
Parallelization of the Lagrangian Particle Dispersion Model
Author(s) -
R. Buckley,
B.L. O`Steen
Publication year - 1997
Language(s) - English
Resource type - Reports
DOI - 10.2172/578683
Subject(s) - message passing interface , supercomputer , parallel computing , lagrangian , computer science , computational science , series (stratigraphy) , atmospheric dispersion modeling , particle (ecology) , code (set theory) , atmosphere (unit) , interface (matter) , dispersion (optics) , turbulence , lagrangian particle tracking , message passing , particle number , mathematics , meteorology , physics , optics , chemistry , programming language , geology , air pollution , oceanography , maximum bubble pressure method , paleontology , plasma , quantum mechanics , bubble , organic chemistry , set (abstract data type)
An advanced stochastic Lagrangian Particle Dispersion Model (LPDM) is used by the Atmospheric Technologies Group (ATG) to simulate contaminant transport. The model uses time-dependent three-dimensional fields of wind and turbulence to determine the location of individual particles released into the atmosphere. This report describes modifications to LPDM using the Message Passing Interface (MPI) which allows for execution in a parallel configuration on the Cray Supercomputer facility at the SRS. Use of a parallel version allows for many more particles to be released in a given simulation, with little or no increase in computational time. This significantly lowers (greater than an order of magnitude) the minimum resolvable concentration levels without ad hoc averaging schemes and/or without reducing spatial resolution. The general changes made to LPDM are discussed and a series of tests are performed comparing the serial (single processor) and parallel versions of the code