Application of Parallel Processing to Numerical Weather Prediction

Abstract

The purpose of this study is to illustrate the application of a parallel network processing computing system to an important class of problems in hydrodynamics. The computing system selected for this study is a prototype of the SOLOMON parallel processing system (cited as SOLOMON II) which was developed at the Westinghouse Defense and Space Center, Baltimore, Maryland. Emphasis is placed on the problem of numerical weather prediction mainly because of the large data storage and manipulation required, plus the extensive numerical analysis that is involved. The mathematical basis for numerical weather forecasting lies in the principles of conservation of mass, momentum, and energy. From these principles, research meteorologists have devised equations which express the physical laws governing the atmosphere. These basic equations are nonlinear partial differential equations and are well suited for solution by the SOLOMON II system. The computational method is to replace the region of interest by a grid system and replace the continuous equations by their finite difference approximations. In this representation, the programmer can take maximum advantage of mode control, neighboring connections, variable geometry, and simultaneous operation of a network of processing elements. Another topic discussed is the numerical solution of elliptic partial differential equations by the parallel processing network. It is felt by the authors that a parallel processing system of this type will offer a significant increase in computational speed over that of a sequentially organized computing system in the field of fluid dynamics as well as in other scientific fields.