Applying terrain_hi terrain_lo to refinements

[ews-ffarella]

I am still wrapping my head around AMR but I was thinking. Isn't it sometimes interesting to only want to refine a bit away from the terrain? Like high TKE values or vorticicity. Close to the terrain, these values will always be high. Can we introduce a terrain_hi / terrain_lo restriction. Like if I only want to refine above 20m until 300m if grad_vel is above whatever.

Does this make sense?

It would be very fast as the terrain height is already computed

[hgopalan]

I usually apply field refinement just above where the met mast is and a cylindrical refinement region around met mast. I have not validated sufficient complex terrain cases to say if this is sufficient.

[ews-ffarella]

Thank you. I have spent the whole week trying to have my case running, but somehow it still take a lot of time to converge.
I have read that you have used RANS on a massively complex Reunion island case. I cannot wrap my head around the fact that 600s would be enough. In my case, it takes a lot more. Would it be possible that you provide your inp files?

When you say field refinement, you mean based on velocity gradient / vorticity?

I have increase the Abl and Rayleigh damping to 2x the elevation difference not to constrain too much the bl development, and also increase the Rayleigh relaxation to 50s.
I will have a coarser mesh rub overnight.

[hgopalan]

Possibly I need to expand more on this. When the upstream region of point of interest is relatively smooth it should converge relatively fast. However, if the upstream region has topography it may take a lot of time to converge. The RANS model is relatively new in the code and has not been extensively tested to see how long it takes to fully converge. 600 seconds was something for my testing and is not a guideline. So I am not surprised it takes a longer time for you. We should hopefully have more guidance on typical time take for convergence in future when we run other complex sites in US.

[ews-ffarella]

It makes total sense! I ll happily test complex cases. I do have a lot!

[hgopalan]

Do add comments in future if you find anything interesting. Will close it for now.