Imprint of atmospheric grid visible on some surface sea ice fields

[njeffery]

A checkboard pattern is visible in monthly climatologies of melt-pond fraction for ne30 atm grids and EC30to60E2r2.

[rljacob]

Can you include a figure showing it?

[njeffery]

[njeffery]

The above figure is from @golaz 's water-cycle spinup, average of about 100 years.

[whannah1]

@njeffery are you able to plot the atmospheric fields on the native grid? If not then can you point me to where the data is? I'd like to check a few fields. This might be the same issue that I documented here: https://acme-climate.atlassian.net/wiki/spaces/NGDAP/pages/2953512527/Checkerboard+noise+problem+related+to+the+dCAPE+trigger

[rljacob]

and the imprint is only in that field? no others?

[njeffery]

I'm looking at some of the other fields now. I suspect it'll show up in albedo, shortwave surfaces fluxes, that sort of thing.

I looked at an RRM-EC30to60E2r2 run that @jonbob had done and the checkerboard disappeared in the Arctic (still present in the Southern Ocean).

Chris's run is archived on NERSC. I have a comparable run on anvil that shows the same pattern. Here: /lcrc/group/e3sm/ac.jeffery/E3SMv2/v2.snow.snowRedisOnly/run

[njeffery]

We made a change in v2 in how ponds form in response to snow melt. There is a namelist flag that is intended to turn off that feature. It's not coded properly at the moment but I'll change that soon and submit a PR. With the snowmelt-to-pond flag false, the checkerboard disappears. Here's a figure from a comparable run:

[rljacob]

Interesting. Would still like to know if shows up in other fields (and if the flag matters for those).

[whannah1]

FYI I looked at a few fields from the run on LCRC that @njeffery pointed to. I mainly just plotted 1 or 10 year averages, but I'm not seeing any detectable checkerboard over the poles. Here's an example of PRECT, but I just a similarly smooth looking field for liquid water path, which is usually the best field to detect the checkerboard problem that I've looked into previously.

[njeffery]

Here are some other fields in MPAS-SI that see imprinting:

[proteanplanet]

Here is another example of atmospheric mesh imprinting on the sea ice and ocean models, this time for the coupler field o2x_So_dhdx passing from ocean to ice:

In the coupler field of sea surface height being passed from the ocean model to the sea ice model, the atmospheric mesh is imprinted on the gradient height field, meaning it has (probably) derived from the atmospheric turbulent flux fields. This is for the WC14 ocean-ice mesh coupled with the standard-resolution atmosphere. One would expect some imprinting with this resolution discrepancy, but the fact it occurs mainly in the Canadian Archipelago and Greenland coastal regions suggests land/atmosphere coupling and/or coastal weighting are somehow involved.

The area around Greenland is zoomed for convenience here for dh/dy:

[rljacob]

Are the areas the imprinting is seen also under sea ice? What controls sea surface height under sea ice?

[proteanplanet]

For completeness, I'm going to post a compressed file here of every relevant coupled plotted flux and state field to the ice and ocean so that you can step through them. But here is the summary so far: Aside from the (dh/dx,dh/dy) o2x gradient field, which results, in winter, from ice-ocean stress, in turn driven mainly by the wind stress (hence affected by the surface radiation balance), the main imprint appears to come from precipitation and radiation, for which I'm sharing x2i snow for the northern hemisphere, and x2i long wave up (which is usually balanced by longwave down).

It'll take me 24 hours to complete the generation of fields, but I've already looked through most of them.

[proteanplanet]

Please see attached important atmospheric coupled fields used to calculate energy fluxes when there is little to no sunlight affecting the northern hemisphere pack. These results are very solid, because they are an average of 500 years form the PI Controls from v2 standard resolution and NARRM:

July_19_Ice_Group_Brief_E3SM.pdf

The problem appears to be temperature related, since the V2 aliasing is apparent in sensible heat, but only mildly in latent heat, and not in the wind field. It is evident in both the NARRM and SR simulations. I'm attaching the PNG here of the latent heat plot shown above, so that you can zoom into that particular figure to see the alias even at high resolution.

There is no aliasing as such in the surface longwave down, although the atmospheric grid is definitely noticable, which is not unexpected:

These fields are as seen from inside the sea ice model.

[proteanplanet]

Addendum: All of the above figures are plotted on the native mesh (no interpolation), and therefore are different from results that would appear in MPAS-Analysis.

[eclare108213]

@proteanplanet @njeffery Has this issue improved in the new version?

[proteanplanet]

@eclare108213: Yes. We are now using non-linear remapping between the ocean/ice and atmosphere, and as a result the imprint of the atmospheric mesh is reduced, and has allowed us to discover other issues associated with the interaction of Redi with ice shelves that previously were aliased-out by linear remapping. Here is an example of sea ice production around the Antarctic, with the non-linear and linear maps, respectively, for sea ice production. There are other differences between these simulations, but the figures highlight the impact of non-linear remapping:

[proteanplanet]

I suggest this issue could be closed and marked resolved, with thanks to @ambrad for the non-linear remappping.

[njeffery]

Much improved with new non-linear remapping between components.