Hello everyone,
I'm trying to reproduce the Southern Ocean reentrant channel configuration of this Abernathey et al. paper using MOM6 as the dynamical core instead of MITgcm.
I'm specifying the exact same horizontal geometry (cartesian 1000km x 2000km) and resolution (5km), and a quasi-identical vertical resolution. It is the same temperature only, linear EOS setting. The surface forcing and sponge restoring are also exactly the same. I've also set the same viscosities/diffusivities as in the paper. I'm using the KPP module to impose a homogeneous 50m MLD.
With all that, I seem unable to satisfyingly reproduce the MOC structure the authors report in that paper. Mainly, the deep "AABW" cell appears considerably weaker (compared to the other two cells) than in the paper, and shows a significant diapycnal return flow in the interior, as one can see in the attached picture (MOC_fixed_MLD.png). Compare with figure 3 of the paper which displays a strong, quasi-isopicnally oriented cell.
The picture MOC_dyn_MLD.png shows the results obtained with a dynamically computed MLD: even though the diapycnal component appears tamed down, the cell is still weak and appears to loop back diapycnally in the interior instead of in the sponge.
Globally, and also in diagreement with the paper, the other two cells also show significant diapycnal flow in the interior: Looking at e.g. the "red" cell, we see that diapycnal flow happens below the diabatic layer as defined in the paper (lower grey line representing the SST 95% quantile ).
In my experimentations, I've seen huge improvements by switching to higher-order remapping and tracer advection schemes (PQM_IH4IH3 instead of PPM_H4 for vertical remapping, PPM instead of PPM:H3 for tracer advection). In the paper, they mention the importance of the Prather advection scheme to reduce spurious diapycnal mixing in the ocean interior.
So, I'm wondering if the advection/remapping schemes implemented in MOM6 would allow recovering the same results, in Z* coordinates, or if the problem potentially comes from somewhere else? I've attached my MOM_parameter_doc.all to this post.
If it is known that the tracer advection schemes implemented in MOM6 are significantly more diffusive than the Prather scheme, do you think switching to Hybrid coordinates could do the trick?
If any other detail would be worth providing, please tell me!
All the best,
Nelson
PS: in the pictures, density is used as the coordinate (left panel) instead of temperature in the paper. The correspondance is exact though since EOS is linear with alpha=-0.2 kg m-3 decC-1. The isotherms (isopycnals) in the right panel are plotted in 0.25 degC intervals (coldest plotted is 0.25 degC).
I'm trying to reproduce the Southern Ocean reentrant channel configuration of this Abernathey et al. paper using MOM6 as the dynamical core instead of MITgcm.
I'm specifying the exact same horizontal geometry (cartesian 1000km x 2000km) and resolution (5km), and a quasi-identical vertical resolution. It is the same temperature only, linear EOS setting. The surface forcing and sponge restoring are also exactly the same. I've also set the same viscosities/diffusivities as in the paper. I'm using the KPP module to impose a homogeneous 50m MLD.
With all that, I seem unable to satisfyingly reproduce the MOC structure the authors report in that paper. Mainly, the deep "AABW" cell appears considerably weaker (compared to the other two cells) than in the paper, and shows a significant diapycnal return flow in the interior, as one can see in the attached picture (MOC_fixed_MLD.png). Compare with figure 3 of the paper which displays a strong, quasi-isopicnally oriented cell.
The picture MOC_dyn_MLD.png shows the results obtained with a dynamically computed MLD: even though the diapycnal component appears tamed down, the cell is still weak and appears to loop back diapycnally in the interior instead of in the sponge.
Globally, and also in diagreement with the paper, the other two cells also show significant diapycnal flow in the interior: Looking at e.g. the "red" cell, we see that diapycnal flow happens below the diabatic layer as defined in the paper (lower grey line representing the SST 95% quantile ).
In my experimentations, I've seen huge improvements by switching to higher-order remapping and tracer advection schemes (PQM_IH4IH3 instead of PPM_H4 for vertical remapping, PPM instead of PPM:H3 for tracer advection). In the paper, they mention the importance of the Prather advection scheme to reduce spurious diapycnal mixing in the ocean interior.
So, I'm wondering if the advection/remapping schemes implemented in MOM6 would allow recovering the same results, in Z* coordinates, or if the problem potentially comes from somewhere else? I've attached my MOM_parameter_doc.all to this post.
If it is known that the tracer advection schemes implemented in MOM6 are significantly more diffusive than the Prather scheme, do you think switching to Hybrid coordinates could do the trick?
If any other detail would be worth providing, please tell me!
All the best,
Nelson
PS: in the pictures, density is used as the coordinate (left panel) instead of temperature in the paper. The correspondance is exact though since EOS is linear with alpha=-0.2 kg m-3 decC-1. The isotherms (isopycnals) in the right panel are plotted in 0.25 degC intervals (coldest plotted is 0.25 degC).
