Conformal octagon: An attractive framework for global models offering quasi-uniform regional enhancement of resolution

Summary With the increasingly widespread adoption of massively parallel processing (MPP) computers for applications in computational fluid dynamics it becomes appropriate to reconsider the geometrical configuration of the computational grid that best suits the problem. In the case of global numerical weather prediction we have recently advocated a conformal spherical-cubic geometry. Among its merits, this grid lends itself naturally to simple domain-decomposition and obviates the need for polar filtering.Here we extend the same principles, but with an emphasis on the problem of regional forecasting. In this case we observe that it is possible to cover the global domain with a conformal grid geometry based on the mapping to the sphere of a back-to-back pair of octagonal regions. In the most symmetrical case, each octagon maps to a hemisphere. By compounding this mapping with a nonhomogeneous conformal mapping of the sphere to itself, one can also arrange to have quasi-uniform enhanced resolution of the resulting grid inside any chosen circle on the sphere, at the expense of relatively coarse resolution degrading gradually with distance outside the circle of interest.With appropriate grid dimensions, the new conformal octagon decomposes naturally into several identical square subdomains for efficient distribution over the nodes of an MPP computer.With 11 Figures