Gridscale noise in regions where the flow should be relatively quiescent might be an indicator of this type of instability; further testing in other GCMs is necessary to check whether this is true. The unphysical mixing effect occurred in both the nonhydrostatic solver and the MITgcm, and as such the authors consider it a general numerical issue that may arise when using anisotropic HIF inhibitor viscosity. To explore further, another set of five simulations was run with an isotropic grid (Δx=Δz=1Δx=Δz=1 m)

and stratification parameters as in Taylor and Ferrari (2009), except that the horizontal viscosity and diffusivity were set to νh=κh=10-4,10-3,10-2,10-1,1νh=κh=10-4,10-3,10-2,10-1,1 m2 s−1. This configuration was chosen because in their original paper Taylor and Ferrari (2009) used isotropic viscosity and diffusivity with νh=κh=10-4νh=κh=10-4 m2 s−1 on an isotropic grid, and obtained full restratification to q=0q=0 and Ri=1Ri=1. The linear stability calculator predicts that full restratification would also be achieved for any choice of νhνh in the set above. Therefore, if the vertical viscosity is held fixed at νv=10-4νv=10-4 m2 s−1 and the horizontal viscosity is increased, any overshoot in either Ri or q can be attributed to anisotropic

viscosity. Indeed, Fig. 7 demonstrates a progressively Selleck INK 128 larger overshoot in both Ri   and q  , as well as more energetic inertial oscillations, as νhνh is increased away from νv=10-4νv=10-4 m2 s−1. These results suggest that the use of anisotropic viscosity is at least partly responsible for the excessive restratification, though this effect does seem to be amplified as the grid aspect ratio Δz/ΔxΔz/Δx becomes smaller ( Fig. 5(b)). The converse scenario (isotropic viscosity and anisotropic grid) was not tested due to the prohibitively small timestep it would require – in order to permit SI the vertical viscosity (and thus horizontal viscosity) must be kept very small, which makes modelling of this situation prohibitively expensive. As the stratification of the mixed layer plays a key role

in communicating atmospheric forcing to the interior of the ocean, excessive or improperly represented restratification Astemizole could negatively impact climate prediction on long time scales. Further investigation of this numerical issue is beyond the scope of this paper. To the authors’ knowledge this effect has not been previously documented, but due to the ubiquity of using anisotropic viscosity in GCMs it is possible that this it would occur in non-SI flow regimes as well. In this paper a set of 2D numerical simulations have been conducted to demonstrate how a combination of model viscosity and grid resolution can affect mixed layer restratification by symmetric instability. Linear theory is used to predict the growth and restratification potential of SI modes resolved in the model.