Add neutral atmospheric boundary layer example

[ewquon]

This is based on the shear-driven ABL case from Moeng & Sullivan's 1994 JAS paper -- and subsequently used to develop the NCAR LES SGS scheme (see also #534). This example uses a prescribed surface shear stress (through ustar). Currently, I'm using this to assess flowfield behaviors in the surface layer (e.g., due to the overshoot problem, cf. Brasseur & Wei 2010 Phys. Fluids). Runtime is ~15 min on a T4 GPU.

[navidcy]

The halo values are all 0. Unless it's a FieldTimeSeries loading issue.

I think it's because we take the Average...? So for some reason the output writer doesn't apply the operation in the halo region and save the Average halo?

I'm trying (with 8286c79) to take the derivative before the Average; let's see.

[navidcy]

Now docs get a bunch of newton_div-related errors. Is this related to CliMA/Oceananigans.jl#5140?

cc @glwagner

[giordano]

Is this related to CliMA/Oceananigans.jl#5140?

Most likely yes

[navidcy]

Was it a breaking change?
I don't know how to fix it. Actually newton_div is never called in Breeze so perhaps this is an Oceananigans bug that isn't captured in the Oceananigans CI?

[ewquon]

It's caught by the unit tests

[giordano]

Yeah, it's failing in CliMA/Oceananigans.jl#5491, which is what we're using here 😅

[ewquon]

Hmm, I tried running RICO separately and it failed for the same reason, this is the fix that Claude came up with #631

[ewquon]

Now waiting on JuliaRegistries/General#153245

[navidcy]

The $J_θ$ looks good now. But the last plot doesn't show any difference between total and resolved. Is something just zeros by accident?

[giordano]

https://numericalearth.github.io/BreezeDocumentation/previews/PR615/literated/neutral_atmospheric_boundary_layer/ looking good now?

[navidcy]

Nice job!