When I activate 2-way nesting, erroneous horizontal circulation (appearing in the time-mean) appears at certain parts of the boundary of my child domain as reported by Vincent, as well greatly (100x) amplified surface-intensified vertical velocities across the area the child domain occupies within the parent grid and a serious SSH discontinuity across the parent-child interface. The vertical velocities and SSH anomalies do not appear within the child grid itself (i.e. croco_his.nc.1), only within the area occupied by the child domain in the parent grid. This issue appears to be entirely related to 2-way nesting being activated - it is not affected by bottom-friction, timestep, resolution, boundary sponging, or initial conditions, and is not present when 1-way nesting is active. What perplexes me is that such a significant SSH discontinuity, if actually in the model, should generate geostrophic currents - but if you ignore the child domain within the parent grid, the solution looks perfect and continuous.
The fix suggested in the other thread (defining AGRIF_UPDATE_DECAL) removes the erroneous horizontal currents at the boundary, but it does not solve the vertical velocities/SSH. So my two questions are as follows:
-
Is it possible that there’s a bug dumping incorrect SSH/w fields for the area occupied by the child domain in the parent grid?
-
Or, are perturbations within the child domain in the parent grid unable to propagate across the child-parent domain interface (and if so, why?)
Figures:
Zeta:
W: (scale goes up to 1E-4 m/s, actual vertical velocities in the child-domain can exceed 1E-2m/s - despite this, the model is stable and returns sensible solutions for other parameters)
