Algorithm for reduced grid generation on a sphere for a global finite-difference atmospheric model

A reduced latitude-longitude grid is a modified version of a uniform spherical grid in which the number of longitudinal grid points is not fixed but depends on latitude. A method for constructing a reduced grid for a global finite-difference semi-Lagrangian atmospheric model is discussed. The key idea behind the algorithm is to generate a one-dimensional latitude grid and then to find a reduced grid that not only has a prescribed resolution structure and an admissible cell shape distortion but also minimizes a certain functional. The functional is specified as the rms interpolation error of an analytically defined function. In this way, the interpolation error, which is a major one in finite-difference semi-Lagrangian models, is taken into account. The potential of the proposed approach is demonstrated as applied to the advection equation on a sphere, which is numerically solved with various velocity fields on constructed reduced grids.