Non-uniform values on refinement zones

[marybau]

Hi everyone,

When running ABL simulations with nested grids, we have noticed some strange and rather large "instabilities/differences" in zones where we refined the grid. It happens with neutral, stable and ustable. It is more visible for the mean velocity and mean tke fields; the instantaneous fields seem to be much less impacted. Could it be the averaging?

Here are some screenshots of precursor simulations, neutral and unstable stratification, with a refinement region at the center of the domain, at the last time step of the simulation.

We tried fixing the "ABL.log_law_height" value on the input file. This seems to reduce the differences, but it was not clear it fix it.

Have you noticed something like this? Is it normal ?

We are using AMR-Wind v3.1.4

Thank you very much

abl_unstable_input.txt

[marchdf]

Thank you for bringing this question to us. This is indeed odd and not something I've seen in our runs. For example, figure 21 b here is a time averaged flow field with refinement zones and there are no such artifacts.

A couple of things I noticed in your input file:

• you have a fixed dt, are you getting any warnings about CFL violations in the log file?
• you are using averaging.type = TimeAveraging, is this what you are plotting in the figures above? I haven't actually used that so I wonder if maybe there's an issue there? If you took the instantaneous slices and averaged them yourself, would you see those artifacts?

[marybau]

Hi Marc, thanks for your help.

• The maximum CFL that I get is 0.28. So it should be good.
• In the screenshots, I am showing the mean values calculated by averaging.type = TimeAveraging. You always average the values yourself ? I will do the average myself and post the results here.

I also noticed that the "non-uniform values" happen close to the wall. The height of my refined region is 500m, but at a height of around 100m, there does not seem to be a difference in the mean velocity fields between the refinement regions. But the tke mean still shows differences up to 700 m.

[hgopalan]

We can expect some difference in the mean velocity fields due to the refinement regions since there is a change in the level of modeled and resolved stresses. This effect tends to be maximum at the surface as the wall shear stress is entirely modeled. I have enclosed the windspeed plots from my testing. I did not use the averaging capability but manually averaged.

One-Equation Model

nested_deardorff.pdf

Non-Linear Backscatter Anisotropy Model

nested_kosovic.pdf

Hybrid Non-Linear Model

nested_hybrid.pdf

The nested simulations had three levels of nests: 64 X 64 X 16, 32 X 32 X8 and 16 X 16 X 4. I sampled a vertical profile within the 16 X 16X 4 region.

[github-actions]

This issue is stale because it has been open 30 days with no activity.

[moprak-nrel]

Hi @marybau, following up to see if you were able to confirm that Marc and Harish's comments were sufficient to explain the discrepancy.

[marybau]

Hi @moprak-nrel, thanks. We had some discussions offline and we found some other way around to avoid this issue. For example: modify the refinement of the mesh, or use a newer version of AMR-Wind with a different wall functions, ... But it is still not clear for me why this discrepancy happens, we cannot explain it. We are still working on it. I will post here the solution as soon as we figure it out.

[moprak-nrel]

Thanks for digging into this, please let us know if there is anything we can do on our end to help out. I would definitely encourage using the newest release (unless you have some strong constraints keeping you at 3.1.4).

[github-actions]

This issue is stale because it has been open 30 days with no activity.