halo update bugs in ACCESS driver?

[aekiss]

Hi @russfiedler, following on from #68, these also look wrong to me (should be field_type_scalar, otherwise their signs will be flipped on the tripole in halo updates)

cice5/drivers/access/cpl_interface.F90

Lines 865 to 866 in edcfa6f

	call ice_HaloUpdate(um_tmlt, halo_info, field_loc_center,field_type_vector)
	call ice_HaloUpdate(um_bmlt, halo_info, field_loc_center,field_type_vector)

and

cice5/drivers/access/cpl_interface.F90

Lines 869 to 874 in edcfa6f

	call ice_HaloUpdate(um_swflx, halo_info,field_loc_center,field_type_vector)
	call ice_HaloUpdate(um_lwflx, halo_info,field_loc_center,field_type_vector)
	call ice_HaloUpdate(um_shflx, halo_info,field_loc_center,field_type_vector)
	call ice_HaloUpdate(um_press, halo_info,field_loc_center,field_type_vector)
	call ice_HaloUpdate(um_co2, halo_info, field_loc_center, field_type_vector)
	call ice_HaloUpdate(um_wnd, halo_info, field_loc_center, field_type_vector)

[aekiss]

I've made a PR to fix this. @aidanheerdegen does ACCESS-CM2 use this same code?

[aidanheerdegen]

I've made a PR to fix this. @aidanheerdegen does ACCESS-CM2 use this same code?

ACCESS-OM2 uses the auscom driver I believe, and yes I think the access driver is used for ACCESS-CM2. @MartinDix would have a better idea, and would probably be very interested.

[MartinDix]

The CM2 driver has these problems. I hope it's not as catastrophic as it sounds.

I've started a new run with a fix.

[aekiss]

The bug was introduced in Dec 2014 - does that mean it doesn't affect previous versions of ACCESS-CM and ACCESS-ESM?
49b80b6#diff-fc19414c40fff80d4d66454635b302d42cec694887445d8c275eaf2a219f9285

[MartinDix]

A CM2 run with a fixed driver gives identical results !! ??

Does that mean these halos aren't actually used?

[aekiss]

Hm, maybe we got lucky then? Would we ever need to take horizontal finite differences of these scalar fields?

[MartinDix]

I think the fluxes would only ever be used in the column.

Could test by setting the halos of these variables to NaNs.