Impact of the ''symmetric instability of the computational kind'' on ocean simulations

Recent simulations performed at high resolution with the NEMO ocean model have been severely damaged because of the development of a computational instability. It has been realized that the origin of this instability -and ways to eliminate it- were early explained by Hollingsworth et al 83 : it is specific to a family of discrete schemes for momentum advection formulated in its vector form ; it is due to unbalance at the discrete level between terms which compose momentum advection, i.e the transverse flux of vorticity and the kinetic energy gradient. This computational instability shares some properties with the symmetric instability which is known to develop at small horizontal and vertical scales in the ocean, it has thus been called ''symmetric instability of the computational kind''.

I will use idealized experiments to describe the instability life cycle from the linear stage to its dissipative saturation, and to show the resolution dependency of the instability. I will then report impacts of the instability on realistic ocean simulations. In a global 1/4° resolution simulation, moderate impacts are found in the eastern equatorial Pacific Ocean and in the Agulhas region south of Africa. In a North Atlantic 1/60° resolution simulation, strong impacts are found in the Gulf Stream current system. The instability affects the circulation by driving a spurious and efficient route to dissipation.