Aerosols dispersion modelling using probabilistic particle tracking

Abstract

A method is proposed which can facilitate parallel computations of particle transport in complex environments, such as urban landscapes. A two stage-approach is used, where in the first stage, physical simulations of various aerosol release scenarios are conducted on a high-performance distributed computing facility, such as a Beowulf cluster or a computing grid, and stored in a database as a set of transfer probabilities. In this stage, the method provides a partially decoupled parallel implementation of a tightly coupled physical system. In the second stage, various aerosol release scenarios can be analysed in a timely manner, using obtained probability distributions and a simpler stochastic simulator, which can be executed on a commodity computer, such as a workstation or a laptop. The method presents a possibility of solving the inverse problem of determining the release source from the available deposition data. Using the proposed approach and developed graphical tools, a case of aerosol dispersion in a typical urban landscape has been studied. A considerable speedup of analysis time for different aerosol dispersion scenarios has been demonstrated. The method is appropriate for the development of express risk analysis systems. Copyright © 2006 John Wiley & Sons, Ltd.