Add harmonics limiter for entr/detr, set GABLS config

[trontrytel]

closing #2124

[szy21]

Looks good to me. Just want to double check that this filter is approved by @tapios :) Should we start to keep track of the filters we have somewhere?

Btw when I tried a quick hack for setting e_int to be the same it didn't work, but maybe I did something wrong.

[tapios]

I do not particularly like this zeroing out approach, but I'm ok with having it if there is no more principled alternative.

It should suffice to have entrainment go to zero as a -> 0. Did you try an approach that does that? In some experiments, I found that limiting entrainment by multiplying it with 2*harmonic_mean([max(a, 0); max(1-a, 0)])) (while increasing entrainment overall) leads to bounded solutions for a. Could you try this here?

[trontrytel]

I do not particularly like this zeroing out approach, but I'm ok with having it if there is no more principled alternative.

It should suffice to have entrainment go to zero as a -> 0. Did you try an approach that does that? In some experiments, I found that limiting entrainment by multiplying it with 2*harmonic_mean([max(a, 0); max(1-a, 0)])) (while increasing entrainment overall) leads to bounded solutions for a. Could you try this here?

I will try it. But I think that here the issue is not that the area does not go to zero. But rather that the area is close to zero, but the updraft still has different properties than the grid mean

[tapios]

I do not particularly like this zeroing out approach, but I'm ok with having it if there is no more principled alternative.
It should suffice to have entrainment go to zero as a -> 0. Did you try an approach that does that? In some experiments, I found that limiting entrainment by multiplying it with 2*harmonic_mean([max(a, 0); max(1-a, 0)])) (while increasing entrainment overall) leads to bounded solutions for a. Could you try this here?

I will try it. But I think that here the issue is not that the area does not go to zero. But rather that the area is close to zero, but the updraft still has different properties than the grid mean

Is the issue the updraft buoyancy?

[trontrytel]

Is the issue the updraft buoyancy?

The way I understand it is that if a<<1 we are setting the updraft e_tot to be the same as the grid mean. But if the updraft vertical velocity is different that will make internal energy different and will cause things to diverge. I'm not sure what is causing the initial divergence.

Here are the timeseries and final profiles from GABLS without the filter I'm trying to add.

[szy21]

It's probably not related, I just would like to comment here so I won't forget: I think it would be better to change this line to ᶠinterp(ᶜentrʲs.:($$j)[colidx]) * ᶠu₃⁰[colidx], so that we can make sure the entrainment tendency of updraft u_3 is zero when it is the same as grid-mean u_3.

[trontrytel]

In the end I'm:

• reducing the dt to 5s,
• adding another entr/detr option: a constant coefficient but limited by the harmonics mean of (a, 1-a). This is so that we can easily test it for other cases.

This doesn't solve the problem. Our GABLS simulation continues to diverge. But maybe we can merge this setup in, while we try to understand how to make the scheme more stable.

Do we want to keep the harmonics limiter as default?

With harmonics:

Without harmonics:

[szy21]

bors r+

[bors]

Build failed:

• buildkite/climaatmos-ci

[trontrytel]

bors r+

[bors]

Build failed:

• buildkite/climaatmos-ci

[trontrytel]

bors r+

[bors]

Build failed:

• buildkite/climaatmos-ci

[trontrytel]

bors r+

[bors]

Build failed:

• buildkite/climaatmos-ci

[szy21]

bors r+

[bors]

Build failed:

• buildkite/climaatmos-ci

[trontrytel]

bors r+

[bors]

Build failed:

• buildkite/climaatmos-ci

[trontrytel]

bors r+

[bors]

Build failed:

• buildkite/climaatmos-ci

[trontrytel]

bors r+

[bors]

Build failed:

• buildkite/climaatmos-ci

[trontrytel]

bors r+

[bors]

Build succeeded!

The publicly hosted instance of bors-ng is deprecated and will go away soon.

If you want to self-host your own instance, instructions are here.
For more help, visit the forum.

If you want to switch to GitHub's built-in merge queue, visit their help page.

• buildkite/climaatmos-ci
• ci 1.9 - macOS-latest
• ci 1.9 - ubuntu-latest
• docbuild
• format