Unstable behavior of embedded (AGRIF) zooms
Moving to longer time steps with RK3 (and weakly dissipative external mode) has revealed an unstable behavior of the VORTEX test case. This has been tracked back to the use of a "double open boundary" scheme for the barotropic mode, inspired by Herzfeld and Rizwi (2019). This method basically assumes persistence during the barotropic mode integration of the parent grid, slightly inside the overlapping area with the child grid (the distance from the grid interface is given by "nn_dist_par_bc=7
" parameter, which is given in number of coarse grid points). This prevents from developping potentially ill behaved solutions during parent grid barotropic mode integration, within the overlapping area, and near coastlines in particular (divergence conservation is broken there in the updating stage). Since this requires further investigation (on the minimum distance between barotropic interfaces in particular), I propose to disable this scheme by unsetting the cpp key PARENT_EXT_BDY
in agrif_oce_interp
module.