Use deep atmosphere equations by default

[akshaysridhar]

Draft PR simplifies result tracking for tests with deep-atmos equations as part of the default configuration.
Target config: 70km depth, 99 vertical centers (100 face levels).

[szy21]

Let's have one shallow atmosphere example in ci (you can choose anything, e.g. baroclinic wave).

[charleskawczynski]

I think we may want to write a few more unit tests in ClimaCore to show that this is working for our finite difference stencils. In particular, I think it may be helpful to show that our FD operators are conservative. Thoughts?

[szy21]

I think we may want to write a few more unit tests in ClimaCore to show that this is working for our finite difference stencils. In particular, I think it may be helpful to show that our FD operators are conservative. Thoughts?

What unit tests do we have now that is only for shallow atmosphere?

[charleskawczynski]

What unit tests do we have now that is only for shallow atmosphere?

I don't know that we have conservation tests in a deep atmosphere

[szy21]

What unit tests do we have now that is only for shallow atmosphere?

I don't know that we have conservation tests in a deep atmosphere

We should add them if we don't have them. Could you point me to the tests for shallow atmosphere in ClimaCore?

[charleskawczynski]

Could you point me to the tests for shallow atmosphere in ClimaCore?

That's a good question, I don't actually see any at first glance. I only some conservation tests for horizontal limiters.

We should probably add some, I think we can spin off a test from one of the column heat/wave examples without too much work.

[szy21]

I'm a bit concerned about the two orders of magnitude increase in conservation error in sphere_baroclinic_wave_conservation_source. This test is not exactly accurate, but I don't expect such a difference between shallow and deep atmosphere. Maybe check if sum is implemented correctly for deep atmosphere?

[szy21]

Also this one aquaplanet_rhoe_equil_clearsky_tvinsol_0M_slabocean_ft64 has several orders of magnitude increase in the conservation error. This test should conserve within floating point error (see the current main build).