Best-performing gray-zone parametrizations (complex terrain context)?

[Pechudin]

This is a question for people who have a lot of experience in running borderline-LES simulations (~100-200 m resolution), preferably in complex terrain.

So, I have been running 200-m resolution simulations (LES, km_opt=2 or 3, bl_pbl=0) over complex terrain (eastern Adriatic coast), where terrain drops from 1.5 km to 0 m in ~10 km. I did this to induce shear instabilities in the lee of the mountain range, since that is what I want to study.

I was succesful in this, and the results seem promising. However, I have noticed several tendencies of these results:

• overestimation of near-ground wind speed (by as much as 50% in some cases)
• seemingly unrealistic superadiabatic lapse rates in a shooting flow with very high wind speed (20 m/s), which should be well-mixed instead (summer case)
• perhaps too strong KHI between the shooting flow and the lee stagnation zone (caused by breaking mountain wave)

Now, this is probably due to insufficient mixing by the LES scheme (LES assumes inertial range, which 200 m is not). The discrepancies above are less bad during the summer, presumably because of the CBL eddies being larger than the wintertime BL eddies.

The question I have now is the following: I have tried using the Shin-Hong scheme (bl_pbl=11) and the 3d_tke option (km_opt=5), both suited for the gray-zone simulations (as per the manual). I have tried them, and noticed improvements in mean speed (especially in 3d_tke scheme), but the instabilities are too smoothed out, and so are basically absent. Additionally, the more realistic speed with the 3d_tke seemed to result from overall lower speed in the BL flowing in from the 1-km mesoscale domain (meaning the YSU scheme, which is what I use in the mesoscale domains, overestimated the inflowing mean speed; YSU is the basis of the Shin-Hong scheme, so that explains the weaker speed improvement with this scheme).

Therefore, what are best practices when employing these parametrizations? Which other parameters should be paired up with the Shin-Hong scheme? What about 3d_tke scheme? What about the static fields? I have extensive notes on sensitivity tests, so if that is helpful I can contribute. I will run some more simulations with a lower resolution to compare the results.

 

[kwerner]

Hi,
Apologies for the delay in response. Per our physics expert, it looks like you are trying the right things. With a PBL scheme, topo_wind = 1 may help wind in the outer domain. If the topography is very smooth, maybe some elevation perturbations could be added to generate turbulence. Is the nest large enough to generate eddies downstream of the nest boundary?

 

[Pechudin]

Well, the 200 m nest is 250 x 250 points in size, or 50 km on each side. For the quasi-LES (no PBL option) and Shin-Hong option, instabilities develop (especially for the LES case, where they develop after first 30% of the domain). Nothing for the 3D-TKE option.

 

[kwerner]

Are you using km_opt=4 for Shin-Hong? That's what you should do, if you aren't already. If so, will you attach your namelist.input file so we can see your settings? Thanks!

 

[Pechudin]

Well, I used km_opt=4 for the outer 3 domains (9, 3, 1 km), but not the inner domain (200 m; here I used km_opt=2). Why should km_opt=4 be used with the Shin_Hong scheme? I haven't seen what the reason was.

Additionally, I have read more literature and it seems the gray zone parametrizations were optimized for the convective BL, not downslope windstorms in complex terrain. Also, both rely on a well-defined PBL height to scale the subgrid terms, which is ill-defined in these conditions. It certainly is interesting how these two things will interplay. I intent to research more, especially if I can somehow extract the behaviour of resolved and unresolved terms and where the algorithms consider the PBL top to be.