Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

High resolution (gray-zone adjacent) simulations and gravity wave drag options

Pechudin

New member
Hello all,

I am currently running WRF simulations with the goal of resolving rather short wavelength (high-frequency) wave motions in the lee of a mountain range (wavelength of several km and period of 1-10 min); in the past these motions were attributed to KH instability between the shooting downslope windstorm flow and the stagnation zone above.

I am currently running a 9-3-1-0.333 km 4-domain nested (feedback on) simulation and have at last managed to get time series and spectra very similar to the observed ones, if of a smaller amplitude. However, I was only able to get these results if gwd_opt=1 was ON on ALL the domains. I have tried gwd_opt=1 on only 1, 2 and 3 domains and none did it for me, only when it was on the finest domain as well.

Now, my question is the following - the WRF guide says that gwd_opt is recommended ON for all domains, but as I understand it, gwd_opt is intended to parametrize subgrid orography wave drag effects (and blocking) ; surely on a 1 km and 333 m domain these effects should be resolved; in that case am I "double-counting" gravity wave drag? Also, the gwd_opt input data is coarse, 10m in resolution, which sometimes (above the ground) results in unsightly squares (picture included, left - terrain, right - U at level 12, around 800 m a.g.l. with dzbot=20 and dzstretch_s=1.05). Vertical level resolution has no effect on this, as 60 and 80 levels provided the same results.

What are your opinions? Should gwd_opt be on for ALL domains (no matter how small they are)? I know the guide says the gwd_opt=3 is recommended for < 1km resolutions. I have not found a study which argues either way. On the other hand, added wave drag should enhance low-level instabilities since the drag enhances wind shear, and to keep the simulation from exploding in complex terrain I had to add vertical damping. Perhaps these two effects cancel each other out partially. The added wave instabilities are not perpendicular to the wave sides so I don't think they are caused by the potential discontinuities at the edges of the squares.

There are other parameters which may impact the simulation; namely the terrain resolution on domains 1, 2, and 3 is the 30-arcsecond default dataset, while on domain 4 it is the ASTER 1-arcsecond dataset. I will run a simulation with gwd_opt=1 on all domains and no ASTER terrain and report back. I will also try the gwd_opt=3 option as well. Don't getme wrong, the result is exciting since it does reproduce the measured spectra well at the required frequencies, I just wonder if the result could be an artefact of using unsuitable parametrizations. Thank you very much in advance,

Petar Golem, MSc
 

Attachments

  • ncview.U_d04.ps
    1.4 MB · Views: 2
  • ncview.HGT_d04.ps
    810.1 KB · Views: 3
Okay, an update. The gwd_opt=3 option did not produce the high-wavenumber instabilities, instead it looked more like the simulations with the gwd_opt turned off. Also, ASTER terrain is not the "culprit", but it is definitely the gwd_opt=1 option when turned on for all domains.
 
Hi Pechudin
Thank you for the detailed description of this issue and the updates.
Yes I have seen those square patterns before with gwd_opt. One possible issue is that we don't have higher resolution GWD files (OA etc.) in geogrid. Anyway, I agree that you should not need gwd for such high resolutions. Maybe the surface flow-blocking part is the important factor.
I will talk to our expert about this issue and get back to you if we can find a solution.
 
Hi Pechudin
Thank you for the detailed description of this issue and the updates.
Yes I have seen those square patterns before with gwd_opt. One possible issue is that we don't have higher resolution GWD files (OA etc.) in geogrid. Anyway, I agree that you should not need gwd for such high resolutions. Maybe the surface flow-blocking part is the important factor.
I will talk to our expert about this issue and get back to you if we can find a solution.

Well, I am not sure if this is an issue? Yes, the squares are unphysical, and the GWD file resolution is the reason (the squares boundaries are exactly where the discontinuities happen in the image), but on the other hand when I do turn it on, WRF reproduces the waves at roughly the right time and at roughly the same wavelength as the measurements - kind of hard to accept that this is unphysical (in fact it makes me quite happy, as it means thta perhaps I can explain some observations in the measurements).
 
Okay, an update - using gwd_opt=1 in high-resolution (333 m) terrain is a problem. I am not sure if this was documented, as I could not find references to it, but what seemingly happens is that it "enforces an elevated critical level". Not sure how to say this, but in the picture there is almost an elevated line which follows the terrain well - too well in fact, and it does not move with time. The pictures are domain 3 (1 km dx) and domain 4 (333 m dx). They are not from the same simulation, but the settings were identical and the event simulated is the same, this is more to illustrate the problem.

I think this creates an upper boundary relatively close to the ground, which then creates spurious reflections and resonances, which is why my simulations showed increased energy in the high-frequency (1-10 min) part of the spectrum. Comments?

Still weird that the timing and the frequency of the motions was very similar to the measurements.

d03_wdir.jpg
d03_wspd.jpg
 
Another update; a senior of mine told me how the gwd_opt works; effectively raising terrain where the unresolved terrain is more bumpy (i.e. higher wavenumber).
 
Update, if someone from the mods is reading. The WRF guide says (the most recent one as well) the following (of the gwd_opt=1):

"gravity wave drag and blocking; recommended for all grid sizes; includes the subgrid topography effects gravity wave drag and low-level flow blocking"

I suggest (in the light of the results above) to put a disclaimer about domains < 3 km (even ~ 1km is suspicious), at least with the default 10-m GWD static fields.
 
Okay, an update. The gwd_opt=3 option did not produce the high-wavenumber instabilities, instead it looked more like the simulations with the gwd_opt turned off. Also, ASTER terrain is not the "culprit", but it is definitely the gwd_opt=1 option when turned on for all domains.
But gwd=3 makes better wind in my case.
 
Yes, I have ran simulations with gw_opt=3 and indeed the bias is significantly reduced. It is not completely reduced, but that could be due to the static data not recognizing surface roughness length properly. The measurements were made over a small town and LANDUSE shows it as a forest instead.
 
Yes, I have ran simulations with gw_opt=3 and indeed the bias is significantly reduced. It is not completely reduced, but that could be due to the static data not recognizing surface roughness length properly. The measurements were made over a small town and LANDUSE shows it as a forest instead.
GWD = 3 uses its default gwd3 data of 10m, but when I switched to higher resolution data, the results were worse.
 
There are higher resolution data? I did not look into it too much but I would very much appreciate it if you could link to it; what is the resolution of that data?

Note on the gwd_opt=3 and the 333 m domain - it did not produce this mess as gwd_opt=1 does.

EDIT2: By higher resolution data, I mean the static gravity wave drag fields.
 
Top