The role of numerical implementation on open boundary behaviour in limited area ocean models

Applying open boundary conditions to the primitive equations in regional ocean domains presents many challenges. The response of the model domain interior may be sensitive to the choice of open boundary condition (OBC), and the performance of the OBC itself may be dependent on the method of its implementation, the physics of the problem under investigation and the type of model used. There have been many assessments of OBCs in the literature, but little dedicated assessment of model performance with respect to the implementation method of the open boundary has been provided. This study assesses open boundary performance in terms of the cell location where normal velocity is implemented for a variety of test cases. Specifically, a ‘normal velocity outside elevation’ node location (NVOE) was assessed, where normal velocity occupies the out-most boundary cell face, and a ‘normal velocity inside elevation’ node location (NVIE) where normal velocity occupies the cell face interior to the boundary cell centre. These implementations were evaluated for steady wind forcing, barotropic relaxation, forced waves of tidal period and a travelling storm on a simple test domain. Results show that different implementations of the same boundary condition result in different model behaviour. Both the open boundary condition and its implementation should be chosen so as be compatible with the dominant physics of the problem at hand.