well address PBL processes at various resolutions.

 

[Pechudin]

Has SMS-3DTKE been compared vs MYNN in literature?

EDIT: Nevermind, I found some papers so I'll see. Mostly tropical storm applications and over urban areas though.