Update namelist options for recent changes to Redi and Adams-Bashforth 2

[xylar]

This merge sets config_Redi_min_layers_diag_terms = 0, consistent with E3SM-Project/E3SM#6077.

It also changes the time integrator from split_explicit to split_explicit_ab2 nearly everywhere.

Checklist

• Document (in a comment titled Testing in this PR) any testing that was used to verify the changes

[xylar]

@mark-petersen and @cbegeman, I will test this to make sure all affected tests still work.

Once that's done, I'd appreciate your feedback to make sure we don't want to keep split_explicit (as opposed to split_explicit_ab2) for any tests.

[xylar]

@cbegeman, would it be better to stick with split_explicit for isomip_plus? (Should we explicitly do this for ice_shelf_2d as well?) I think this might be necessary for wetting and drying work you're doing, right?

[cbegeman]

@xylar This change is fine. When we are happy with split_explicit wetting and drying, I will change config_time_integrator='RK4' to config_time_integrator='split_explicit' in namelist.thin_film.forward_and_ssh_adjust. Let's not worry about ice_shelf_2d. I was only planning on making enhancements (incl. adding additional tests) in polaris and not migrating those changes back to compass.

[xylar]

Testing

I ran the pr suite, the gotm test, the two internal_waves default tests, the isomip_plus z-star Ocean0 test, and the 4 sphere_transport tests on Chrysalis with Intel and OpenMPI. All tests ran successfully. I haven't looked at the output carefully or compared with baselines because the changes are expected to be larger than machine precision ("climate changing").

[xylar]

@mark-petersen and @cbegeman, this is ready to review. It's not a high priority but it would be nice to be testing as consistently with E3SM itself as we can be.

[cbegeman]

@mark-petersen Could you weigh in on whether this internal_wave test result is satisfactory? I wouldn't have expected there to be negative values.

AB2

split_explicit

[mark-petersen]

@xylar thanks for changing these tests. I agree that we should change to AB2 in all compass tests, since we made the switch in E3SM.

@cbegeman thanks for noticing the negative relative RPE in the internal wave test case. Note that the vertical scale is a factor of 10 smaller, which means less spurious mixing over time. The negative values are concerning. It means that "anti-mixing" is going on, which would be gradient steepening or grid-scale noise. We should view some horizontal slices of temperature at those negative RPE points and compare the two time step schemes, to see what is going on.

It might be that the timestep needs to be shortened a bit for AB2 relative to split explicit, because there are fewer iterations within each timestep, and we are seeing the result of a numerical instability.

[cbegeman]

@mark-petersen Do you have time to explore the time step sensitivity? If we want to get this PR in soon, we could remove this change from all of the internal_wave tests.

[xylar]

@mark-petersen, I think we're still waiting for your response here.

[mark-petersen]

Sorry I missed this. I'm out for a week, so I think it would be better to remove the flag change from the internal_wave test case.

[mark-petersen]

Please merge after removing changes from internal_wave tests. The AB2 was tested at length for global simulations.

[xylar]

@cbegeman, I switched the internal waves test back to split_explicit. Would you be in a good position to test it again? If things work out, I think we can hopefully get this branch merged.

[cbegeman]

@xylar and @mark-petersen I don't feel the need to test again now that the internal_wave test case changes are removed. The other tests checked out fine when I ran them earlier. If you would really like me to retest I can.

[xylar]

@cbegeman, sure that's fine.

[xylar]

Thanks for the earlier testing and review!