High Resolution (dx/dy ~200m) Simulations and PBL Schemes

[TyBuck]

Hi all,

I recently computed a four domain (25km, 5km, 1km, 200m) simulation and received realistic output. However, after looking through the guidelines I came across the following:

"PBL schemes can be used for most grid sizes when surface fluxes are present; however, at grid size dx << 1 km, this assumption breaks down. To get around this, you can use 3d diffusion instead of a PBL scheme (coupled to surface physics). This works best when dx and dz are comparable."

My simulation was computed with:

bl_pbl_physics = 2, 2, 2, 2,
km_opt = 4, 4, 4, 4,

Yet it appears I should have run it with km_opt = 3 and no pbl scheme for the highest resolution domain.

What kind of impacts might this have had on my simulation? Is it still reasonable?

Thanks,
Ty

 

[rdumais63]

Hi Ty,


That is a turbulence "grey area" at 250m grid spacing. This is much more the case as the atmosphere gets convectively unstable. Under convective conditions I would say that you may want to use bl_pbl_physics=0 (treat the nest as an LES with 3D diffusion option for turbulence). Pay attention to your diff_opt and km_opt settings for this nest as well. Watch setting of sfs_opt too for LES nest. Also set epssm to a little higher value (say 0.5 to 0.9) if you have steep slopes in your sub-km nests. Only use a cu_phys option on at most your outer two nests. Radt is another consideration- I might set namelist to 25, 5, 1,1 although there has always been debate on this forum as to whether radt for all nests is really limited to what is set for radt of outermost nest. Guidance always has suggested 1min per km grid spacing for a nest radt value but this has always seemed to conflict the other thinking I just mentioned for radt. In really deep convective PBLs even 1 km might have a few issues using a PBL scheme but it is clearly not ideal for LES treatment. Maybe a scale-aware hybrid PBL option like Shin-Hong scheme might be useful for this kind of setup of 25-5-1-250m?

Now if you restrict to night runs under neutral/stable conditions you can probably get away using PBL schemes even to 250m grid spacing. Certainly the scale-aware scheme is ok. LES may be ok but not as clear-cut as under convective as largest energy-containing turbules are much smaller typically at night. Under stable stratification you might be able to get away testing placing lowest model level closer to ground than you could during daytime convective.

Watch your aspect ratio (deltaz/deltax,y) remaining less than/equal to 1 at least in lowest 1-2 km agl but not much more than 2 or 3 most of the rest of the column. If model top is 50 mb, you may need to stretch this a little bit more (maybe ratio of 4-6) above about 150 mb unless you add additional sigma levels up near model too.

 

[rdumais63]

This pdf put together by Dr. Jimy Dudhia at NCAR should be helpful for approaching LES. There is also a relstively new vertical nesting option in WRF you can experiment with. I have another forum topic open right now trying to iron out a few issues in the boundary zones I have been having with it.

 

[Ming Chen]

Hi Ty,
I would suggest that you try the option 3DTKE for this case. Note that you should set
bl_pbl_physics = 0, 0, 0, 0
km_opt = 5, 5, 5, 5
diff_opt = 2, 2, 2, 2
sf_sfclay_physics = 1, 1, 1, 1
Please try nd let me know whether this options can give you better results.