Many aquaplanet simulations are broken

[szy21]

See this build: https://buildkite.com/clima/climaatmos-gpulongruns/builds/242

Recent major changes are

• using SSP + CAM_SE (Add SSPKnoth and make it default #3019)
• removing filter in radiation (remove filter for radiative fluxes #3099)
• updating to RRTMGP v0.15.0 (Update RRTMGP #3098)

Things to try

• Test atmos v0.26.0 (the differences between this version and main are radiation filter and rrtmgp version): https://buildkite.com/clima/climaatmos-gpulongruns/builds/244
• Use ARS and/or ClimaHyperdiffusion
• Use RRTMGP main branch which contains the clamping of tau: https://buildkite.com/clima/climaatmos-gpulongruns/builds/243
• Add back the radiation filter

[szy21]

Here is a summary.

• The topography runs (longrun_aquaplanet_rhoe_equil_55km_nz63_clearsky_0M_earth and longrun_aquaplanet_rhoe_equil_55km_nz63_clearsky_0M_earth_sleve) are broken because of the timestepper change. They were broken before too, but run for longer. Maybe we should revisit it after the topography work.
• The aquaplanet 1M (longrun_aquaplanet_clearsky_1M) breaks faster with the rrtmgp version update (v0.15.0).
• The latest rrtmgp version (with clamping of tau) fixes the aquaplanet 0M runs without topography (longrun_aquaplanet_rhoe_equil_55km_nz63_clearsky_tvinsol_0M_slabocean and amip_target_diagedmf).
• However, the change in timestepper breaks the conservation test in longrun_aquaplanet_rhoe_equil_55km_nz63_clearsky_tvinsol_0M_slabocean. There is a ~20% change in total water. We should look into it. I will open a separate issue.

[Sbozzolo]

Deep atmos passing here: https://buildkite.com/clima/climaatmos-gpulongruns/builds/248#0190363d-b25a-4199-92fc-b3150c045db9

[szy21]

I think we can close this now. The rrtmgp update has been merged, and I opened an issue for the aquaplanet conservation. Topography work will be tracked somewhere else.