The effect of vertical coordinates on the accuracy of a shelf sea model

The vertical coordinates (VC) are one of the most important set of configuration options of an ocean model. Optimisation is, however, a non-trivial exercise. We compare nine configurations to investigate different VC options and contrast the Vanishing Quasi-Sigma (VQS), partial step z-level, s-z hybrid and Multi-Envelope (MEs) approaches. Using NEMO model simulations, a hierarchy of experiments are conducted, including: unforced simulations, multi-year climatological simulations with comparisons against tracer profile observations, and tide-only simulations. Hydrostatic pressure gradient errors on the continental slope in the VQS coordinates are found to be consistent with reduced domain-averaged accuracy in both unforced and realistic simulations. Reduced accuracy on the continental shelf is associated with larger advective tracer transports at the shelfbreak. Accuracy is improved by using separate definitions of the computational surfaces on the shelf and slope using the MEs and s-z hybridisation approaches. MEs configurations employing VQS on the continental slope with a computational slope steepness parameter, rmax, of 0.04–0.07, perform comparably with s-z hybrid configurations. Restrictions on the tilt of computational surfaces on the shelf and upper slope appear less important. In contrast, tide-only experiments without stratification show that tidal simulation quality is linked with accurately representing the shelf bathymetry, which favours terrain-following systems. The experiments support transitioning the vertical coordinates across the shelfbreak using either a MEs or hybrid s-z approach as a flexible route to improving accuracy in regional and global models.