Improved representation of topographic effects by a vertical adaptive grid in vector-ocean-model (VOM). Part I: Generation of adaptive grids

Abstract

This is the first part of a publication that describes the generation of adaptive grids (this part), and simulations with vector-ocean-model (VOM) in unstructured grids resulting from the adaptation (part II). A static vertical adaptive grid in z-coordinates allows improving the approximation of topography and vertical resolution at slopes. Adaptive grids use elements from a set of grid sizes by multiplying a basic smallest cell size with powers of two, as in cell division. Grids with locally isotropic vertical resolution at surface, seabed, and slopes can be generated whereby resolution decreases towards the ocean interior. The adaptation to topography yields unstructured grids that are organised in a one-dimensional vector by column-wise storage of cells, discarding land cells. The vector storage suggested the model’s name. Grids are generated by an iterative procedure that relies on rules, i.e. criteria and directives to control the grid structure in favour of a good representation of physics and smooth numerical operations. The directives govern vertical resolution at sea surface and seabed, and at slopes. For the latter vertical resolution is extended in the horizontal. In the ocean interior horizontal distances between changes in grid size can be controlled for the sake of smooth numerics. The use of a z-grid that avoids transformation errors, the depth-independence of vertical resolution, and the lateral extension of vertical resolution at slopes towards the ocean interior are the most significant differences of adaptive grids in comparison to vertical coordinate transformations. Unstructured grids do not rely on a smoothing of topography and can be used within any of the horizontal Arakawa-grids. For the same topography directives allow creating various grids as demonstrated for a shelf-ocean topography. The number of cells per column in two unstructured grids generated for the North Atlantic may locally well exceed typical layer numbers in conventional model matrices. But the domain average is similar to layer numbers of today’s ocean models. Thus, with the same investment of cells per domain a higher resolution in slope regions can be achieved by unstructured grids as compared to conventional z-grids.