A Reduced Grid Method For A Parallel Global Ocean General Circulation Model

A Reduced Grid Method for a Parallel Global Ocean General Circulation Model

A Reduced Grid Method for a Parallel Global Ocean General Circulation Model

A limitation of many explicit finite-difference global climate models is the timestep restriction caused by the decrease in cell size associated with the convergence of meridians near the poles. A computational grid in which the number of cells in the longitudinal direction is reduced toward high-latitudes, keeping the longitudinal width of the resulting cells as uniform as possible and increasing the allowable timestep, is applied to a three-dimensional primitive equation ocean-climate model. This ''reduced'' grid consists of subgrids which interact at interfaces along their northern and southern boundaries, where the resolution changes by a factor of three. Algorithms are developed to extend the finite difference techniques to this interface, focusing on the conservation required to perform long time integrations, while preserving the staggered spatial arrangement of variables and the numerics used on subgrids. The reduced grid eliminates the common alternative of filtering high-frequency modes from the solution at high-latitudes to allow a larger timestep and reduces execution time per model step by roughly 20 percent. The reduced grid model is implemented for parallel computer architectures with two-dimensional domain decomposition and message passing, with speedup results comparable to those of the original model. Both idealized and realistic model runs are presented to show the effect of the interface numerics on the model solution. First, a rectangular, mid-latitude, at-bottomed basin with vertical walls at the boundaries is driven only by surface wind stress to compare three resolutions of the standard grid to reduced grid cases which use various interface conditions. Next, a similar basin with wind stress, heat, and fresh water forcing is used to compare the results of a reduced grid with those of a standard grid result while exercising the full set of model equations. Finally, global model runs, with topography, forcing, and physical parameters similar to those used for ocean-climate studies, are advanced to a near equilibrium state for both the reduced grid and the standard grid. Differences between the two are presented for typical fields of interest, and very little degradation of the solution due to the reduced grid is observed.

Parallel Computational Fluid Dynamics 2004

Parallel CFD 2004, the sixteenth international conference on Parallel Computational Fluid Dynamics and other modern scientific domains, has been held since May 24th till May 27th, 2004 in Las Palmas de Gran Canaria, Spain. The specialized, high-level Parallel CFD conferences are organised on travelling locations all over the world, yearly because of multidisciplinary subject of parallel CFD and its rapidly evolving nature.The conference featured 8 invited lectures, 3 Mini Symposia, contributed papers and one Tutorial & Short Course. More than 80 multi-disciplinary presentations of the Parallel CFD had been presented, with participants from 17 countries. The sessions involved contributed papers on many diverse subjects including turbulence, complex flows, unstructured and adaptive grids, industrial applications, developments in software tools and environments as parallel optimization tools. This Book presents an up-to-date overview of the state of the art in parallel computational fluid dynamics.- Report on current research in the field.- Researchers around the world are included.- Subject is important to all interested in solving large fluid dynamics problems.- It is of interest to researchers in computer science, engineering and physical sciences.- It is an interdisciplinary activity. Contributions include scientists with a variety of backgrounds.- It is an area which is rapidly changing.