When selecting one of the ‘nn_ahm_ijk_t’ option to supply 2D or 3D fields for ahm{t,f}, should the values supplied for ‘t’ and ‘f’ be the same (if so why require two variables?), or are there cases when they can or should vary? For example, at coarse resolutions with an equatorial refinement?
I know the {t,f} refer to the point on the cell (either centre or corner), and looking at the code (e.g. DYN/dynldf_lap_blp_lf.F90) it seems that ahmf is used for calculating the curl (or shear) terms, and ahmt is used for calculating the divergence (or tension) terms. However, I don’t have a good understanding of what this means, and whether there is good reason to use different values for ahmt and ahmf.
Formally, the identity used for the vector Laplacian (i.e. the split using the divergence and vorticity that you picked up) is valid for horizontally constant coefficients. This said, chosing different values for divergence and vorticity is possible, and required anyway from the aforementioned discretization. What matters at the end, and pleasingly this does not depend on the values you choose at t/f points, is that these terms will be individually (hence globally) energy dissipating.
Practically speaking divergence/vorticity dissipation, if spatially varying, do generally follow the same rule, i.e. you could interpolate both from a given viscosity field given at T-points.
chosing different values for divergence and vorticity is possible, and required anyway from the aforementioned discretization.
By this, are you thinking/referring to latitudinal (or grid size) dependent scaling, where there will be small differences between values at the t,f points on the same cell?
I have a case where the ‘eddy_viscosity_3D.nc’ file has reduced values of ahmt for the equatorial refinement in eORCA1, but ahmf remains constant globally. There is literature which describes the benefit of anisotropic viscosity at the equator (e.g. Jochum et al. 2008 and references therein), but I couldn’t find anything to support an ‘anisotropy’ between the divergence (ahmt) and curl (ahmf) as we have in NEMO. I’m wondering whether the two fields ought to be set the same, or is the choice of different values a way to enable better representation of the strong zonal equatorial currents and countercurrents.