The stability of passively nested ocean general circulation models

The nesting of one ocean model inside another is a developing technique for the study of detailed local circulation which is dependent on the larger scale flow. Coupled versions of such models are often not possible because of the computational cost so we investigate the passive behaviour of a nested model forced at its open boundaries by an outer model. Our actual example is a limited area model of the Vema Channel region of the South Atlantic, forced by the Fine Resolution Antarctic Model. We show that simulations much in excess of a year are likely to fail, or drift excessively, unless all parameters in the inner and outer models are the same. The best nested simulation possible still drifts significantly From the outer model, even with identical physics, grid-sizes and bathymetry, because of the modification of the equations of motion required at the boundary and the near impossibility of using actual outer model boundary forcing at each timestep. Choosing open boundaries where flow tends to be cross-boundary rather than parallel to it is likely to reduce numerical problems. The initialisation conditions may also lead to numerical instability. Particularly energetic periods in the seasonal cycle should therefore be avoided as initial states.