MPAS seaice critical error: rescale_mesh: inconsistent radii

[wlin7]

The error occur in a testing run of WCYCL1850 using ne4pg2_oQU480:

CRITICAL ERROR: rescale_mesh: inconsistent radii: sphere_radius: 6371230.00000000, earthRadius: 6371229.00000000

Can the coupled model run at resolution ne4pg2_oQU480 at all?

[xylar]

Hmm, and neither of those seems to be the "correct" earth radius defined in: https://github.com/E3SM-Project/E3SM/blob/a301f2ad8be666beaf9738dc1f6543ed91197930/share/util/shr_const_mod.F90#L20

[jonbob]

@wlin7 - that error happens when the compset and ocn/ice resolution are inconsistent, due to missing spunup initial conditions. For oQU480, we do not have those files created, so the coupled model can run but needs to use WCYCL1850NS instead of WCYCL1850

[wlin7]

@jonbob , nice to know this. Is it convenient to create the spunup ICs for this grid? We would like to use this grid for concept proving coupled model auto-tuning. @xylar , do you feel at this resolution, would the simulation still have reasonable value? If ocean and/or sea ice are totally unrealistic, we would need to try a higher resolution.

[jonbob]

@wlin7 -- I think this resolution is much more meant for testing than anything scientifically valid, which is why we have been hesitant to support it more fully. It's easy enough to use with the WCYCL1850NS compset, so is there a reason you would want spunup initial conditions?

[wlin7]

@jonbob , thanks for pointing this out. If just to get the code run, WCYCL1850NS compset should fit well. But we do want the simulation to be scientifically valid -- at least to some extent. Would a run with spunup ICs be any more meaningful than without? We look to find a working grid to run lots of perturbation simulations, if not oQU480 (or oQU240), is there another more affordable grid than the standard one for the E3SM-LR configuration?

[jonbob]

@wlin7 - I think part of the decision not to have spunup IC's for the test configurations is to dispel the idea that results with them are scientifically useful. You might get relative information for comparison from those toy grids, but nothing meaningful in a scientific sense. As far as I know, the reason the LR configuration has that name because it represents the most affordable resolution that produces scientifically valid results. But that may be a question for the coupled model group -- @golaz? I certainly can spinup initial conditions if we decide they're useful, but I believe there's a reason we have not done so

[proteanplanet]

All, just to make sure it's understood what's being used for the ice-ocean mesh, here are the polar views: You would almost certainly want to switch off sea ice dynamics for such a configuration, and I'm sure the ocean configuration would need attention, too. However, I can see that it may be useful for certain tests.

[wlin7]

All are great points, @jonbob , @proteanplanet . This should be a good topic for today's coupled group meeting (added to the agenda). I also have another thought, if current LR (EC30to60E2r2) is the lowest horizontal resolution for scientifically valid simulations, what if we sacrifice some vertical resolution? I can imagine much more works will be needed to support such a configuration.

[golaz]

@wlin7, @jonbob : I have no first hand experience with resolutions lower than the standard E3SM low-res. My understanding is that ultra low resolutions in the atmosphere were explored as part of a previous PPE but quickly abandoned due to a lack of fidelity.