We are planning to apply eddy-resolving global ocean model to climate simulations on the Earth Simulator, a massively parallel vector computer. The model will be with about 0.1 deg resolution and longitude-latitude grid model will be adopted. However, in longitude-latitude models, meridians converge around the poles and it causes inefficient computation. The computational efficiency is critical in ocean climate simulations because it takes more than thousand years to spin up the global ocean. In order to construct numerically efficient model without deterioration of physical performance, we are exploring both numerical schemes and computational techniques.

We are now trying to introduce a quasi-uniform grid. Currently, we are using quasi-homogeneous cubic grids of Purser and Rancic(1998). This kind of cubic grid has non-orthogonal coordinate and has 8 singular points on the sphere. However, by using B-grid and by employing an adequate definition of metric tensors, we found that the dynamical core of the cubic grid is simple and treats those drawbacks appropriately. In the presentation we will talk about the development of global ocean model in FRSGC. We will discuss finite difference schemes on the cubic grid, treatment of the singular points and will show preliminary results of the cubic grid global model.