ACCESS-CM2-025 coupling weights

[wghuneke]

CM2-025 currently uses the same regridding weights as the 1° version, are these appropriate or should they be adapted following OM2-025?

[MartinDix]

Regridding weights used in CM2 at 1 and 0.25 degrees are (with example variables)

evap2d: rmp_um1t_to_cice_CONSERV_DESTAREA.nc
heatflux: rmp_um1t_to_cice_CONSERV_FRACNNEI.nc
taux: rmp_um1u_to_cice_BILINEAR.nc
tauy: rmp_um1v_to_cice_BILINEAR.nc
ocn_sst rmp_cice_to_um1t_CONSERV_FRACNNEI.nc
sunocean: rmp_cice_to_um1u_CONSERV_FRACNNEI.nc
svnocean: rmp_cice_to_um1v_CONSERV_FRACNNEI.nc

These are all generated by OASIS in an initial model run (scaling is poor and it takes ~ 2hours for the 0.25 degree version).

[wghuneke]

For OM2, flux fields are remapped using conservative interpolation (first-order for OM2-01, second-order for OM2-025 and OM2-1). Scalar fields are remapped using patch interpolation.

A discussion on which order to use for the conservative interpolation for OM2 is here.

[wghuneke]

Looks like the weights in CM2-025 are not ideal, the wind stress curl shows an imprint of the coarser atmospheric grid. The figure shows the coupled model on top and the ocean-only model at the bottom for a region somewhere in the Southern Ocean.

Below for comparison the wind stress fields (tau) without taking the gradient:

[aekiss]

ACCESS-OM2-025 uses ESMF 1st order conservative for all fluxes (swflx, lwflx, rainfall, snowfall) and ESMF's "patch" method for everything other than runoff (pressure, air temp, humidity, wind components) - see https://github.com/COSIMA/025deg_jra55_iaf/blob/master/namcouple. Runoff is handled separately via YATM in ACCESS-OM2, but should be remapped conservatively in a coupled model. Likewise for iceberg calving.

We needed to use smooth interpolation (ESMF's "patch" method, which is smoother than bilinear) for non-fluxes to avoid the coarse atmospheric fields being blocky in the ocean/sea ice model grid - particularly important for wind, since 1st order conservative creates spikes in the wind stress curl, and therefore Ekman pumping - one symptom is stripes in sea ice: see COSIMA/access-om2#52

The GMD paper and comment above are out of data - we no longer use ESMF's 2nd order conservative interpolation for fluxes in any ACCESS-OM2 config because it can go negative due to spurious overshoots so is not suitable for fields that are sign-definite: COSIMA/access-om2#74 (comment). ESMF are working on a monotonic 2nd-order conservative interpolation (see last slide here), but AFAIK it isn't available yet. So we're stuck with 1st order conservative, which imposes blocky forcing on the higher-resolution model components.

There are some notes on using ESMF to create remapping weights here https://github.com/COSIMA/access-om2/wiki/Tutorials#3-changing-the-oasis-remapping-files and here https://github.com/COSIMA/access-om2/wiki/Technical-documentation#creating-remapping-weights. We use ESMF rather than OASIS because it is much faster at high resolution, and also more accurate.

[aekiss]

ACCESS-OM2 remapping weights are in /g/data/ik11/inputs/access-om2/input_20201102/common_*deg_jra55 (rmp_jra55_cice_2nd_conserve.nc are not used)

[wghuneke]

@MartinDix updated the interpolation weights in CM2-025, a 6 month test run shows the improvements on the wind stress curl field. You can still slightly see an imprint of the coarser atmospheric grid, but much less.

[aekiss]

Looks good - @MartinDix is this using ESMF's "patch" method?

[MartinDix]

Suite to calculate the coupling weights and ancillaries is u-cx958. This uses a modified version of make_remap_weights from the OM2 tools that adds support for the CM2 grid https://github.com/MartinDix/access-om2/tree/CM2_grid

CM2 suite u-cz789 uses the ESMF coupling weights with the patch method for the wind stress. For all other fields the ESMF weights are almost identical to the original OASIS ones.

This suite also uses the 20230515 version of the updated 0.25 ocean topography COSIMA/access-om2#265 (comment)

[access-hive-bot]

This issue has been mentioned on ACCESS Hive Community Forum. There might be relevant details there:

https://forum.access-hive.org.au/t/access-cm2-025-storage-request/1649/1